DOI: 10.31038/MGSPE.2022212

Abstract

We present the results of two Mind Genomics cartographies, each focusing on the discovery of common points of view in areas of political disagreement (discussions about Israel among New York City Jews), and areas of open hostility (discussions about what is needed to stop hostilities and begin a peaceful future enjoyed by Palestinians and Israelis). The cartographies searched for specific messages which drive respondents to feel that these messages ‘will work.’. We suggest that an iterative, inexpensive, rapid set of such cartographies can quickly reveal common mind-sets in mutually argumentative and even openly hostile populations. The promise of such iteration is the discovery of topics of mutual interest and agreement in a manner which is practical, which increases insight and wisdom, and which can produce reference databases about the mind of the antagonist, becoming a guide for productive negotiations. We suggest that the effort to find the ‘right messages’ through extensively reasoned efforts followed by an effort to validate these messages be replaced by a mechanistic, iterative, rapid, inexpensive system requiring just a few hours from start-to-finish for each iteration.

Introduction

The world of people is no stranger to difficult issues. Whether the issues are people vs. the environment (viz., global warming), people vs. economic realities (we all cannot really live equally well), or people vs. people (long standing political disagreements, quick starting personal disagreements), there are always conflicts. And there are no prospects for the conflicts to end, or even to be resolved simply. Efforts as grand as the League of Nations and the United Nations, or as intimate as couples counseling continue to reveal the almost absolute impossibility of imposing harmony in an environment when there are different goals, different viewpoints, different levels of power and influence.

The foregoing is a truism. Anyone reading newspapers, sitting in a political debate, or sitting in a home with adolescent children and their parents know about the need to find ways to have meaningful conversations. The real issue is how to do it. Conversations themselves will not solve problems, but without the conversations the problems will never be solved, except through force. And force will not keep the problem solved when the balance of power shifts, when the anger builds. Conversations will help the parties involved come to an agreement, accepting each other. This paper focuses on two difficult problems, one involving the New York Community of identified Jewish people, and the other involving Palestinians and Israelis. The objective of both studies was to assess whether one apply a method hitherto used by marketers (Mind Genomics) to issues of public discourse, and specifically, divisions about Israel. The first was from the point of view of Jews in the United States, where there are many opinions about Israel, not all positive, not all negative [1-5]. The second was from the point of view of self-described Israelis and Palestinians [6-9]. The two projects were run during the two-year period 2012-2014 to determine what Mind Genomics might contribute.

Moving from Opinion to Experiment

It is obvious to anyone who observes the behavior of people, whether dealing in the worlds of goods or services that people differ from each other. Differences can end up being a simply a nagging reality which gets in the way of efficient activities, whether social, economic, and so forth. Indeed, there is an old French proverb, translated into English as ‘Of taste one does not dispute.’ The smart professional knows these differences exist, the experienced professional knows about the nature of these differences, and the effective professional knows how to work with these differences to achieve objectives. In their minds people live in different worlds; one size does not fit all. People need to hear different things. The smart marketer recognizes these different worlds, often establishing them by one or another research effort such as focus groups, in-depth interviews, surveys, and now more recently observing the behavior of a person over time in terms of what the person searches for or how the person shops on the Internet. The smart marketer then digests this information and feeds back a (presumably) motivating offer to the individual using messages which seem to be in line with what the person has done, said, or even looked at.

How can this knowledge-based approach be used in situations where there is natural disagreement, conflict? Is there a way to deeply understand the mind of people in conflict, whether the conflict is open and mutually aggressive, like war, or simple civil disagreement? In other words, if the knowledge-based approach can be used to sell soap, can it be used to sell or at least to message ‘peace’.

The Mind Genomics Worldview

In the 1960’s mathematical psychologists R. Duncan Luce and Patrick Suppes introduced what then was an esoteric paper, presented in the first issue of the Journal of Mathematical Psychology, and as the first paper of that issue. The title of the paper was daunting: Simultaneous conjoint measurement: A new type of fundamental measurement [10]. The focus of psychologists in the 1960’s was to establish the science of measurement, doing so in what was called the ‘axiomatic measurement’. The elegance was at first mathematical. Over the years, the importance of the thinking behind this new method, conjoint measurement, would overshadow the mathematical elegance. This then-esoteric approach would be a way to deep, often profound knowledge about how people think, knowledge that could be applied in a practical way to problems of everyday life. Conflict is one of those problems.

The basic idea behind conjoint measurement is that one could ‘measure’ ideas by putting them in combinations, getting responses to the combinations, and somehow deconstructing the response to the contribution of the separate ideas. The mathematics of the approach, the various postulates, lemmas, and so forth, are no longer of interest to most people, although remain of great import to mathematical psychologists. What is important is the notion that by measuring responses to combinations which simulate ‘reality’, one can deduce the part-worth contribution of each component of the combination.

The immediate importance of this discovery comes from the realization that the typical approach in science including social science, is to isolate the factors, control everything, and measure the response to those factors. Thus, in a study of conflict, for example, or in a study of tomato sauce in contrast, one might look at a set of single factors in a questionnaire and ask how important the factor is to drive negotiations for conflict or drive acceptance for tomato. The process breaks down for tomato sauce because tomato sauce is meaningless unless the mixture is created. Asking a person ‘how important is sweetness’ in tomato sauce is meaningless because the taste profile and smell and appearance profiles must be correctly balanced. So, when it comes to products, importance is not a meaningful topic. It is the mixture.

Moving now to conflict, it is hard to understand how to rate ‘openness to negotiation’ vs. rating ‘Having evidence about one’s point of view’, in terms of importance. It can be done. Most questions about ‘importance’ and aspects of the topic rely on this one-at-a-time effort. The effort certainly does not work for products, although it may work for more complete descriptions of products. The effort may or may not work for the topic of ‘discussions’, although showing that the one-at-a-time approach does not work for topics of ‘discussions’ and ‘arguments’ is harder to prove because we are dealing with simple ideas which have meaning.as.

Evolving from Compound Mixtures to Systematic Variation – The Contribution of Mind Genomics

Mind Genomics emerged in the early 1980’s, with work done under contract for a number of companies, one of which deserves both acknowledgment and thanks. This was the Colgate Palmolive Company of Canada, and its visionary general manager at the time, the late Mr. Court Shepard. As the general manager, it was Mr. Shepard’s simple objective to increase the company’s sales, rather than to simply do his job well and protect his job in any way he could. Mr. Shepard confessed at a meeting that he did not know what to say about Colgate Dental Cream to increase sales, a statement which led to the discussion of how a ‘different’, and possibly more complex but powerful method might help. The method involved mixing messages together in a systematized way using the statistical discipline of experimental design, presenting the combinations, obtaining ratings to the combinations, deconstructing the ratings to the contribution of the different messages, and then discovering which messages ‘worked’. To his credit, Mr. Shepard said ‘yes’, and within a week the study was run, analyzed, and the results implemented. Sales increased. The lesson from that early experiment, fall, 1980, was that systematics work. The respondent, the person evaluating the test combinations need not be an expert. In fact, what happened was that the respondent became disinterested in the task, and answered almost ‘automatically’, without thinking.

The approach ended up descending deeply into the mind of the respondent, even for such a mundane product as toothpaste or dental cream. What becomes important in this regard is that the results could not be faked. There could emerge a ‘stance,’ driven by one’s conscious beliefs. The mixtures of messages prescribed by experimental design, whether of dental cream or of social issues and feelings, were simply too hard to disentangle at an intellectual level. When the messages comprised three or four different ideas, seemingly thrown together, even the person who wants to ‘fake’ the study cannot figure out what to do to bias the results. One could answer randomly, but that is quickly revealed by a statistical analysis of how well the ratings co-vary with the presence/absence of the elements. In the end, the respondent relaxes, and in a somewhat bored way, reads the vignettes, and ‘guesses’.

To summarize the first part of the Mind Genomics effort emerged from the creation of mixtures of messages (elements) using experimental design, the evaluation of these mixtures by respondents, and then the deconstruction of the responses to the contributory impacts of the separate elements. There is a second part of this, one which is just as important. This second part is the discovery of mind-sets, of groups of people who respond in similar ways to sets of elements. These groups, mind-sets, are not necessarily similar in who they are, or what they buy. They are similar in the pattern of their responses to the elements but may be radically different otherwise. Yet, knowing these mind-sets allows the marketer to tailor the messages to a mind-set. That knowledge would prove invaluable for marketers, because it was simple to change the message in a knowing way to appeal to the mind-set. The discovery seemed to be like having an experienced salesperson involved in every messaging effort, a person who would know ‘what works.’ Only there was no person, just a simple algorithm to create these mind-sets, and then to uncover these mind-sets in new populations, groups who may never even have been encountered before.

Applications of Mind Genomics

The foregoing presented the theory of Mind Genomics. The two experiments presented here show the approach of Mind Genomics to issues of prospective conversations about Israel first, and then the Palestinians and Israel. The objective of the two experiments, both done around 2012, ten years ago, was to identify the common topics of conversation which could heal potential fractures in the relationship of people. The Mind Genomics process follows a set design and analysis approaches, created to produce ‘actionable’ data sets in days, and thus be amenable to continual improvement, at low cost, and in a speedy, efficient, knowledge-building way.

The background and processes of Mind Genomics have been written about for more than 15 years, and can be found in a variety of papers, some dealing with the general method and applications (e.g., [11]) others dealing in more depth with specific applications, such as the law [12], charity donations [13], environmental considerations [14], digital marketing [15], and so forth. Some of the seminal experiences in the formation of the emerging science of Mind Genomics appeared almost 15 years ago in a book Selling Blue Elephants: How to Make Great Products that People Want Before They Even Know They Want Them [16].

The Mind Genomics process follows these seven steps:

Step 1 – Define the Topic, Create a Set of Questions Which Elaborate the Topic, and for Each Question Create a Set of Answers

The Mind Genomics ‘template’ provides a structured system to focus the researcher’s effort. For these two projects presented in this paper, the experimental design comprises a topic, six questions elaborating the topic, and in turn six answers to each question. It is important to keep in mind that the Mind Genomics effort lends itself naturally to fast feedback and iteration. Thus, the elements need not be worked and reworked until ‘perfect.’ Rather, it suffices to have a reasonable set of elements. It is very straightforward to run the study, considering it as a first iteration. The results, when returned, can always be improved, and the study re-run in a few hours using new elements to replace the elements which did not perform well.

Step 2: Create Vignettes, Combinations of Elements

The vignettes are put together according to an underlying plan, the so-called experimental design [17]. The design for these studies prescribed precisely 48 combinations, vignettes, 36 of these vignettes comprising four elements, 12 of these vignettes comprising 3 elements. Each element appears exactly five times in 48 combinations and absent from the remaining 43 combinations.

An important feature of Mind Genomics is the ability to cover as much of the design space as possible. The design space is another way of describing the many possible combinations that could emerge from creating 48 vignettes, and it is in this precise point that Mind Genomics differs from conventional research. Conventional research would create 48 vignettes, and then have many respondents test the same 48 vignettes, with the objective of reducing the variability of measurement. That is, conventional research implicitly limits the focus of the effort, creating what is thought to be correct, and spending the time, money, and effort on validating the guess. There is little learning to be gained as one goes along the research path. The hope is that the research can ‘intuit’ what to do next when the effort fails, recognizing that the research steps are not iterative, but rather evaluative. Iteration thinking is not built in, but rather becomes an unwanted necessary step when the research fails to confirm the ingoing intuitions manifested in the test stimuli, here the 48 vignettes combining the 36 elements.

In contrast to the above, Mind Genomics takes as its cue the approach represented by the MRI, the magnetic resonance imaging, used in medicine. The MRI takes many pictures of the underlying tissue, each picture from a different angle. Afterwards, through a computer program, the MRI combines these pictures to get a better idea of the underlying tissue, one in three dimensions. No single ‘picture’ is right. Each picture is ‘noisy’, and not useful by itself, but it is the pattern emerging after combining the pictures which is realistic.

Step 3: Define the Rating Scale

The rating scale is the way that the respondent can communicate with the research and give her or his opinion. In these two studies, the rating scales were 9-point scales (Likert or category scales), anchored at each end.

The 9-point scale is simple, easy to use. The ratings for the scale, however, the nine points, are hard to understand for the manager, despite being easy and widely accepted by researchers. Common practice for the past decades is to transform the scale to a binary scale, a yes/no scale. Managers can easily understand the scale. Ratings of 1-6 were transformed to 0; ratings of 7-9 were transformed to 100. This transform converted the scale to a format that managers can more easily understand. The transformation is accompanied by the addition of a vanishingly small random number (~ 10-4). This small number prevents the analysis program (OLS, ordinary least-squares regression) from ‘crashing’ when it tried to deal with a set of data all of which have either value 0 (all vignettes had been rated 1-6), or values of 100 (all vignettes had been rated 7-9).

Step 4: Invite the Respondents to Participate

The respondents are invited by professional groups, called online panel providers. It is always tempting to save money and provide panelists from one’s ‘network’, but the reduced cost turns of out to be one of the most expensive ‘savings.’ With 304 respondents participating for the New York study, and with 158 respondents for the Israel Palestine study (about half Israelis, half Palestinians), the judicious approach is to hire a professional organization to provide the panelists. The organization does so at a reasonable fee, which allowed each study to be completed in less than 24-48 hours, without effort by the researcher.

Step 5: Orient the Respondents in the Task

It is best to provide as little information as possible about what is really expected, and instead simple introduce what the study is about, and some of the necessary information, such as the length of the interview (very important), the fact that the all the vignettes are different, and that there are either one or two rating scales. Often respondents who participate, evaluating 48 vignettes, feel sure that they have ‘seen this vignette before.’ They could not have, but it matters little, and it is important to assure them that they are seeing stimuli that have been meaningfully crafted.

Step 6: Create Equations (Models) Relating the Presence/Absence of the 36 Elements to the Binary Ratings (0/100)

The equation is written as: Dependent Variable (Top 3) = k₀ + k₁(A1) + k₂(A2) … k₃₆(F6). The foregoing equation deconstructs the response to the vignettes, so that we begin with an additive constant, and then estimate the part-worth contribution of each element to the dependent variable. The additive constant is the estimated value of the dependent variable (0 or 100), in the absence of any of the 36 elements.

The additive constant is a purely estimated parameter. It has no real existence, but it can be used to estimate the proclivity of the respondent to rate a vignette as 7-9 in the absence of elements. Of course, by design, all of the vignettes comprised 3-4 elements, so the additive constant cannot really exist. Nonetheless, as we will see, the additive constant gives us a sense of the degree to which a respondent is ‘ready’ to say something, even without evidence.

When we see high additive constants, we can be sure that the respondents feel strongly and positively towards the topic. In contrast when we see low additive constants, we can feel strongly that the respondent is not predisposed to rate the vignettes high but rather waits for the momentum imparted by just the right elements to carry matters forward.

The 36 coefficients each reflect the marginal, or part-worth contribution of the individual element to the value of TOP3. When the coefficient is 0, we conclude that the element has no ability to drive the response. When the coefficient is positive, we conclude that adding the element to a vignette increases the percent of respondents rating that vignette 7-9. For example, a coefficient of +5 for an element means that when the element appears in the vignette, an addition 5% of the responses will be 7-9. In contrast, when the coefficient turns out to be -5, we conclude that when the element appears in the vignette, 5% fewer of the responses will be 7-9. We don’t know whether the 5% fewer will migrate to very strong negatives (viz., 1-3) or migrate to mere indifferences (viz., ratings of 4-6).

From the point of view of statistics, the coefficients usually end up with standard errors of approximately 4-5, meaning that we should pay strong attention to elements with coefficients of 8 or greater. We should pay attention to all positive coefficients, but the elements with high positive coefficients, 8 or greater are really important.

The Mind Genomics method ends up producing many numbers. For example, just looking at the total panel coefficient can be overwhelming. There are 37 numbers to consider when searching for a pattern, the additive constant and be coefficient for each element A good practice, one adopted here, is to present only those coefficients that are noticeably positive (2 or higher) and ignore those coefficients which are 1 or lower. Furthermore, when an element has no coefficients for any group which are positive, it is not instructive to present that element. The element ends up taking up room, and not teaching anything. In the data tables presented later in this paper many elements do not appear because they fail to produce impactful statements.

Step 7: Create Individual Level Models and Use the Coefficients as Inputs to Clustering

Clustering will identify new-to-the-world groups (mind-sets) based on the patterns of responses to the messages. We are taught to think of people in terms of who they ARE, what they DO, and what they say they THINK/BELIEVE. Marketers call this segmentation.

The standard ways of dividing people, so-called geo-demographics, can generate a large vector of information about a person, based upon gender, age, income, education, marital status, and so forth. In the 1960’s, William Wells, a market researcher working in advertising introduced the notion of psychographic segmentation [18], a way to divide people by the pattern of what they believed when the topics were lifestyle, beliefs, etc., so-called macro-topics. Today’s technologies allow people to be divided by more micro-patterns, such as the way they search on the Internet for specific ‘things.’

In the spirit of dividing people, Mind Genomics looks for groups as well, but this time groups based upon the pattern of responses to a limited, focused issue. These are called mind-sets. A mind-set comprises individuals who think alike in a limited topic, such as the patterns of discussions that they prefer (viz. mind-sets emerging from Study 1 on Jewish discussions or mind-sets emerging from Study 2 on positive outcomes in the conflict between Israelis and Palestinians). The mind-sets are obtained from the pattern of coefficients of the respondents who participated. Recall that each respondent evaluated a unique set of vignettes, but that the mathematical structure of the 48 combinations evaluated by each respondent was the same. The 48 combinations sufficed for a valid experimental design that one creates the set of coefficients for each respondent separately. It is the set of 36 coefficients for each respondent in the study which becomes the basis on which individuals are separated into mind-sets. Individuals with ‘similar’ patterns of 36 coefficients are put into a cluster or mind-set by a mathematical algorithm (k-means; [19]). The outcome is a small set of clusters, which comprise individuals within a cluster having similar patterns of coefficients, and with the patterns of averages of the 36 coefficients different from cluster to cluster. One might envision this as a set of globes, far away from each other (the clusters or mind-sets), but a group of points (respondents) swirling around inside the globe and being close to each other. The number of such mind-sets is left to the researcher. Two criteria have been used to select the number of clusters, parsimony and interpretability, respectively. Parsimony refers to the number of clusters or mind-sets. Fewer is better. Interpretability refers to the fact that the pattern of coefficients, the strong performing elements are similar, and tell a ‘coherent story.’ Not a perfect story, of course, but something which seems to make sense.

As the number of clusters increases, parsimony decreases, but interpretability increases. The act of clustering respondents calls into play a balancing act between creating sets of respondents whose data can be easily understand, and creating at the same time a large number of such groups, so that at the end of the process one is not sure whether the clusters or mind-sets are ‘real’ They may tell interesting stories, but there may be simply too many clusters on which to make generalizations.

We now move to the two studies, first the study of a cohesive group, the Jewish population of New York, and then study of two historically opposed populations, Palestinians and Israelis. Both studies are about discussion, about finding common ground, the first with New York City Jews to bring an ethnic group together, the second with Palestinians and Israelis to reduce tension, and begin to bring the groups together.

Study 1: Search for Common Ground for Discussions about Israel among NYC Self-defined Jews

This study was run under to aegis of Dr. Jonathan Cummings of the Jewish Community Relations Council of New York to determine the features of a venue for productive conversations. We show the elements, the mind-sets, and then where relevant information about the mind-sets (for study 2). Table 1 presents the elements.

Table 1: Elements for the Mind Genomics study regarding common ground for discussions about Israel among NY self-defined Jews.

The orientation was simple, focusing primarily on the process, and providing few specifics about the topic. It is the topic which will be ‘particularized through the elements.

Today you will be taking a survey regarding conversations about Israel in New York’s Jewish Community. Sometimes, talking about Israel can be difficult. Sometimes, we may not want to talk about Israel at all. We are interested in finding out what might make those conversations more satisfying and would like to know your opinion regarding different kinds of conversations with others in the local Jewish community, outside of your inner circle of contacts and friends. It will take you between 10-15 minutes to complete the survey. During this survey, we will show you several scenarios describing different conversations in various discussion settings. Although they may seem similar, please note that each screen combination is UNIQUE.

You will be asked the same question for each test screen: How satisfying would a conversation about Israel be with members of the Jewish community with whom you do not generally converse based on the above: 1 = not at all satisfying…, 9 = very satisfying

Figure 1 shows a sample vignette that the respondent would see, except for the boxed information on the left. These are the ‘questions’, which the respondent never sees. The role of the question is to provide a stimulus for the six different answers.

Figure 1: Example of a 4-element vignette shown to the respondent. The respondent does not see the boxed information on the left.

The study was run with 304 respondents. Table 2 shows the strong performing elements. Table 2 suggests that:

Table 2: Result from models for the conversations about Israel among NYC Jews (Study 1).

1. The basic level of expected satisfaction from the conversation (additive constant) is moderate for total, for Mind-Set 1 and Mind-Set 2, but quite low for Mind-Set 3
2. No element drives a feeling of strong satisfaction when we look at the total panel of 304 self-identified Jews in the New York region. The 304 respondents would be considered a homogeneous group, discussing a topic of concern among Jews.
3. The data suggests dramatically different mind-sets. What appear to be irrelevant elements at the level of the total panel end up being strong performers for the mind-sets?

Study 2: Discussion among Palestinians and Israelis Regarding What Will It Take to End the Conflict (Question 1) and Create Lasting Peace (Question 2)

The objective of Study 2 was to develop a system which could deal with conflicts in a way consonant with the vision of Mind Genomics, namely treat the issue as the conflict of different mind-sets. The second study was run, under the aegis of Professor Peter Coleman and his associates at Teacher’s College, in Columbia University, and under the aegis of Professor Martin Braun of Queens College, City University of New York. The study was run 2011-2012, a decade ago, using the same experimental design as had been used for Study 1 on Discussions about Israel.

The elements appear in Table 3. The elements were created by the team led by Naira Musallam at Teacher’s College. They were designed to be short, easy to read ideas. The elements were created through a process involving depth interviews, ethnography, brainstorming, competitive analysis. Finally, the elements were developed with a psychodynamic and psychoanalytic orientation, dealing with different aspects of needs, wants, and perceptions.

Table 3: Elements for the Mind Genomics study regarding cessation of hostilities and lasting peace.

It is important to note that there is no fixed process for developing ideas. in the end, it is always a matter of creative thinking, of merging the richness of language to describe and the need to portray what should be described.

The respondents read the test vignettes, but this time rated the vignette on two rating scales. The first was the likelihood that this described the situation where the mutual hostilities would stop. The second scale was whether this described a situation which would move to lasting peace. One can look at these two sales as intellectual and emotional, as evaluating what will happen, and what could happen.

Half the respondents were Palestinians, recruited by friends, and half the respondents were Israelis, also recruited by friends. There was no pre-screening about attitude. Rather, the respondents simply were introduced to the topic with an explanation of the scale.

The study was set up so that the 158 different experimental designs were divided, so that the first set of unique designs was allocated to the Palestinians (with study totally in Arabic), the second set of unique designs was allocated to the Israelis (with study totally in Hebrew), and then the process repeated. The data for the Palestinians and for the Israelis were treated as one large group, both for clustering into mind-sets, and for reportage of results. This is possible because to the computer we are dealing with 158 respondents, all evaluate vignettes from the same large design.

Once again, the focus of the study is the identification of groups of like-minded individuals existing in groups which are in conflict.

Figure 2 shows the orientation pages in Hebrew and Arabic. Figure 3 shows examples of what the respondents saw in terms of vignettes. Since there were two rating questions, the vignette remained on the screen, the rating question changed. When the respondent completed the second rating, the vignette changed to the next prescribed by the underlying experimental design.

Figure 2: The orientation screen in both Hebrew and Arabic.

Figure 3: Example of a screen showing a vignette and the rating scale, in Hebrew and Arabic.

After the vignettes were completed, the respondents completed a short, self-profiling questionnaire, with some results shown in Table 5.

Table 5: Self-profile of the respondents in the two key mind-sets who are enthusiasts: Mind-set 3 (End conflict enthusiasts) and Mind-Set 2 (Peace enthusiasts).

Table 4 shows the significantly positive elements and the strong performing elements. The respondents were clustered simultaneously on ratings for Question 1 (cessation of hostilities) and Question 2 (lasting peace). This joint clustering was done by combining the two sets of coefficients to create a vector of 72 numbers, to which the k-means clustering was applied.

Table 4: Result from models for ending the conflict (Q1) and establishing long-lasting peace (Q2), from Israelis and Palestinians. Each mind-set comprises both Israelis and Palestinians.

The results suggest four interpretable mind-sets. Across both questions Mind-Set 2 appears to respond to elements which are positioned as ‘end the hostilities’ and Mind-Set 1 appears to respond to elements which are positioned as peace.

It is striking and somewhat disconcerting that in Table 4 that most of the cells are blank, having generated coefficients of +7 or lower. To help the patterns emerge, we show only those cells with strong performance, viz., coefficients of +8 or higher dramatically, we look only for strong performing elements.

The four mind-sets each comprise a mix of Palestinians and Israelis. This is important, because it gives hope that there can be found like-minded individuals in hostile populations, with perhaps some of the mind-sets capable of negotiation.

The total panel contains no strong performing elements at all. Nor do Mind-Sets 3 and 4, comprising 67 of the 158 respondents. There are, however, strong responses in the mind-sets, viz., suggesting that there are areas of agreement. All that we have to is find them.

The messages which drive interest for ending the conflict are:

1. Freedom from violence and oppression are individual human rights
2. Ongoing community exchanges between Israeli and Palestinian youth help our situation
3. More and more people everywhere are developing a stronger sense that they are all members of one global community

The messages which drive work for peace (Mind-Set 1) are:

1. Lasting peace and justice in Israel-Palestine will only happen by Israelis and Palestinians working together
2. There are currently many areas of economic, technological, cultural and educational cooperation between Israelis and Palestinians
3. A solution to the conflict and compromise over Jerusalem and the refugees is possible

The traditional approach to understanding people is to create a surface understand of their minds, and in turn probe deeply into who they are. Table 5 shows a breakdown of self-profiling classification of mind-sets 2 (End Conflict enthusiasts) and mind-set 3 (Peace enthusiasts). The information is enlightening, but the important information is missing, viz., the reason for the strong performing messages. We know about the enthusiasts, but would never have predicted which elements in Table 4 would have performed well

Discussion and Conclusions

At the start of the efforts underlying these two studies a decade ago, the vision for Mind Genomics was to identify the ‘optimum messaging.’ The size of the study, 36 elements (six questions, each with six answers) provided a large array of possible ideas to include in the study. The rapid turn-around time, less than a day, was not considered a particularly strong ‘positive’, but the possibility of testing many messages was a positive. The notion was that Mind Genomics provided a testing platform for many well-thought-out ideas, rather than one or two ideas. The worldview that accuracy, even with long cycle times, was critical. It was better to expend a great effort, to get deep thinking, and then to do the study.

The outcomes of the studies, reported here, were “interesting” but failed to find a receptive audience. Part of the failure was lack of visibility of the results. Another part was lack of knowledge about what Mind Genomics was, and what Mind Genomics could deliver. The target audiences, those in policy, those in academia, were stuck on the traditional methods, the slow, often tedious, eventually self-correcting, one-at-a-time thinking.

Over the decade, however, Mind Genomics has evolved to an iterative system, one providing virtually instantaneous results, at very low prices, with the objective of creating an ongoing database, the wiki of the mind. Coupled with this is the recognition that the traditional methods of science, the tortuous one-at-a-time hypothetical-deductive system, the creation of hypotheses and careful testing, does not work well in a world of language and feelings, where there are many ways to express winning ideas, and many more ways to express losing ideas. Furthermore, speed, once a negative because ‘if it is so fast it can’t be particularly good!’ has been supplanted by a culture of speed.

There is something, so obvious that it may be boggling. A continuing finding of Mind Genomics is that virtually no one really ‘knows what will work.’ These studies, run today, often show a lot of blank cells, not so much for products and services which are tangible, but rather for social issues which have been the food and drink of policy makes for generations. A better system was needed. The plethora of empty ‘data’ cells in Tables 2 and 4 are witness to the fact that the ‘best guess’ elements from experts do not drive the response. People in the business of ‘knowing’ may not know. This is not a criticism. The same plethora of empty cells for results occurs for marketing services and products, as well as social issues, legal statements in litigation or patients in the hands of medical professionals looking for guidance in the way which is most appropriate to their mind-sets.

Some of the answer to ‘fewer empty cells’ comes from the use of Mind Genomics, perhaps in a more abbreviated, simper form, not with 36 elements (6×6; 48 combinations) but with fewer than half that number, 16 elements (4×4; 24 combinations). There are three aspects to the opportunity all embodied in a publicly available tool, BimiLeap (www.BimiLeap.com).

Up-front Thinking

It may be daunting to have to think of 36 elements, but to think of 16 elements should be a far simpler task.

Speed, Collapsing the Process to Minutes and Hours

The second is to produce a system with collapsed timelines, a system which is templated, so that the Mind Genomics Project’ can be set in up in 15 minutes, launched, and the fully analyzed reports, ready for presentation,, emerge within 15 minutes, or at most minutes.

Change from Confirming to Iterating

Analysis paralysis, one of the banes of progress, perhaps the cause of the results here, emerges from the dutiful action of ‘measuring nine times, cutting once’. It might be better to think quickly, iterate quickly, update, and iterate again. Each iteration, in turn, to be done from front to back in the space of 60-90 minutes, at low cost, with ongoing updating, keeping and expanding the good, throwing away the ‘bad’, the ‘poor’, the irrelevant.

The prospect of going from no knowledge at 9am to nine iterations by 9pm or earlier, is simply ‘game changing.’ One can just imagine the number of elements which emerge over time as potentially strong messages, as the researcher iterates to better and better messages, simply by the mechanical effort of testing, evaluating, discarding, expanding, and retesting. Furthermore, the proof is immediate, manifested in the number of ‘filled cells,’ the magnitude of the positive coefficients, and the practical results from messaging.

One last part of the vision from 2012 deserves mention, a vision which is now becoming almost a trivial application of the above. As noted above segmentation and the use of mind-sets has been the domain of the world of marketers. But would about databases of mind-sets for conflicts around the world? What if each conflict could be studies with an iteration of 20 studies, as noted above, so that one arrives as a database of elements which reveal what can be agreed upon? It is quite likely that the results will require mind-sets. What if the process of iterating could be continued, the strong performing elements validated, and then a ‘PVI’, personal viewpoint identifier incorporated into the database [20]. The PVI would allow people in the conflict on both sides to be assigned to one of the mind-sets involved in the conflict. From there, negotiations could begin between antagonists on the two sides who happen to share the same mind-set, a mind-set which holds a point of view allowing for peace. One can imagine a library of 100 books, the Library of Today’s Conflicts, one book for each conflict, created in weeks, at low cost, from 10-20 iterations, and with its own PVI!

Acknowledgments

The authors wish to acknowledge the support and encouragement of these individuals a decade ago who made the studies possible, both through financial support, and through direct participation.

Peter Coleman, Columbia University, Teacher’s College, New York, USA

Jonathan Cummings, Jewish Community Relations Council of New York, New York, USA

Naira Mussallem, Columbia University, Teacher’s College New York, USA,

Janna Kaminsky, Moskowitz Jacobs, Inc. White Plains, USA

John Lightstone, Lightstone Capital (Deceased), White Plains, USA

Taly Marian, Teacher College, Columbia University, New York, USA

Nora el Zokm, Teacher’s College, Columbia University, New York, USA

Steven Onufrey, The Onufrey Group (Deceased), Philadelphia, USA

Note: The current (free) program for Mind Genomics is located at www.BimiLeap.com. The only fees are processing fees on a per respondent basis.

References

1. Cohen SM (2002) Relationships of American Jews with Israel: What we know and what we need to know. Contemporary Jewry 132-155.
2. Coleman PT, Bass B (2019) Facing uncertain times together: Strengthening intercultural connections. Journal of Intercultural Communication 22: 1-14.
3. Dessel AB, Yazbak Abu Ahmad M, Dembo R, Ben Hagai E (2017) Support for Palestinians among Jewish Americans: The importance of education and contact. Journal of Peace Education 14: 347-369.
4. Keysar A (2010) Distancing from Israel: Evidence on Jews of no religion. Contemporary Jewry 30: 199-204.
5. Waxman D (2010) The Israel lobbies: A survey of the pro-Israel community in the United States. Israel Studies Review 25: 5-28.
6. Ben Hagai E, Zurbriggen EL (2019) Bridging narratives: Predictors of Jewish American and Arab American support for a two‐state solution to the Israeli–Palestinian conflict. Analyses of Social Issues and Public Policy 19: 177-203.
7. Hagai EB, Zurbriggen EL, Hammack PL, Ziman (2013) Beliefs predicting peace, beliefs predicting war: Jewish Americans and the Israeli-Palestinian conflict. Analyses of Social Issues and Public Policy 13: 286-309.
8. Mearsheimer JJ, Walt SM (2009) Is it love or the lobby? Explaining America’s special relationship with Israel. Security Studies 18: 58-78.
9. Rosenthal ST (2001) Irreconcilable differences? The waning of the American Jewish love affair with Israel. Brandeis University Press.
10. Luce RD, Tukey JW (1964) Simultaneous conjoint measurement: A new type of fundamental measurement. Journal of mathematical psychology 1: 1-27.
11. Moskowitz HR (2012) ‘Mind Genomics’: The experimental, inductive science of the ordinary, and its application to aspects of food and feeding. Physiology & behavior 107: 606-613. [crossref]
12. Wren JE, Williams TC (2009) Selling blue elephants to the jury: Potential application of rule developing experimentation in Litigation. Baylor Law Review. Rev 61: 1.
13. Gabay G, Moskowitz H, Gere A (2019) Understanding the donating mind and optimizing messaging- public hospitals. In 12th Annual Conference of the EuroMed Academy of Business.
14. Gere A, Zemel R, Papajorgij P, Radványi D, Moskowitz H (2020) Public driven and public perceptible innovation of environmental sector. In: Innovation Strategies in Environmental Science 69-106.
15. Salom J (2021) Mind Genomics with big data for digital marketing on the internet. In: Handbook of Research on Methodologies and Applications of Supercomputing. IGI Global 282-289.
16. Moskowitz HR, Gofman A (2007) Selling Blue Elephants: How to Make Great Products that People Want Before They Even Know They Want Them. Pearson Education.
17. Gofman A, Moskowitz H (2010) Isomorphic permuted experimental designs and their application in conjoint analysis. Journal of Sensory Studies 25: 127-145.
18. Wells WD (1975) Psychographics: A critical review. Journal of Marketing Research 12: 196-213.
19. Kaufman L, Rousseeuw PJ (2009) Finding groups in data: an introduction to cluster analysis (Vol. 344). John Wiley & Sons.
20. Gere A, Moskowitz H (2021) Assigning people to empirically uncovered mind-sets: A new Horizon to understand the minds and behaviors of People. In: Consumer-based New Product Development for the Food Industry 132-149.

DOI: 10.31038/MGSPE.2022211

Abstract

In six parallel studies, selected as relevant for education from a total of 35 studies about ‘giving’ (Give It! project), respondents evaluated the appeal of messages for donations to education causes. These specific causes included, respectively, Importance of Reading, Education about Art, Technical Education, Art in Education, Alumni Efforts and University Scholarships. In each study, respondents recruited by an online panel recruited provide 70-80 respondents in each study. Each respondent selected the study of interest from a list and participated in the study they selected. Respondents each evaluated unique sets of 60 vignettes, created from 36 elements, presented in different combinations for each respondent, with the vignettes created according to an underlying design, permuted for each respondent. The pattern of results revealed three different mind-sets cutting across the six studies. These three mind sets were: MS1 (Commitment) Because I Care….it’s about what I can personally do to make the issue better; MS2 (Actions) Showing Support It takes more than just effort and good wishes to make things change… it takes money, time, items’; MS3 (Effect) It Makes a Difference….it’s about what can be done to help those affected by the issue. The discovery of mind-sets, and their presence in different proportions in the six studies, suggest that on a practical level communications seeking donations for the various education causes would be best served by mixing together strong performing and mutually compatible messages appealing to each mind-set.

Introduction

In 2002 author Moskowitz along with Jacquelyn Beckley and Hollis Ashman of the Understanding and Insight Group, Inc. created a set of studies called ‘Give It!’ The objective was to use the emerging science of Mind Genomics to create a database of patterns of response to messages about ‘charitable donations.’ The focus of these then-called ‘It!” studies was to explore the way people responded to these messages with the aim of uncovering basic mind-sets in the population. The previous studies in the It! series dealt with foods (Crave It!), beverages (Drink It!), and insurance (Protect It!), as well as anxiety (Deal With It!), and shopping (Buy It!). The Give It! studies, funded by the O’Grady Foundation, broke new ground in understanding the messages which would drive people to say that they were intrigued. The focus was not to drive giving, but rather to find out the messages that would put people into a positive frame of mind for a specific cause.

The rationale underlying the It! studies was the recognition that our knowledge of what drives donations is extensive, but piecemeal. In the words of a recently published paper [1]:

“Charities operate in a highly fragmented environment with many players competing for individuals’ support. The limited resources available for campaign development (creative, filming) and execution (media planning, on-air time) means that charity marketers need to use the most effective principles to ensure return on investment. Commercial marketers can use clear guidelines published on how to execute the brand to enhance advertising effectiveness and, more specifically, brand recall and recognition. Whether such guidelines are adhered to by charity marketers is unclear as no known research exists on this topic.”

A glance into the academic literature through Google Scholar® for the phrase charitable donation messaging reveals 21,200 as of this writing (Fall, 2021), with the academic literature focusing on general theory of why people give, and in turn, messaging which works. This focus on trying to understand the deeper WHY something works is admirable because it increases our understanding of the mind of people. Thus, it should come as no surprise that the academic literature focuses on the general types of messages used for different causes, the modes of donating e.g., Chen [2], and of course the nature of the giver. As in most academic studies of these social issues, the objective is to work from the top down, from general classes of ideas to the effectiveness of those ideas in particular uses [3]. Thus, one might see studies focusing on ‘guilt’ as a topic of the message, and its effectiveness. For example, consider crowd funding for a cause. Chen [4] reported that three types of messages work best: guilt, utilitarian products, and emotion messaging, respectively. Do we find this troika reflected in giving for causes?. Occasionally one might encounter papers dealing with specific phrasing, but the focus on the performance of such phrase is motivated by the fact that the phrase itself is unusual (e.g., even a penny will help; Shearman) [5].

Moving further, from the general to the particular, we find that in a great deal of the scientific literature there does not seem to be a systematic review of the power of specific messages, as a focus of the research, although one might speculate that such information is a staple of private databases used for seeking donations. It is there, in the world of the everyday specifics, the world of the granular, that Mind Genomics makes its mark, and its contribution.

The Mind Genomics Approach

To understand Mind Genomics and the large topic of ‘giving’ we first turn to the world of conventional research and specifically the types of experiments that are done. In the world of conventional research, a typical experiment to understand the mind of the ‘giver’ for creates an experiment with one or two conditions, often exaggerated changes of what might occur in everyday life, executes the experiment, and determines which antecedent method drives the greater amount of the criterion response. This traditional approach creates its knowledge base by aggregating together the results of such isolated experiments, establishing the pattern by a meta-analysis of these many findings. There may be a desire to create a library of practical information, ‘vetted’ by science, but the one-at-a-time process allows such library to emerge years, often decades after the data has been collected, the individual experiments reported, and then re-considered as a totality to create the library.

Mind Genomics differs from the conventional methods in worldview, execution, and types of data collected, and types of inferences made, respectively. Mind Genomics focuses on decision rules emerging from systematic experiments with many variables, doing so in ways which have become rapid, scalable, affordable, and amenable to iteration. Mind Genomics presents the respondent with many systematically varied messages, each respondent evaluating a different set of messages. It is the pattern of responses to the set of systematically varied combinations which provides the necessary data, but only after the pattern id deconstructed into the contribution of the individual messages. At a more global level, Mind Genomics looks for patterns across stimuli, and for emergent groups in the same topic area who show meaningfully different patterns of responses to the same set of stimuli. These are so-called mind-sets, appearing again and again across all topics explored by Mind Genomics, with the natures of the mind-sets driven by the topic itself [6-8]. These mind-sets vary in nature by category (nature of product, nature of service), and most important, clearly transcend most conventional ways of dividing people (viz., gender, age, country etc.). During the past 30 years, since 1993, Mind Genomics have evolved from a one-off system for research, based on conjoint analysis, to a templated system set up so that anyone can become a researcher. We present here the template adapted to the set of studies run in 2002 [9-11]. Note that processes which started out manual, such as combining files for the ‘reporting’ have become totally automated as of this writing.

The Mind Genomics Research Process Applied to the World of Giving

Step 1 – Select the Topics about ‘Giving’ to Explore

Figure 1 shows the 35 topics. At the time of the actual study the respondent would select the topic of interest, go to the study, and then participate. The topic would ‘disappear’ from the wall after a certain number of respondents completed the study. Unknown to the respondent, the ‘test material’ viz., the ‘elements, ‘ viz., phrases, full set of studies were almost mirror image of each other across the 35 different studies except for eight of 36 elements which were specific to the topic.

Figure 1: The wall of 35 Give It! studies. The respondent selected the study and was led to the actual Mind Genomics experiment corresponding to the study

Step 2 – Create the Basic Structure of the Experiment, Comprising a Specific Number of ‘Questions, and a Specific Number of Answers’ for Each Question

The IT! studies conducted during the first five years 2001-2005 used four questions, each with nine answers, or in the language of today’s Mind Genomics, four questions, and nine elements (answers). The choice of this 4×9 experimental design was based upon the joint desire to acquire as much information as possible, and the popularity of Mind Genomics designs around 2003, the early phrase of the internet in consumer research.

The designs had to fall into the class of permuted experimental designs, which could be created in the hundreds, so that each respondent would evaluate answers (elements), albeit in different combinations. That structure, allowing for strong individual-level analysis, pre-determined the set of viable design structures. Table 1 presents the 36 different messages, groups into four categories, or in today’s usage, four questions. Each category (or question) comprises nine elements (or answers). It is critical that a group of related elements, viz., elements of the same type but which may carry different information, just appear in the same category or question. This requirement is a ‘bookkeeping device’ to ensure that two elements of the same type, but containing mutually contradictory information, can never appear together, since the vignette specifies as much. The elements or answers are direct statements, painting a word picture. As the vignettes will show below, simple one- or two-word answers to a question do not suffice to paint a word picture. The objective was to create answers or elements, which in combination, painted such a word picture without the need of a question to set the stage.

Table 1: The 4 questions or categories, and the 36 answers or elements, nine per question/category. The elements pertain to supporting alumni efforts.

It is important to keep in mind that any large-scale investigation of a vertical, such as donations with Give It!, must sacrifice a great deal of the specifics of a cause in the interests of comparability of causes. The objective of Mind Genomics applied to the vertical is to discover general patterns from sets of common elements. The alternative approach, doing individual studies for each topic, would provide deeper information, but the meta-analysis of the results might require a great deal more effort, and require involve luck at the end, rather than planning at the start. In Table 1, eight of the 36 elements have an ‘x’ added to their code. These are elements which are similar across the six different mind-sets, but also contain topic-specific language that was changed in a minor fashion from study to study.

Step 3 – Create Small, Easy to Read Vignettes, Using an Experimental Design

Step 3 creates the test stimuli, the combinations f messages. In the language of Mind Genomic, these combinations are called vignettes. Figure 2 shows an example of a vignette. The stimuli comprise 60 different combinations, all similar in format to Figure 2, except that some comprised two elements, some three elements, and some four elements. The spacing and design of the vignette is such that the respondent could easily read the vignette. Experience with Mind Genomics suggest that the combination of messages in a spare form, with open space allows the respondent to quickly ‘graze’ the information and make a rating. The design of the test stimulus in Figure 2 goes contrary to approaches, which present the respondent with a crafted paragraph. In the end, comments from respondents who, having been presented with these spare looking vignettes like that of Figure 2, concur that it is easier, less fatiguing, less frustrating to deal with form of design when rating many vignettes, rather than working one’s way through what a dense paragraph.

Figure 2: Example of a vignette comprising four elements

The vignette is created according to an underlying experimental design [11]. The design prescribes the exact composition of each vignette, specifying which specific elements are combined. To the novice unfamiliar with the design structure, such as the respondent, the vignette looks as if the elements had been thrown together at random. The truth is precisely the opposite. The compositions are carefully crafted to ensure that each element appears equally often, that each element is statistically independent of every other element, and that there are sufficient of compositions or vignettes which lack one or two elements. The latter feature of the experimental design ensures the data can be used by OLS (ordinary least=squares ) regression to deconstruct the response into the contribution of the individual 36 elements. Furthermore, the built-in incompleteness of some vignettes prevents the statistical problem of multi-collinearity, which would eventuate in the crash of the statistical analysis for that data set. Finally, and most important, the coefficients emerging from the OLS regression have ratio-scale values, and are comparable from study to study, from period to period, and even across different individuals. Two other features of the experimental design are important to note. The first is that no question or category can contribute more than one element to a vignette. The rationale for this constraint is that quite often the question or category comprises elements which mutually contradict each other. Were these mutually contradictory elements to appear in the same vignette they would corrupt the response to the vignette. The second significant feature in the Mind Genomics system is the permutation of the basic design, so that the mathematical structure is identical, but the actual combinations differ from one another. The use of the permuted design at first seems to be merely a statistical enhancement, but the reality is that it is a frontal attack on some of the thinking of conventional science, and an alternative to the oft-quoted proverb ‘measure nine times and cut once.’ This permuted design (Gofman and Moskowitz) emerged out of a recognition that the standard research approaches are based on reducing statistical error by repeating the same experiment with dozens or hundreds of people. The implicit assumption of the conventional research procedure is that the ‘correct answer’ is known, and that the research is going to confirm or disconfirm that guess. Yet, the ‘reality on the ground’ is that no one really knows what messages will work, and what messages will fail to work. Thus, the choice of the messages and the combinations becomes one’s best guess. The permuted design avoids the need to select a limited set of vignettes, or test combinations at the start of the study. The key benefits of the permuted design used by Mind Genomics are ability to explore a wide number of alternative ideas (36 in this study), and at the same time explore a great deal of the underlying ‘design space’ of different combinations. Each respondent evaluated a unique set of 60 vignettes. Across the approximate 70-80 respondents, this means that for each topic (e.g., technical education), the Mind Genomics experiment investigated the response to many different vignettes, albeit each vignette evaluated just a few times The creation of a model relating the presence/absence of the elements to the ratings was stabilized because of the many different combination. Even if one or several, or several dozen were mis-judged, the weight of the approximate 60×75, viz. 4500 judgments of different combinations sufficed to ensure that no systematic error could affect the result. In contrast, conventional research testing the same 60 vignettes 75 times each might be well advised to make sure that the 60 vignettes are the correct vignettes. Choosing the wrong single vignette to test or having an aberrant reaction to that vignette is not quite as serous in Mind Genomics as it is for conventional research. To summarize this point, it should thus be kept in mind that the permutation of the combinations ensures a wide coverage of the possible combinations, producing a better experiment. It is simply very difficult to introduce a strong bias when the combinations change all the time. It is the underlying pattern, emerging for 4800+ vignettes which is critical.

Step 4: Invite the Respondent to Participate, Introduce the Respondent to the Subject, and Execute the Actual Interview

The It! studies were run with the Canadian on-line panel, Open Vue Ltd. Their panel comprised both USA respondent and Canadian respondents, among many others. The respondents were selected to be residents of the United States. Open Vue sent out email invitations to its panelists. Those who answered were led to the screen shown in Figure 1, where they selected a study of interest to them. During this early period of research with Mind Genomics and with the It! studies, it became obvious that an efficient way to do 35 studies with approximately similar numbers of respondents was to let the respondent choose the study. Once the study quota was filled, the study disappeared from the available choices. Figure 3 shows the orientation screen for the study. Most of the screen is taken up with bookkeeping details, about the length of the study, the fact that the screens (viz., the vignettes) differ, the rating question, and the expected time of the study. During this early period of internet research, the respondents were not yet saturated by requests to participate in simple studies or evaluations of their experience, and thus were more likely to donate 15 minutes of their time to the study. Nonetheless, it was important to incentivize the respondent with a monetary reward, a drawing for a prize. The three prizes were the incentive across all 35 studies. That is, all respondents across all studies were entered into the drawing, and three respondents were selected as winners.

Figure 3: The orientation page for the Give It! studies. Each study was introduced by the same page, with the only difference being the specific topic.

Step 5- Transform the Rating to a Binary Dependent Variable and Create Individual-level Models

The respondent rated each vignette on the simple scale ‘How much does this giving situation appeal to you?’ Note that the respondent was not asked to state whether or not the respondent would donate, or how much, although those could have been legitimate questions to ask. Rather, the respondent was asked a question about feelings, about a sense of ‘appeal to me.’

The 9-point scale, a category or Likert Scale, can be easily analyzed. The problem with the scale, however, is how to interpret the scale. When managers receive data, they often ask the simple question ‘what do these ratings MEAN?’. To a manager, the fact that one can easily analyze the data with sophisticated statistics means very little when the results cannot be easily understood and acted upon. Thus, it has become standard procedure to transform these Likert scales, usually to a binary scale, yes/no. The manager using the data has no problem understanding yes/no. The transformation is straightforward. Standard practice has evolved to transforming the ratings of 1-6 to 0, and ratings of 7-9 to 100. This division of the scale makes thee interpretation easier. As a prophylactic measure, we add a vanishingly small, random number to every transformed variable to ensure that the transformed variable, viz., the newly created binary variable, has some minimal variation. If the respondent were to rate all 60 vignettes as 7-9, or as 1-6, respectively, then the transformation as just specified would create a set of 60 number, all 100, or all 0, respectively. The analysis of the data by OLS (ordinary least squares) regression would immediately crash. Adding a vanishingly small random number to the newly created binary value ensures that this unhappy event does not occur. Every vignette will have its own number, around 100 or around 0, respectively, depending upon the original rating assigned to the vignette. Once the data have been transformed the 60 rows of data from each respondent is subjected to an OLS (ordinary least squares) regression. The regression is called ‘dummy variable’, because each of the 36 element corresponds to an independent variable, and takes on only one of two values, 0 or 1, as follows: The element is either present in or absent from a vignette, so its corresponding independent variable is coded ‘1’ when present, or coded ‘0’ when absent.

The equation is: Binary (0/100) = k₀ + k₁(A1) + k2(A2) .. k₃₆ (D9)

The regression analysis created 60 rows of input data for each respondent. Each row comprised 37 numbers, additive constant (k₀) and the 36 coefficients, k₁-k₃₆. With 453 respondents participating, the regression analysis generated 453 rows of coefficients. It would be these 453 rows of coefficients that would be used to create mind-sets. The 453 rows, viz. the full data set, was subject to k-means clustering, the inputs for the clustering being the 36 coefficients k₁-k₃₆. The additive constant was not used for the clustering. To make the analysis easier, we extracted three clusters, a number usually found to reveal strong patterns, but not unwieldy to analyze. The clustering was done by the k-means method [12], which looks at the distance between each pair of respondents and tries to put respondents into a set of mutually exclusive groups so that the distance between the respondents in a cluster is small, while at the same time the distance between the centroids of the three clusters is large. The clustering does not take into account any of the ‘meaning’ of the elements, but simply tries to satisfy a mathematical criterion. Table 1 shows eight elements with the element code having an ‘x’ as the suffix. These were elements deemed too specific to the topic and were not included in the clustering. The rationale was that the clustering should comprise only those elements common in meaning to the six different topics of giving. These eight elements did not satisfy that criterion of being ‘topic-agnostic.’ They will, however, be presented in the results. Within any group, whether total, donation topic, mind-set or topic x mind-set, the corresponding additive constants and 36 coefficients were averaged to generate the results shown in the data tables. More recent approaches simply combine data together for the respondents in a defined group and rerun the regression model on the total data for the relevant group. The results are similar for both forms of analysis.

Results

Total Panel and the Six Different Topics

The Mind Genomics analysis generates a substantial amount of summary data. Our objective is to discover patterns and generalities, not to show all of the data, which would hide the patterns which exist. In order to make the discovery task simpler, we will eliminate from consideration all elements with coefficients of +7 or lower and report the element when it has a coefficient of +8 or higher. The element will not appear at all in the case that all of the coefficients for the key subgroups are lower than +8. This pruning action brings the really important elements into the foreground. Table 2 presents the results from the total panel, combining the six studies, and all of the respondents. The additive constant is 41, meaning that on average two of five responses to the vignettes will be rated as appealing (viz., rated 7-9 on the nine-point sale). The messages range from belief in the organization (D5) to affiliation (A2X), to focus on the recipient (A3X). These are the key messages that any organization seeking donations should incorporate.

Table 2: Strong performing elements from the total panel, combining all respondents across the six studies.

The array of strong performing elements increases when we move from combining all the data into one group (Total) and do the analysis on a topic-by-topic basis. The results appear in Table 3. Once again the table shows only those elements which generate at least one coefficient of +8 or higher. Thee first data column shows the sum of the strong performing coefficients and used to sort the elements from strongest performing elements to the weakest performing elements. In addition, the six studies are sorted by the magnitude of the additive coefficient, viz., the likelihood to find the vignette appealing in the absence of elements. Stated differently, the additive constant might be considered to represent the basic proclivity of the respondent towards the topic.

Table 3: Strong performing elements for each of the six giving topics.

The additive constants suggest that the most appealing topic is ‘importance of reading’, the least, but still strong being university connected topics, ‘alumni efforts’; and ‘university scholarships’, respectively. One of the properties of Mind Genomics is the fact that the coefficients have ratio scale properties. Thus, we can conclude that ‘importance of reading’ is 50% more appealing than the two university topics. Table 3 is characterized by a great number of blank spaces, suggesting that the strong performing elements do not transcend the different topics. No element drives strong appeal to more than three of the six topics The two strongest elements appear to focus on different directions, first a focus on the topic itself (D6X), and second a focus on the social aspects (A2X). There is a third focus, that of helping the person who is associated with the giving cause. These three directions suggest three different foci of appeal, directions which will emerge as mind-sets

D6X: Ensure that a strong interest in TOPIC remains a priority

A2X Sharing a love of your TOPIC with others

A3X: Ensuring that TOPIC become productive citizens

Moving from Total to Mind-sets

Table 3 hinted at the possibility that there might be different ways of evaluating the messages. Although at first glance we might consider the key factor to be the recipient of the donation so that certain topics are more attractive than another, there might be a far deeper factor at work, mind-sets. The hallmark of Mind Genomics is the discovery of these different patterns of response to messaging. The metaphor is white light, which seems to be colorless, but when the light is diffracted through a prism, the spectrum of colors emerges. We see white perhaps because the different colors interfere with each other. Mind Genomics posits that for virtually all conventional aspects of daily experience, there are different patterns of focus, of importance. What one person thinks to be important (viz., more is better) another person might as consider to be utterly irrelevant, even off-putting. The discovery of these groups, so-called mind-sets, it a matter of experiment. Furthermore, once these mind-sets are established through analysis, some of the data begins to make more sense. We may hypothesize about the possible mind-sets, but an easier way to establish these mind-sets is through a set of experiments, such as the experiments run here. The analysis to establish these mind-sets is simple OLS regression as we have done, followed by clustering to create groups of individuals with similar patterns of responses. The set of individual coefficients comprises raw material for the creation (or discovery) of these mind-sets, the permuted experimental design provides us with what we need to create the individual-level set of coefficients. As discussed above, the OLS regression analysis was straightforwardly able to create an individual level model for each of the 453respondents. The OLS regression estimated the additive constant and the value of each of the 36 coefficients, one coefficient for each element. Table 1 showed the expression of the elements for the topic of Alumni Efforts. Eight of the 36 elements appear to be specific to the topic and are marked with an ‘X’ in the element code. As noted above, these eight elements will not be used to establish the mind-sets by clustering, but then will be included in the later analyses after the mind-sets are created. The clustering method of k-means created two clusters for the 453 respondents, and then created three clusters for the same 454 respondents. The clustering procedures are a purely objective one, attempting to satisfy certain mathematical criteria. The criteria previously adopted for Mind Genomics studies for choosing the appropriate number of clusters (now called mind-sets) are not statistical, but rather qualitative. The two criteria are that there be as few clusters or mind-sets as possible (parsimony), and that each mind-set tells a story (interpretability). The criteria suggested a three-cluster solution, rather than a two-cluster solution. These clusters become the mind-sets. The clustering itself was done, as noted, on 28 of the 36 elements. Once each respondent was assigned to a cluster or mind-set, it was straightforward to estimate the additive constant and the value of the coefficient for each of the original 36 elements. That is, we resort to the 28 ‘general’ elements ONLY to create the clusters or mind-sets, and then revert back to the full set of data for further interpretation.

Three segments emerged, based on a qualitative ‘sense’ of what is communicated by the strong preforming elements. No element is strong across all six giving topics, so the interpretation of the meaning of the mind-sets become a simple heuristic with which to discuss the results. Furthermore, the clustering does not dramatically separate the three mind-sets. It’s a matter of emphasis. This is important. The dynamics of appealing to the heart of the donor become a matter of combining messages of different types, rather than focusing on one specific factor, such as EFFECT (viz., the benefit to the recipient).

MS1 (Commitment) Because I Care….it’s about what I can personally do to make the issue better.

MS2 (Actions) Showing Support It takes more than just effort and good wishes to make things change… it takes money, time, items,

MS3 (Effect) It Makes a Difference….it’s about what can be done to help those affected by the issue.

The Baseline Proclivity of the Mind-sets towards ‘What Appeals’

The additive constant tells us the estimated rating of 7-9 (appeal to me), in the absence of elements. Although the additive constant is a purely estimated parameter, it can be used to indicate the proclivity of the respondents to say ‘appeals to me’. Table 4 presents the additive constants estimated separately for the six different causes, and the three mind-sets that were developed for all the causes combined. The additive constants are sorted by average, first in descending order of cause by averaged across the three mind-sets, and then by mind-set averaged across causes. There are remarkable differences in the additive constant of the three mind-sets and in the six studies. The strongest ‘pull’ emerges from donations to help teach reading (average 50), and the weakest from alumni efforts (average 34) and university scholarship (32). This teaches us that the strongest pull, on average, is exerted by causes which pull toward young people, to give them an opportunity. Universities will have a more difficult time reaching the donors’ heartstring. In terms of the three mind-sets, we also see radical differences. Mind-Set 1 (commitment) shows the strongest proclivity to feel positive (additive constant 50), As the array is presented, there is also clear evidence for some interactions, specifically for reading. Mind-Set 2 (actions) finds its strongest pull with reading.

Table 4: Additive constants for the three mind-sets and the six donation ‘causes’, sorted by cause and by mind-set.

What Elements ‘Drive’ Positive Feelings about Giving for the Three Mind-sets?

Tables 5-7 show the strong performing elements for each of the three mind-sets. Note again that the tables present only the strong performing elements for at least one of the three mind-sets, and that all elements were considered for inclusion. In the case of the eight elements which were topic-specific, the topic is replaced by the word ‘TOPIC.’ One gets a sense of the specific thrust of the communication by reading the complete element, even with the word ‘TOPIC’ replacing the actual topic.

Table 5: Mind-Set1: The table shows the strong performing elements for MS1, labelled ‘COMMITMENT’’.

Table 6: Mind-Set 2. The table shows the strong performing elements for MS1, labelled ‘ACTIONS’.

Table 7: Mind-Set 3 . The table shows the strong performing elements for MS3, labelled ‘EFFECTS’.

The Composition of the Three Mind-sets

A hallmark of conventional research is that WHO a person is often covaries with what a person does or what a person believes. It is for this reason that so many consumer researchers spend a great deal of time collecting so-called classification questions about the respondent. What attracts many conventional researchers is the possible covariation of the easy-to-measure-behavior with additional information about the respondent.

In the world of Mind Genomics, the focus is on a better understanding of the individual. Only secondarily is the focused on establishing the relation between who a person IS versus, what the person THINKS To a great degree the lack of focus on the covariation between mind-set and behavior is due to the belief that the most pressing task is to understand the mind-sets, rather than to link the scarcely understood mind-sets to other variables. The It! studies captured a great deal of individual level data regarding attitudes and behaviors involving ‘giving’. Some of the data appears in in Table 8. Table 8 shows the complicated relationship between the three mind-sets and both WHO the person is, as well as how the person BEHAVES with respect to donating to causes. There are many patterns emerging, depending upon the way the respondent self-classifies, but no simple pattern which can be said to be common to the mind-sets.

Table 8: The percent of respondents in each of the three mind-sets, the range of percentages across the three mind-sets, and the base size. Each row constitutes a classification variable in the self-profiling classification.

Discussion and Conclusions

The academic study of ‘giving’ typically focuses on higher level motive, looking at the individual material from either actual campaigns, or creating an experiment. The important thing to note is that these studies generate a certain kind of knowledge, understanding the general drivers of donations. That information is important to understand donating to causes in the context of theories about why people do what they do. Being able to put a person’s ‘giving’ behavior, or response to different appeals allows the academic to understand yet another part of the mind of the person, for the world of the everyday. The Mind Genomics approach presented here, with its focus on the specific messages, give us a different point of view. The goal of Mind Genomics is to work with the stimuli of the everyday, in this study the stimuli being ‘messages.’ Rather than look for underlying patterns to fit into a theory, the effort is to identify what really works, and then point to what might be happening. Mind Genomics is atheoretical, but systematized experimentation. There is no theory in which to place the response patterns of giving, or at least no theory which drives the effort. Rather, the objective of the study is to see ‘what works’, with the test material being the type of messages that would be used in actual campaigns. The important results from this study are simple to summarize, namely that most of the messages really don’t work very well in terms of the ratings by respondents, and that the nature of the mind-sets which emerge is not a case of ‘polarization’ but rather ‘emphases. It’s not that the mind-sets respond only to one type of message, but rather the mind-sets respond to the messages, the elements, but some messages are stronger for one mind-set, and still positive but weaker for another mind-set. There is a strong practical side to the data presented here. That side is the fact that the patterns emerging from messages can be used immediately. There is no need to translate the test messages used in the experiment to actual messages that might be useful in a practical situation. The messages from the Mind Genomics experiment come from actual campaigns, although edited to have general application. Finally, the finding emerges once again that although there are mind-sets that are clearly different, there do not seem to be any simple co-variation of the mind-sets with who the respondent IS, or the self-stated patterns of involvement with the world of giving. It is that finding, a continuing revelation, which continues to surprise. The practice has always been to stratify the efforts by dividing people by WHO they are, assuming that people who appear similar on the criteria of who they are or how they involve themselves with the world of giving will be similar in their response to messages about giving. It just not the case.

Acknowledgments

These studies were run under the aegis of It! Ventures, Inc. The authors acknowledge the contribution of the late Hollis Ashman, as well as the contribution of Jacquelyn Beckley of the Understanding and Insight Group, New Jersey, USA. The studies were sponsored by Kathleen O’Grady of the O’Grady Foundation.

References

1. Nguyen C, Faulkner M (2020) In pursuit of effective charity advertising: Investigating the branding and messaging execution tactics used by charity marketers. Third Sector Review 26: 66-87.
2. Chen W, Givens T (2013) Mobile donation in America. Mobile Media & Communication 1: 196-212.
3. Duncan T (1995) Why mission marketing is more strategic and long-term than cause marketing. In: 1995 AMA Winter Educators Conference: Marketing Theory and Applications, Vol. 6, (eds: Stewart D, David W, Vilcassim N, Chicago: American Marketing Association) 469-75.
4. Chen S, Thomas S, Kohli C (2016) What really makes a promotional campaign succeed on a crowdfunding platform?: Guilt, utilitarian products, emotional messaging, and fewer but meaningful rewards drive donations. Journal of Advertising Research 56: 81-94.
5. Shearman SM, Yoo JH ( 2007) “Even a penny will help!”: Legitimization of paltry donation and social proof in soliciting donation to a charitable organization. Communication Research Reports 24: 271-282.
6. Luckow T, Moskowitz HR, Beckley J, Hirsch J, Genchi S (2005) The four segments of yogurt consumers: preferences and mind-sets. Journal of Food Products Marketing 11: 1-22.
7. Foley M, Beckley J, Ashman H, Moskowitz HR (2009) The mind-set of teens towards food communications revealed by conjoint measurement and multi-food databases. Appetite 52: 554-560.
8. Rabino S, Moskowitz H, Katz R, Maier A, Paulus K, et al. (2007) Creating databases from cross‐national comparisons of food mind‐ Journal of Sensory Studies 22: 550-586.
9. Moskowitz HR, Gofman A, (2007) Selling Blue Elephants: How to make Great Products that People Want Before They Even Know They Want Them. Pearson Education.
10. Moskowitz, H.R., Gofman, A., Beckley, J. & Ashman, H., (2006) Founding a new science: Mind genomics. Journal of Sensory Studies 21: 266-307.
11. Lundstedt T, Seifert E, Abramo L, Thelin B, Nyström Å, et al. (1998) Experimental design and optimization. Chemometrics and Intelligent Laboratory Systems 42: 3-40.
12. Likas A, Vlassis N, Verbeek JJ (2003) The global k-means clustering algorithm. Pattern Recognition 36: 451-461.

DOI: 10.31038/MGSPE.2021114

Abstract

As preparation for a negotiation involving the merger of two corporations through direct purchase, an experiment was conducted to determine whether the negotiations could be enhanced by understanding how ‘uninvolved third parties’ felt about the different aspects to be negotiated. These aspects were topics such as dividing shares of the merged company, and policy toward retaining employees and assets. The research effort was to assess the operational viability of a process which required about an hour from beginning to end to provide that ‘third party view of the issues. Aspects of the merger were surfaced and combined into vignettes comprising 2-4 element, and evaluated by an outside panel of respondents, unknown to the negotiating parties. The panel responses to the elements were deconstructed into the potential ability of each element to drive agreement (MERGE – YES) or disagreement (MERGE – NO). The process quickly revealed the elements on which there would probably be agreement, and elements over which there might be conflict. A segmentation of the test respondents showed two different mind-sets, uncovering types of sticking points for each mind-set.

Introduction

A great deal of the practice of the law involves negotiation and coming to an agreement. The negotiations may be left to the parties, to a professional negotiator/mediator, to the lawyers involved, and so forth. When there are opposing parties with different interests, how can negotiations be expedited, using knowledge, to reduce time, and reduce expense? Is there room for an application of the scientific method, which can provide a sense of what the parties can agree upon? Discussions with law professionals continue to suggest problems with ‘access to the law’ [1,2]. By access to the law is meant an easy, affordable, rapid way to get legal advice. Many lawyers are happy to give a free hour or so of consultation before they take on the case and request payment for their legal services. Despite this gesture, which is often welcome by businesspeople making deal as well as by parties seeking to sue another, the access to the law is not what it could be. The lawyer or the legal aid group must put time against the situation, understand it, and then decide whether there is a sufficient opportunity to monetize the time put against the effort. One unhappy consequence is that the ordinary small efforts are given short shrift. Sometimes the unhappy result is the oft-heard plaint ‘the only ones who made money from the situation were the lawyers.’ As denigrating as the statement might seem, it is hard to refute, especially when one tries to look at what the law provides for the small issues. The literature on negotiation, whether for business transactions or legal issues, continues to grow. The value of sensitivity in negotiation is obvious, and serves the negotiator well [3]. In fact, the importance of such sensitivity, and its practical application are subjects taught in law schools and business schools [4-6], as well as in the world of medicine. What then might be an appropriate technology or at least technique to introduce into the world of law to create a way to access the law? The approach might have to avoid taking up the time of a legal professional, because that defeats the purpose, especially when the case or situation is relative minor. The equivalent would be to find an approach to access knowledge in a set of printed material, without having to involve a librarian or even a legal assistant. In other words, the approach would have to rely on automatic computing, and analysis, to some people bordering or actually using ‘artificial intelligence.’ The idea of having technology assist in the negotiation process is not a new one. With the advent of computers and the recognition that there can be decision support systems of an electronic nature, interest has focused on the features of such a system [7-9].

The foregoing problem has been the focus of author HRM for 20 years, since 2001. The issues then, twenty years ago, were to understand how to evaluate the feelings of people presented with scenarios of a societal nature. The approach used by author HRM is called Mind Genomics [10]. Mind Genomics grew out of work beginning in 1980, trying to understand the patterns of preference of people towards foods, and the application of those patterns to the creation of commercial products [11]. The pioneering work, first with products eventually migrated into a variety of areas, some dealing with food, others dealing with social issues [12], and finally with the law [13], and with bigger issues in society [13]. The original efforts migrated to consulting projects in the legal and business areas, with work in different places around the world. It was clear from the projects that people from different countries often had markedly different styles of negotiating, an observation supported by published studies [14]. What was also interesting was the range of different responses to the same offers, suggesting the need to treat aspects of the negotiation process in a way which respects and understands these profound individual differences.

Demonstrating the Opportunity Through a Short Case History

During a meeting with lawyers in the Albany region of New York State, the opportunity emerged to demonstrate the approach. The topic was a merger of two companies. The opportunity was to identify how the owners of two companies could find an area of agreement. Separately from the private meeting with their lawyers negotiating the merger, the parties agreed to discuss the issues with author HRM, in an informal manner, and strictly for purposes of science. From the short, 10-minute background discussion, it become possible to ‘create’ a matrix of different issues, elaborating on the topics raised. At this point, the participants in the merger, here presented as Charles and Rebecca, respectively, returned to the meeting, after having given HRM permission to do a small demonstration ‘experiment’ using the material surfaced in the meeting. The relevant information was disguised where necessary.

Table 1 shows the set of four questions emerging from the discussion. The questions pertain to the topics of the merger. The 16 answers or elements present alternatives raised in the discussion, as well as several added by HRM afterward, based on the discussion, but not directly raised. The Mind Genomics process provides a template by which the researcher can quickly record the topic, the four questions, and the 16 answers, as shown in Table 1.

Table 1: The four questions and the four answers to each question, for the project pertaining to the merge of two companies.

Figure 1 shows three panels, each a screenshot from the actual project. The left panel shows the selection of the project name (business case Rebecca). The middle panel shows the four questions. The right panel shows the four answers to Question 1. The Mind Genomics program follows this right-hand panel with three additional panels (now shown), allowing for the remaining three sets of four answers each. It is important to note that the project can be set up ‘live,’ viz., in real time. The www.BimiLeap.com website is set up to guide and structure the thought processes, making the approach feasible in the middle of the meeting to gain quick feedback. The website can be freely accessed and easily used. Figure 1, as well as Figures 2 and 3, show the set-up of the experiment, requiring about 15-20 minutes at most.

Figure 1: The set-up screens for the Mind Genomics project showing the selection of the name, the list of four questions, and the set of four answers to the first question.

Figure 2: Screen shots showing communication to the respondent, including the third self-classification question (left panel), the rating scale and anchors (middle panel), and the short orientation which introduces the topic to the respondent.

Figure 3: The final user screen (left), and the two respondent screens (middle, self-profiling classification; right, sample vignette to be rated).

The next set of screens shown in Figure 2 instruct the user first to add a question pertaining to the respondent (left panel), then the rating scale, and finally the orientation to the study that the respondent will read. The Mind Genomics project is entirely private, so that the respondent is only identifiable by age gender and the third question.

Have you negotiated in the last five years in business?

1=no 2=yes 3=no but occasionally give advice 4=Not applicable

The rating scale provides the opportunity for the respondent to voice her or his opinion about the merger (whether or not the offer for merger will be rejected (rating = 1) or accepted (rating = 9)). This scale is called the Likert Scale, showing the magnitude of feeling. More recent practice has been to use a shorter 5-point scale. The scale is anchored at both ends, serving as a tool to show the respondent’s opinion AFTER the respondent has read the vignette, the test stimulus, described below.

Low Anchor: Rating question                                                   1=reject offer

High Anchor: Rating question                                                 9=accept offer

The orientation provides the user with a way to tell the respondent about the project. Creating the orientation is easy, but the user should be sure to provide as little specific information as possible. It will be the vignettes, small combinations of 2-4 messages which will provide the necessary ‘real’ information about the communications pertaining to the proposed merger.

Rebecca and Charles are merging companies. Here are negotiation suggestions. Read each screen as a suggestion and rate whether it be accepted by both by both Rebecca and Charles

Figure 3 shows the final screen of the user’s set-up experience (left panel), and then the respondent’s experience (middle and right panels, respectively). The user is given a set of options, to declare the study a business study or an academic study, to define the number of respondents to participate, and then to define the sourcing of the respondents. The study here involved the selection of 25 respondents, a sufficient and affordable number of respondents to provide necessary information about the ‘case.’ The user specified recruiting from the preferred provider (Luc.id, Inc.), and did not choose any specifics about the respondent. The final step is either to review and edit, or to ‘launch’ the study, and pay with a credit card.

The respondents are selected by country, and by other criteria available through the Luc.id system of literally hundreds of user-specified criteria. The respondents are sent notifications, and within minutes, many of the respondents from the ‘blast email’ participate. The respondent goes through two major steps for this study. The first step is completion of a self-profiling questionnaire (middle panel), which asks for gender, age, and response to the classification question shown in Figure 2 (left panel. The second step is the evaluation of 24 vignettes, set up like the vignette shown in Figure 3 (right panel).

The right panel of Figure 3 shows all of the information that the respondent needs to decide, but using information presented in an unusual way. Every one of the 24 vignettes that the respondent will see is set up the same way:

a. Orientation about the topic

b. Reminder to consider the entire vignette (viz., all the elements) as one idea

c. The rating question/scale

d. Three elements put together seemingly ‘at random,’ left justified

e. The response scale and the anchors

The vignette itself comprises 2-4 elements, combined according to an experimental design. The combination may look random, but the combination(s) is set up according to a strict structure called an experimental design [15]. Each of the 24 screens has a defined number of elements, and a defined listing of the specific elements to be incorporated. As a consequence, each of the 16 elements appears exactly five times, and is absent from 19 vignettes. Each question or grouping of four elements is allowed to contribute either one or no elements, but never two or more elements. This design feature means that for bookkeeping purposes, one should put into the same question two, three or four elements which are mutually contradictory. Finally, the 16 elements are statistically independent of each other, allowing for regression analysis. The novel part of the design is that by the correct permutation each respondent can evaluate the same ‘structure’ of vignettes, but the combinations are different. One can liken this metaphorically to the MRI, magnetic resonance imaging, which takes many pictures of the same tissue, all from different angles, and during the processing phase recombines them to arrive at a single in-depth image with 25 respondents, each viewing 24 DIFFERENT combinations, we end with 600 pictures of the topic, and the associated rating of that vignette or ‘picture’. The study was launched approximately 20 minutes from the start, although the novice may require at first 30-40 minutes to set up, and then launch. The actual data collection and basis, automated analysis, required 30 minutes. The results were ready for discussion approximately 60 minutes from the start, and available in printed form (easy-to-read EXCEL booklet). The speed and cost of the process are worth emphasizing before we look at the data. If nothing else, the process actually helped the merger negotiation by surfacing issues and the responses to the issues.

The First Experience with the Data – Average Ratings by Total and by Key Subgroups

Mind Genomics experiments generate a great deal of information, much of it usable. Our first analysis looks at the averages. We will look at complementary groups, shown in Table 2. The averages are computed on four measures

a. Rating = average of the 1-9 rating (1=merger offer not accepted .. 9=merger offer accepted)

b. TOP3 = a new binary variable, showing either strong acceptance (ratings 7-9) or all else (1-6)

c. BOT3 = a new binary variable, showing either strong rejection (ratings 1-3) or all else (4-9)

d. Response time in seconds = The Mind Genomics program measured the time from the presentation of

Table 2: Average values for ratings, binary variables and response times for Total Panel and key subgroups.

The test vignette to the time when the respondent rated the vignette. The authors’ experiences in a variety of studies suggest most response times of approximately 1.5-4.5 seconds for a vignette. Typically, response times of 8 or more seconds suggest that the respondent was multi-tasking. These longer response times (about 1/5 of the data) were simply eliminated from all analyses, but the remainder of the data from the respondent was kept.

Based upon Table 2 we see a simple story emerging when we look data from the total panel.

a. An average rating of 4.9 on the anchored 9-point scale, suggesting neither MERGER-YES (higher averages) or MERGER-NO (lower averages).. This may be due to most of the ratings clustering in the middle, or the decisions about equally divided between TOP3 (YES to the merger), and BOT3 (NO to the merger). Table 1 shows that the responses are divided about equally among YES (27%), NO (32%) and the rest MAYBE (100% – 27% – 32% = 41%)

b. The response time is short, about 1.6 seconds. The information in this merger is not difficult to comprehend and does not require much thinking. The information appears to be more emotionally driven than fact driven.

The self-profiling questionnaire allows us to identify respondents by gender, by age group, and by involvement in negotiations. Further analysis to uncover mind-sets (groups of people who think alike) reveal two mind-sets. These two mind-sets will be further explicated below. For the current analyses, it suffices to measure the average ratings for each of these defined groups. Table 3 suggests some differences, such as the fact that the younger respondents (ages 14-29) are far more negative about the prospects for the merger (BOT3 = 48), almost beginning with a negative attitude), and read the vignettes on average twice as quickly than do the older respondents (1.0 vs 2.1 seconds). For those with experience in negotiating, the average is overwhelmingly positive, and the time to read the vignettes is shorter. The two mind-sets differ from each other and are explicated below in depth.

Table 3: How the elements drive a third-party group (respondents) to feel whether there will be a merger (TOP3) or there won’t be a merge (BOT3).

Beyond Averages to the Stability/Instability of the Averages across the 24 Vignettes

A continuing issue in attitude research concerns how stable the responses are over time, especially when the respondent is evaluating many test stimuli. Practitioners have discovered the so-called ‘tried first bias’ [16], which means that the stimulus evaluated first may score aberrantly higher or lower than it would score when tried in the middle of a set of similar stimuli. This bias, sufficient to affect the validity of the data, has led to different ‘best practices’ such as testing only stimulus per person (so-called pure monadic), evaluating many products and rotating the order of the products to minimize the ‘tried first bias.’ The Mind Genomics system ensures that the respondents each evaluate a different set of vignettes, so that there is no tried first bias. Yet, there is always the possibility that the vignette evaluated first is biased, even though we cannot measure the effect of that bias due to the different combinations. Figure 4 shows two panels. The left panel shows the average TOP3 (merger = YES), and average BOT3 (merger = NO). The right panel shows the average response time. The graph shows the change in the averages across the 24 positions. Figure 3 does not suggest a systematic bias in the ratings for TOP3 or BOT3, although one might make a case for the variation in averages being at the start of the evaluation. The effect of repeated evaluations is far clear when the dependent variable is average response time. Over time the response times become shorter, presumably because at some point the respondent both knows what to do and responds more quickly when recognizing those elements which are important. It might an interesting study to compare different sets of messages around the same topic of mergers, to see whether the pattern of decrease of response time with experience in rating time is affected by the type of message.

Figure 4: The change in the average responses (TOP3 – merge; BOT3 – no merge; Response time) as a function of position in the 24 vignettes evaluated by the respondent.

Linking Elements to Response to Determine ‘What Messages’ Work

The most important aspect of the Mind Genomics effort is the ability to link together the elements and the responses, and by so doing discover what elements might be driving the response. The benefit of the Mind Genomics design is that cognitive richness of the test stimuli. Up to now we have simply looked at the pattern and surmised what might be happening. Up to now we had to be content with discovering that there are regularities in the data, such as the drop in the response time with increasing experience, or the difference in the average rating by key subgroup. For practical applications, such as study of the efficacy of messages, we must move beyond general patterns of responses, and into the specific elements themselves. The strategy of combining the messages by underlying experimental design ensures that that the combinations have some semblance of reality, and that the respondent cannot ‘game the system.’ The elements are combined in a way which precludes the respondent from changing the criterion of judgment. Such change of criterion may occur when the messages, the elements, are presented one at a time. The respondent might well adopt one criterion when the issue is division of ownership, and another criterion when the issue is which employees and assets to retain. By combining the elements into vignettes, Mind Genomics makes it virtually impossible for the respondent to adjust the judgment criterion. As explicated above, each respondent evaluated a unique 24 different vignettes, with the elements statistically independent of each other [17]. The underlying experimental design makes it feasible to use OLS (ordinary least-squares) regression to relate the presence/absence of the 16 elements to the newly created binary dependent variables, TOP3 and BOT3, respective, as well as Response Time. The equation deconstructs the newly created binary variables into the part-worth contribution of each element, as well as a baseline value the additive constant.

The equation is written as: Dependent Variable = k₀ + k₁(A1) + k₂(A2).. k₁₆₍D4)

The additive constant, also called the intercept, shows the expected value of the dependent variable (e.g., TOP3) when all 16 elements are absent. Of course, the experimental design ensures that every vignette comprises a minimum of two and a maximum of four elements, at most one element from each question. Thus, the additive constant is strictly theoretical, but does provide a sense of the baseline. Table 3 shows that the additive constant for TOP3 is 40, and the additive constant for BOT3 is 38. We conclude from that the basic likelihood is equal for votes for (TOP3) versus against (BOT3) the merger. It is in the coefficients where matters become interesting, informative. A positive coefficient means that including the element in a vignette will increase the vote, either for the merger (TOP3) or against the merger (BOT3). A negative or a 0 coefficient men that including the element in a vignette will not increase the vote, either for the merger or against the merger. In the interest of making the study simple to report, and patterns easy to spot It has become customary in Mind Genomics studies to report only the positive coefficients, and to highlight the strong positive coefficients, viz., those around 8 or higher. The negative and 0 coefficients do not tell us much. For TOP3 they tell us the strength of failure to push for TOP3. When we are really interested in the elements which actively drive away agreement (Merger – NO), we are better served by looking at the coefficients for BOT3. Positive elements for BOT3 are those which actively drive away agreement. Table 3 presents the positive coefficients for TOP3 and for BOT3, respectively. When an element fails to have a positive coefficient for either TOP3 or BOT3 the element does not appear. In this way it becomes easier to see the patterns. The data suggest that the there is an equal proclivity for Merger and No Merger. The elements which push for a merger are those about the way the merger will combine the companies. The elements which push away from a merger are those about ownership. It becomes clear that control is a major issue, as perceived by an outside group of people evaluating the different propositions for merger. The data do not mean that these are the actual issues that will be discussed, but rather perceived to be potentially contentious.

Mind-Sets and Negotiations

If we were to stop at the results in Table 3, the effort to understand the ‘sticking points’ of the merger would have emerged, in a matter of 30 minutes, from a small group of 25 respondents acting as ‘consultants,’ albeit unknowingly since their job was to evaluate the likely outcome of a set of discussion points. We could stop here and have our job more or less compete. Yet, there is more to be learned. That ‘more’ is the discovery of different ways of looking at the same information and arriving at different decisions. These different way are called mind-sets. Mind-sets emerging from segmentation have been a hallmark of marketing for decade [18], and is now interesting, or even better carving out new areas of the practice of business and law. Even with as few as 25 respondents it is possible to discover meaningful mind-sets. The researcher creates individual-levels, two per respondent, one for TOP3 vs elements, and the other for BOT3 vs elements. The models do not have an additive constant. The database comprises 25 rows (one per respondent), with 32 columns (16 for TOP3; 16 for BOT3). The numbers in the body of the data matrix are coefficients. The researcher clusters the respondents into two groups, based upon pattern of the coefficients. Within each cluster or mind-set are respondents whose pattern of 32 coefficients are ‘similar to each other, and dissimilar to the patterns of the 32 coefficients generated by respondents in the other cluster or mind-set [19]. After all is done, Table 4 reveals two clear mind-sets. Both mind-sets are equal in their desire for a merger, with the additive constants of 37 and 40. It is the elements which are important. Mind-Set 1 wants equity in the division. Nothing really turns off Mind-Set 1, viz., there is nothing driving BOT3. In contrast, Mind-Set 2 wants a rational merger, and is turned off by either unequal division of stock, or loss of control. The important thing about Table 4 is a sense of the fine-grained needs of the different mind-sets, setting the agenda about what to discuss, and what to ‘take off the table.’

Table 4: Strong performing elements for the two emergent mind-sets, created the combination of the 16 TOP3 coefficients, and the 16 BOT3 coefficients.

A Shortened, Which is Both Inclusive (Participatory) and Objectively (Data Centric)

The analysis above suggests a rich database can be developed quickly, viz., in less than 60 minutes, from start to analysis. The nature of the Mind Genomics approach forces the use into a disciplined presentation of the ‘case.’ Some may consider the speed and the concomitant ‘structuralizing’ of the process as a negative, viz, that those involved may be forced to study the topic without having a chance to think deeply about the topic. This criticism is absolutely correct. The spirit of the Mind Genomics process is founded on the alluring combination of structure, speed, and depth. The very design, as a computer-based app, with almost automatic front-to-back effort, and an automated basic analysis, prevents deep thinking, at least at the time of the evaluation. The focus is on pulling out the salient ideas, putting them into a template, involving a third part as judges in a way which prevent judgment biases, and h nth return with structured data. What might be the way such a system could be used in the world of the everyday? The first use is a subtle one. The structure forces the people involved to think about alternatives or options facing them. The user must contribute the question and the four elements for each question. The thinking, therefore, is to support one’s position, but rather to focus on the different aspect of the topic. Ongoing work with Mind Genomics suggests that simply requiring the participants to offer ideas in a structured manner improves their thinking. There is a second benefit as well. That benefit is the ability to identify what specifics work, and whether there exist hitherto unknown or only suspected mind-sets of individuals having different points of view [20]. Such information is important to the people involved in the case because it demonstrates the very real possibility that there are different ways to approach the same topic. The disagreements between people become more explainable. Even more promising, however, is the possibility of finding ideas which are very acceptable to one mind-set and to another, or at least ideas which are acceptable to one mind-set, and do not turn off the other. There is a third benefit, perhaps the most important. That benefit is improved access to the law, something being regularly recognized as a major need. Howard [21-25]. With an opposing party threatening to sue, or at least to damage by driving up legal fees, there is a need for rapid, inexpensive DIY (do it yourself) methods. It is quite possible that Mind Genomics might be one of those methods, a simple DIY system, executed collaboratively by the different groups involved in the negotiation, leading to speedier, fruitful negotiations, filled with mutual understanding, less expensive, and ultimately being far more productive.

References

1. Frandino J (2018) (Personal communication).
2. Moskowitz H, Wren J, Papajorgji, P (2020) Mind Genomics and the Law. LAP Lambert Academic Publishing.
3. Arunachalam, V, Dilla WN (1995) Judgment accuracy and outcomes in negotiation: A causal modeling analysis of decision-aiding effects. Organizational Behavior and Human Decision Processes 61: 289-304.
4. Chetkow-Yanoov B (1996) Conflict-resolution skills can be taught. Peabody Journal of Education. 71: 12-28.
5. Gettinger J, Koeszegi ST, Schoop M (2012) Shall we dance?—The effect of information presentations on negotiation processes and outcomes. Decision Support Systems 53: 161-174.
6. Nadai E, Maeder C (2008) Negotiations at all points? Interaction and organization. In Forum Qualitative Research (online), 9: 1-19.
7. Julian V, Sanchez-Anguix V, Heras S, Carrascosa C (2020) Agreement Technologies for Conflict Resolution. In Natural Language Processing: Concepts, Methodologies, Tools, and Applications pg: 464-484. IGI Global.
8. Kersten GE, Lo G (2003) Aspire: an integrated negotiation support system and software agents for e-business negotiation. International Journal of Internet and Enterprise Management. 1: 293-315.
9. Neijens P, Swaab R, Postmes T (2004) Negotiation support systems: communication and information as antecedents of negotiation settlement. International Negotiation 9: 59-78.
10. Moskowitz HR, Gofman A, Beckley J, Ashman H (2006) Founding a new science: Mind genomics. Journal of Sensory Studies 21: 266-307.
11. Gladwell M (2004) Choice, happiness and spaghetti sauce.
12. Moskowitz HR, Gofman A (2007) Selling Blue Elephants: How to Make Great Products that People Want Before They Even Know They Want them. Pearson Education.
13. Kover A, Moskowitz H, Papajorgji P (2020) Applying Mind Genomics to Social Sciences, Book submitted, In Review.
14. Gabrielidis C, Stephan WG, Ybarra O, Dos Santos Pearson VM, Villareal L (1997) Preferred styles of conflict resolution: Mexico and the United States. Journal of Cross-Cultural Psychology 28: 661-677.
15. Ryan TP, Morgan, JP, (2007) Modern experimental design. Journal of Statistical Theory and Practice 1 ; 501-506.
16. Malhotra N (2008) Completion time and response order effects in web surveys. Public Opinion Quarterly 72: 914-934.
17. .Gofman A, Moskowitz H (2010) Isomorphic permuted experimental designs and their application in conjoint analysis. Journal of Sensory Studies 25: 127-145.
18. Wedel M, Kamakura WA (2012) Market segmentation: Conceptual and methodological foundations (Vol. 8). Springer Science & Business Media.
19. Likas A, Vlassis N, Verbeek JJ (2003) The global k-means clustering algorithm. Pattern Recognition 36: 451-461.
20. Sternberg RJ, Soriano LJ (1984). Styles of conflict resolution. Journal of Personality and Social Psychology 47: 115-126.
21. Howard D (2001) The Law School Consortium Project: Law Schools Supporting Graduates to Increase Access to Justice for Low and Moderate-Income Individuals and Communities.” Fordham Urb. LJ 29 (2001): 1245.
22. Nader L (1980) No Access to Law: Alternatives to the American Judicial System (pp. 64-67). New York: Academic Press.
23. Trebilcock M, Duggan A, Sossin L (2018) Middle Income Access to Justice. University of Toronto Press.
24. Ware SJ (2012) Is adjudication a public good: Overcrowded courts and the private sector alternative of arbitration. Cardozo Journal of Conflict Resolution 14.
25. Wren J (2021) (Personal communication).

DOI: 10.31038/MGSPE.2021113

Abstract

In a rapid, affordable, and scalable experiment, 50 respondents each evaluated unique sets of 24 ‘messages’ dealing with ‘offers to switch mobile phone providers’. The focus of the study was to show what could be learned from a simple experiment dealing with an everyday topic. For each test vignette of 2-4 messages, the respondents rated both likely to switch, and selected an emotion. The analysis revealed the emotional signature of each element, showing the feeling(s) most associated with each element, as well as the degree to which the element or message was rated as driving the respondent to switch providers. In terms of convincing respondents to say that they would be likely to switch providers, no elements performed strongly for the total panel. Only after clustering the respondents into four mind-sets by the patterns of their responses to the elements did the opportunities emerge, corresponding to specific messages. The paper shows the power and contribution of Mind Genomics to understanding a person’s person’s decision criteria, as well as providing immediate guidance to solve a practical problem. The attractiveness of the approach comes from the combination of the power of the approach, combined with the practical benefits of simplicity, speed, and affordability.

Introduction

Today (2021) it would be no exaggeration to say that the smart phone is ubiquitous. The millennial generation has grown up with the smart phone. It is unimaginable to many that there could have been an era when the telephone was just that, a device into which people spoke to other people either on the next street, or more rarely in the next state, or even more remarkably (and with great expense) people living in a foreign country. Those who are old enough to marvel at the change in technology like to compare the technology of today’s smart phone with the available technology in 1969, when John Glenn went to the moon and back.

As a consequence of the accelerated acceptance of the smart phone, the phone itself has become a commodity. What attracted attention 30 years ago in the 1990’s is now a part of everyday life? And of course, with the ubiquity of the smartphone is the ubiquity of the provider. The smartphone is everyone’s entry point into today’s powerful technology. Virtually world-wide people can do things such as make video calls at low price, and with the abandon of something which is virtually free.

A search through today’s literature (Google Scholar®, done October 18, 20201) for the topic of ‘switching mobile phone providers’ showed an astonishing 57,800 hits, and changing the word from provider to carrier reduced the number of hits to 38,000. The numbers become even more remarkable when we break up to ‘hits’ for mobile phone providers, doing the analysis in by five-year periods, and then compute the approximate hits/month. Table 1 shows these statistics.

Table 1: Hits for ‘switching mobile phone providers’.

Clearly the academic literature reflects the interest in mobile telephony, and the services provided. What is just as interesting are the topics. What emerge from the literature search is what might be expected, namely studies of the topic in different countries, namely what are the important aspects to which people attend. Here are two representative titles of papers. The focus is on the general process of how people think about the issue of switching:

Switching behavior of mobile users: do users’ relational investments and demographics matter? [1]

Drivers of brand switching behavior in mobile telecommunications. [2]

The search through many of the Google Scholar® hits reveals two patterns. The first pattern is the academic effort to simplify the topic of switching into a set of actions, or need states, viz., the need to systematize understanding. The second pattern, far more frequent, is the analysis of switching behavior, along with motivations for doing so, in the many countries around the world. This second effort gives the reader an awed sense of the power, and the ubiquity of mobile telephony.

What is missing, however, is the practical understanding of the type of messaging which drives consumers to feel that they would switch mobile phone providers. The literature may contain some of these messages, but the focus is generally from the ‘outside in’, viz., looking at the patterns of behavior in the increasingly important world of mobile telephony. Like so many other areas, the focus ends up rarifying the specifics, the messages, which become ephemera, to be discarded in the search for lasting ‘truth’ or at least ‘general patterns.’

The focus on Mind Genomics is on these ephemera, specifically the messaging. The objective of Mind-Genomics, as emphasized below, is the understanding of the messaging, and by so doing, understanding the topic from the ‘inside-out’, from the mind of the customer faced with the array of messages, and the competing cacophonies of merchants hawking their wares, shouting their offers.

Faced with this topic, the experiment reported here, done 10 years ago, is still relevant. The technology may have changed, but as we will see, the minds of people then made sense. Nothing discovered a decade ago has not changed very much, nor surprises very much. In this paper we resurrect data a decade old to show how understanding the way the consumer mind works provides data which has a very long shelf-life. Underneath the technology is the benefit, the appreciation of that which does not change from year to year.

Ways to Solve the Problem

Better messaging to convince buyers is a hallmark topic of consumer research. The publications one sees in the scientific and business literature about the topic is dwarfed by the amount of information retained in the archives warehousing corporate research. It should come as no surprise that the proper messaging of a company’s offer is a key to attracting customers, at least new customers, who have almost no other opportunity to know what is available unless they are told.

The above being said, it is rare that a corporate executive will truly know the messages which appeal to customers. There may be some answers of an obvious nature, but the reality is that most of the corporate knowledge is based upon such things as ‘this is the way we have advertised before…. it works. …let’s not take a chance, let’s not change.’ Of course there are situations such as the recent Covid-19 pandemic which has changed the way people behave, but there is the seemingly eternal reticence to explore new ideas.

The typical approach to testing messages, so-called ‘promise testing’, evaluates the messages one at a time, looking for messages which either simply score well, or upon probing, appear to convey messages of the right tonality. It is often averred by the advertising agency and by marketing gurus that one requires a sensitive and developed ear and mind to ‘know’ what will succeed in the marketplace, and that simply testing many messages does not reveal the truly breakthrough idea. The result of such statements is the continuing fight between the artist who ‘knows’ what will work, and the researcher, who must test to know what will work. The artist feels that one or two of the ‘right talented’ individuals, like the copyrighter, can do the job. The researcher feels that it will require a representative group of target consumers to ensure that one is making the correct choice.

The Mind Genomics Process

The Mind Genomics process has evolved over the past thirty years, during which time is has evolved into the beginning of a science, with many published paper, and an increasing cadre of practitioners around the world [3-6].

The Mind Genomics process is a systematized, templated process, requiring the researcher to break up the problem into a series of small, easily managed steps. By so doing, Mind Genomics forces the researcher (really the user) to think in both a creative and a structure way, respectively. The steps below, moving from thinking to discovery, were done over a period of two days. Today, a decade later, the process would be reduced to about two hours.

Step 1 – Create Materials

Table 2 shows the set of four questions and six answers for each question. Mind Genomics forces the researcher to think a structured fashion, beginning with a topic, continuing with a set of questions, and then for each question a set of alternative answers. The Mind Genomics system comprises a variety of different experimental designs, layouts or sets of combinations comprising a specific number of ‘questions’ and for each question a set required number of ‘answers’. The array shown in Table 2 is called a 4×6 (four questions, each with six answers). Today’s practice has been reduced to the much easier and faster design, the 4×4 (four questions, each with four answers).

Table 2: The four questions and the six answers for each question. Brand names are disguised.

The questions in Table 2 ‘tell a story’. There is no need for the researcher to deeply understand the topic in order to develop the questions and the answers to the questions. Rather than forcing the researcher to select the answers that will be best, thus delaying the process until everything is ‘just right,’ the Mind Genomics process has been designed to be simple, affordable, and iterative. The researcher is encouraged to ‘just do it,’ find the results, and ‘do it again, changing what didn’t work with new guesses. Thus, Table 1 presents a ‘first guess.’ With ‘n’ cycle time of 1-2 hours, the second iteration would see some new questions and answers replacing the ones which performed poorly, viz., simply ‘did not convince the respondents’.

As one might surmise, the hardest part of this first section is coming up with the four questions which ‘tell a story’. The questions require the researcher to think more deeply and critically about the topic. When presented with the notion about asking a question, most researchers begin with a question that can be answered yes or no, or with some specific one-word answer. It takes a while for the researcher to think in terms of questions which require a phrase as an answer, rather than a simple no/yes. In contrast, once the questions are asked, the answers in the form of a declarative phrase are easy to create. The questions provide the structure for the answer.

Step 2 – Creating an Experimental Design Which Mixes and Matches 2-4 Elements in Each Element

The hallmark of ‘field work’ in Mind Genomics is the evaluation of combinations of elements (so-called vignettes), with these vignettes systematically composed according to an underlying experimental design [7]. The design specifies the composition of each vignette. No effort is made to link together the elements of the vignette, an example of which appears in Figure 1.

Figure 1: The orientation page to the Mind Genomics study on purchasing a mobile phone.

Each vignette comprises a specified number of 2-4 elements. Each respondent evaluates 48 vignettes. The mathematical structure of the 48 vignettes is the same from one respondent to another with the mathematical structure ensuring that the 24 elements are statistically independent of each other. The Mind Genomics system presents each respondent with a totally new set of combinations of the 24 elements but a set of combination following the SAME underlying structure. Although the structure remains the same, the actual combinations different. Rarely do respondents ever test the same combinations. That difference in combinations is a property of the underlying design [8].

Step 3 – Execute the Study, through a Third Part On-line Field Service

Since the early years of this century, on-line panel providers have made available respondents to participate in studies conducted on the internet. What was unusual in the late 1990’s is today the ‘norm.’ Most people have been invited to participate in a variety of ‘studies’, whether these studies deal with limited topics such as the satisfaction of their last transaction with a company, or the studies deal with longer, more involved topics conducted as polls. In contrast, the Mind-Genomics process can be thought of as an experiment conducted on the internet, but in the form of a simple set of answers to systematically varied stimuli.

The respondents received an invitation to participate in the study, which would last about 10-15 minutes. The respondents interested in the study pressed on the linked embedded in the email invitation, and were taken to the study. The study began with an orientation, which is shown in Figure 1. The orientation screen used at the time of the study in 2012 was substantially longer than the orientation screen used today. Of special interest is the effort to reassure respondents that they are evaluating different vignettes. This effort to communicate that all the vignettes are really different from each came from a few complaints from professional and students that the vignettes seemed all the same. The early efforts in Mind Genomics focused on establishing the usefulness of the approach, one way of doing so being an effort to anticipate problems and avoid them. Thus the effort to reassure the respondent that the vignettes all differ from each other. Another remnant of the early effort, still in force, is the reassurance that the evaluations will last 10-15 minutes. This effort comes from the complaint from some professionals (but not panel respondents) that the effort is ‘overly long’, and that ‘how much is left?’

The IdeaMap program presented the respondent with a set of 48 vignettes, each vignette to be rated on two scales. The first scale instructed the respondent to rate the vignette on likelihood to switch to the provider. The second scale instructed the respondent to select one of five emotions experienced after reading the vignette.

Figure 2 shows an example of the vignette, set up for Rating Scale #1 (how likely are you to switch to this mobile service provider?). The combinations of elements are dictated by the specific permuted design for the respondent. The program makes no effort to beautify the combination, viz., by providing connections between the elements. The elements are presented as centered phrases, in unadorned fashion. Despite the apparent starkness of the stimulus, few respondents complain. Rather, over the 40+ years that this format has been used (since 1980), many respondents have made the unsolicited comment that the format actually helped them to ‘scan’ the vignette, and make their decision.

Figure 2: Example of a vignette set up for Rating Scale #1.

For each vignette, the respondent assigned two ratings, for the likelihood of switching, and then for the selection of emotion. The vignette remained the same. As soon as the respondent rated the vignette by selection the closest emotion to what was being experienced, the vignette closed, and the next vignette was immediately presented. At the end of the evaluation, the respondents provided answers to four classification questions, including gender, age, and two on patterns of usage.

From Vignette to Mind – The Templated Analyses of Mind Genomics

The focus of Mind Genomics moves to the elements. The vignettes are only a convenient way to ensure that the elements are presented in a more typical fashion, approximating the typical type of offer, rather than being presented one-at-a-time. Presenting vignettes, all differing from each other, makes it almost impossible for the respondent to be politically correct, to ‘game the system,’ and provide the answers that one might deem to be ‘the right answer.’

Given focus on individual elements, Mind Genomics moves from the combinations, the vignettes, to deconstructing the vignettes into the contributions of the individual elements. Recall that each respondent evaluated 48 different combinations, and that the elements appeared 2-4 times in the combinations. Furthermore, each element appeared an equal number of times, and the elements appeared in an uncorrelated fashion.

The analysis begins by creating a data matrix, each row corresponding to one vignette from one respondent. The matrix comprised a column to identify the respondent, 24 columns corresponding to the 24 element, and two final columns corresponding to the rating assigned on Rating Scale #1 and Rating Scale #2, respectively. The final four columns contained the answers to the self-profiling questions (age, gender, and two questions about phone use).

The data matrix comprising 1’s (element present in vignette) and 0’s (element absent from vignette) presents the 24 elements as so-called dummy variables, absent or present. There is no metric information about the dummy variables. The objective of the analysis is to determine the degree to which the element drives estimated switching (Rating Scale #1) or links with certain feelings/emotions (Rating Scale #2).

The actual data matrix comprises a set of 48 rows for each of the 50 respondents, or 2400 rows of data.

The 9-point rating for Rating Scale #1 (1=not likely to switch at all … 9=very likely) is converted to a binary scale, with ratings of 1-6 converted to 0, and ratings of 7-9 converted to 100. To each of the newly created binary numbers is added a very small random number (< 10^-5).

The rationale for converting the 9-point scale to a binary scale is the proclivity of users of data to demand simple yes/no statements, viz., will the respondent switch or not switch, based upon the elements in the vignette? It is technically correct to say, ‘the data shows a rating of 7, closer to switch and further away from not likely to switch’. That answer is not useful in a business situation, where the answer should be all or none. The 9-point Likert scale could be replaced by a simple binary scale (no/yes) at the outset, but there is always interest in precision for other analyses that may be of interest.

The same type of transformation is done for the emotions, except that five new binary variables are created, one each for curious, interested, positive, hesitant, and uncomfortable, respectively. For each vignette, the emotion selected in Rating Scale #2 is given the value 100, and the four emotions not selected in Rating Scale #2 are each given the value 0. Again, and afterwards, each of the five newly created emotion variables has another vanishingly small random number added.

The rationale for adding a small random number to each newly created binary variable is ensure that the binary variables exhibits some small degree of variation each at the individual level. Were a single respondent to rate all the vignettes 1-6, for instance, or select the emotion ‘hesitant’, the conversion would transform all of the respondent’s ratings into the same value. The OLS (ordinary least-squares) regression would fail. Adding a the vanishingly small random number is a prophylactic step, not affecting the results, but protecting against a crash of the regression program used to relate the elements to the ratings.

Linking Emotions to Elements

Our first analysis focuses on the link, if any, between the element and the selection of an emotion. Recall that the respondent selected one feeling/emotion for each vignette. The analysis is straightforward, promoted by the foresight of creating the vignettes according to the permuted experimental design.

The analysis creates five OLS (ordinary least squares) regression equations, each expressed in the same way: Linkage (to an emotion) = k₁ (A1) + k₂ (A2) … k₂₄ (D6)

The foregoing equation says that the linkage between the feeling/emotion and be expressed by a simple equation, which shows the linkage of each element to the emotion. Coefficients (k₁ to k₂₄) are estimated using OLS regression, without estimating the additive constant. Coefficients of about 10 or higher are statistically significant and relevant, based upon previous observations across many projects using Mind Genomics conjoined with ratings of emotion.

For our presentation here, and to allow the strong patterns to emerge, we show the data from the total panel, but show only those coefficients or linkages 10 or higher. Table 3 shows those strong linkages. The elements not appearing at all in the table are those which do not show a strong linkage to the feeling/emotion. nt.

Table 3: Linkage between elements and feelings/emotions. Only strong linkages of 10 or are shown.

The important thing to note is that improvement in our understanding of the elements, simply by learning the linkage of emotions and elements. From the entire array of 50 x 48 or 2400 vignettes we see that despite the imagined difficulty of the task, the linkages exhibit face validity, making sense.

The elements which are interesting are pricing.

The elements which are strongly positive are first service, and then features.

Sometimes an element links to a positive element and to a slightly negative element (hesitant). For some people the element (Reasonably priced extended warranties for all of our phones) provokes the feeling of interested, for others the same element provokes the feeling of hesitant. There may be different mind-sets among the respondents, viz., and ways of thinking.

Finally, one message actually makes the respondent feel uncomfortable (Star Wars lover… Our new (Product 2) has an awesome Star Wars theme)

Linking the 24 Elements to the Likelihood of Switching

We now return to the original focus of the study, viz.., what messages, if any, are likely to get a person to consider switching mobile phone providers. Although the study focused on mobile phone providers, the question is universal in the world of business. The Mind Genomics process provides an approach to answer the question, doing so quantitatively and efficiently.

The analysis once again begins with a transformation, this time with ratings of 1-6 transformed to 0, rating 7-9 transformed to 100, and a vanishingly small random number added to each of the transformed ratings. There is not fixed about the criteria of transformation, but the bifurcation of 1-6 and 7-9 has been the standard one for decades. Sometimes the division is at 7 (1-7 transformed to 0; 8-9 transformed to 100). This is done for respondent populations which tend to up-rate vignettes, and corrects for the exceptionally large number of positive responses.

The analysis creates one OLS regression equation, of the form: Likely to switch = k0 + k₁ (A1) + k₂ (A2) … k₂₄ (D6)

This time the equation has an additive constant. The additive constant, k₀, is a measure of the likelihood to switch in the absence of elements. Of course, by design all the vignettes comprised 2-4 elements, so there are no vignettes without elements. Nonetheless, the additive constant gives a good sense of the likely reception of one’s offers, information valuable to have in a marketing campaign. Without a Mind Genomics experiment of this sort, one would have to ask a respondent directly, or mine the switching data of the respondent. With Mind Genomics the additive constant convenient provides this measure of proclivity to switch. The above-mentioned equation is calculated at the level of the group, with the group defined by total, by specific ages, and by gender, respectively.

When the topic is switching, the information emerging from the analysis suggests the following findings, information that would be useful both to the marketer facing the business problem, but also to the researcher trying to understand what motivates people. Table 4 shows the relevant elements of the model for five groups; total, two genders, two age groups, respectively. There are only 42 of the 50 respondents shown in the age groups. The remaining groups, younger and older, did not comprise a sufficient number of respondents to show.

We begin with the additive constant. As noted above, the additive constant is the estimates likelihood of switching providers in the absence of messages, and should be considered a baseline. Table 4 shows a basically low likelihood of switching in the absence of a compelling message, with the additive constant of 20 for the total panel. Females are more likely to switch than males are (constant 29 vs 21). The age groups are similar, although the older respondents are slightly less likely to switch.

At this point, the common criticism is the small base size. With larger base sizes the additive constant will remain the same, with the usual ‘variability’ encountered with subjective data. What is important is that a 3-4 hours excursion into an experiment suggests topics, messages, and even opporgtunities, perhaps even not hitherto expected or perhaps conjectured but not demonstrated.

The body of Table 4 shows only those elements which generate a coefficient of +6 or higher for any one of the five groups. Only four elements do so, with the male respondents being most positive. For the total panel and for females no elements generate strong performing coefficients.

The low additive constant and the lack of strong performing elements among the total panel and key demographic subgroups are not the results of a low base size. That is, increasing the number of respondents from 50 to 100 or even 200 or 500 is not likely to produce results too different from what we see in Table 4. We conclude, therefore, that there are simply no elements which really drive switching, and that the team must go back to create new messaging. It is better to find this information out in an hour or two than in a month or two.

Table 4: Additive constant and strong performing elements for total panel, gender, and age. Only elements showing a coefficient of +6 or higher are shown.

There is, however, another possibility, viz., that there are different ways of thinking about the offers, ways which do not emerge when we just know age and gender. This is known as mind-sets,

Clustering uncovers hitherto unexpected groups of people in the population, mind-sets in the language of Mind Genomics. Within a cluster the respondents see the world similarly, at least the world of offers regarding switching mobile phone providers. Clustering does not pretend that these mind-sets are actually fixed in stone. Rather, clustering is an analytic ‘heuristic’, trying to make sense out of variation which inevitable occurs in data concerning choices. The clustering provides insights which would otherwise not emerge

The mechanics of clustering is straightforward. There are different ways to cluster data, all of which are equally ‘correct,’ but simply a matter of decision. Clustering attempts to uncover groups in the data, not based on who the groups ARE but rather on how the groups perform.

The mechanics of clustering begins with the generation of the information on which the clustering will be done, such information obtained at the level of the individual respondent. The subsequent analysis creates the clusters or mind-sets. Recall that each respondent tested the same structure of 48 combination, prescribed according to by a single experimental design that was permuted to create new combinations evaluated by each respondent. . The design allows for the estimation of individual-level models or equations, just as we estimated the group model. Thus, the foregoing equation with the additive constant is created at a respondent-by-respondent level to provide a matrix of 50 rows, one per respondent, and 25 numbers in each row, the additive constant and the 24 coefficients. That data matrix is then analyzed by through clustering, using the 24 coefficients (but not the additive constant) to define first two clusters, then three clusters, then four clusters of respondents. A cluster comprises individuals whose patterns of coefficients are similar to each other, and quite dissimilar to the coefficients of the respondents in the other clusters [9].

Table 5 shows new opportunities for messaging when we break the respondents into mind-sets based upon the pattern of responses to the messages that a mobile provider would likely use With 50 people the objective is not to create the science of messaging for this topic, but rather at a tactical level to identify messages which seem to work. There are certainly no promising messages when we look at the Total Panel. When we move to two mind-sets we three elements emerging. When we move to three mind-sets we see seven elements emerging. When we move to four mind-sets, we also see the same seven elements emerging, and some very strong performances. The key group on which to focus is Mind-Set 4F, first because it is the largest mind-set (19 respondents), and second because it has the highest basic likelihood to switch, based on the additive constant of 29. Table 4 is sorted by Mind-Set 4F of the four mind-sets. These are the price-focused, and the serious users. The key messages are:

A2: Check service problems online! Our online support staff can keep you updated on all Service problems.

C1: Individual plans start as low as $39.99 a month

The offering can be improved by choosing a message which also appeals to Mind-Set 4G, those who like to explore, but are not technically adept (not ‘techy’)

A5: Check your phone and voice your concerns at any of our retail stores

Table 5: Additive constant and strong performing elements for total panel, two complementary mind-sets, three complementary mind-sets, and four complementary mind-sets, respectively. Only elements showing a coefficient of +6 or higher are shown.

Discussion and Conclusions

Everyday life is replete with opportunities to understand the way people make decisions. The common approach is to look at the problem from the ‘outside-in,’ searching for regularities, and patterns. This approach characterizes a great deal of what we know about consumers. The academic literature focuses on the pattern, the generalities, the so-called ‘nomothetic’, coined from the Greek word Nomos, pertaining to the general, the normative.

We can trace this focus of outside-in to the development of science, where the focus is on discovering patterns, and where there was no ‘mind’ to report the experience, other than the mind of the researcher. This attitude of searching for patterns is important in the world of science, where the focus is on discovering patterns in a nature which has no ‘communicating mind.’ The reality is that searching for patterns, running experiments and measuring results, are the only ways of making sense out of nature which is mute, but lawful.

The opportunity learns about patterns of thinking and patterns of behavior are much different when we work with people who can talk. Two different measures emerge. The first is what people say they will do, and second is what people actually do. Up to now the focus of ‘real science’ has been on measuring what people actually do. That measure is considered the ‘real’ information. What people say they do is cast off as attitudes, something to measure, but not necessarily something on which to establish a science. It is precisely the patterns of what people ‘say they will do’ with different stimuli and in varying situations which constitutes the basis of Mind Genomics.

At the practical level, the data just shown suggests a richness of understand to be had of the world of the everyday by doing the simple experiments prescribed by Mind Genomics. The data may well enhance business performance on the one hand, as it enhances our knowledge of people and motives on the other. Examples include studies on attendance at museum by teens [10], and the recognition that entire world of new knowledge awaits the Mind Genomics researcher [11,12].

Acknowledgment

The author wishes to thank Professor Martin Braun of Queens College, and Professor Sue Henderson of New Jersey City University, who were instrumental in the work at Queens College leading to these Mind Genomics studies. The studies were done by the Ms. Janna Kaminsky and the late Stephen Onufrey, in Math 110.

References

1. Ranganathan C, Seo D, Babad Y (2006) Switching behavior of mobile users: Do users’ relational investments and demographics matter? European Journal of Information Systems 15: 269-276.
2. Grigoriou N, Majumdar A, Lie L (2018) Drivers of brand switching behavior in mobile telecommunications. Athens Journal of Mass Media and Communications 4: 7-28.
3. Moskowitz HR (2012) ‘Mind Genomics’: The experimental, inductive science of the ordinary, and its application to aspects of food and feeding. Physiology & behavior 107: 606-613. [crossref]
4. Moskowitz HR, Gofman A, Beckley J, Ashman H (2006) Founding a new science: Mind genomics. Journal of Sensory Studies 21: 266-307.
5. Porretta S (2021) The Changed Paradigm of Consumer Science: From Focus Group to Mind Genomics. In Consumer-based New Product Development for the Food Industry Royal Society of Chemistry 21-39.
6. Porretta S, Gere A, Radványi D, Moskowitz H (2019) Mind Genomics (Conjoint Analysis): The new concept research in the analysis of consumer behaviour and choice. Trends in Food Science & Technology 84: 29-33.
7. Lundstedt T, Seifert E, Abramo L, Thelin B, Nyström Å, et al. (1998) Experimental design and optimization. Chemometrics and Intelligent Laboratory Systems 42: 3-40.
8. Gofman A, Moskowitz H (2010a) Isomorphic permuted experimental designs and their application in conjoint analysis. Journal of Sensory Studies 25: 127-145.
9. Fraley C, Raferty AE (1998) How many clusters? Which clustering method? Answers via model-based cluster analysis. The Computer Journal 41: 578-588.
10. Gofman A, Moskowitz HR (2010b) Improving customers targeting with short intervention testing. International Journal of Innovation Management 14: 435-448.
11. Milutinovic V, Salom J (2016) Mind Genomics: A Guide to Data-Driven Marketing Strategy. Springer.
12. Gofman A, Moskowitz HR, Mets T (2011) Marketing museums and exhibitions: What drives the interest of young people. Journal of Hospitality Marketing & Management 20: 601-618.