DOI: 10.31038/ASMHS.2019354

Abstract

We present a new way to understand the mind of the investor, moving away from technical models of investing to determining in simple terms of type of information that the person feels to be important when deciding to invest. The experiment identifies the relative importance of different types and sources of information to which a person might pay attention. The approach, Mind Genomics, combines simple messages or elements to create vignettes, concepts which present several different aspects of the news and information. The respondent rates these different combinations in terms of the degree to which the combination reflects the respondent. Mind Genomics reveals three clearly different mind-sets of individuals, from those who focus on the source of the information, those who focus on the story and information, and those who focus on what their friends are saying and their own insights from mathematical models. Mind Genomics presents a new approach to understanding decision making under uncertainty, based upon the nature of the mind of the person.

Introduction

The world of investing combines a plethora of disciplines, some economic, some mathematical, most psychological, and a great deal of simple human behavior which constitutes the quotidian behavior of daily life. There is vast literature on what the person ‘might be doing,’ what type of information does the person take into account, predispositions riskiness, financial acuity, and comfort. Entire courses about Finance are devoted to the stock market, to the psychology of investing, and so forth. An excursion in the world of the psychology of investing usually begins with economics as the foundation, and the human player linked into the economics, either obeying the laws of rationality proclaimed by economic theory, or behaving as human beings filled with emotion and biases, proclaimed by the new discipline of behavioral economics. The focus may be on risk taking [1], on the nature of information [2], on the tonality of information provided by the corporation such as social conscientiousness [3], on gender [4, 5] even the susceptibility of the respondent investor to the messaging of others [6]. Yet, with all this information we do not get a sense of the mind of the investor as separate from the act of investing, although there are papers on ‘investor psychology’ [7, 8].

This study steps back from the increasingly sophisticated analysis to look at the simple presentation of investing behavior, the type of presentation that a person in psychology might find interesting, e.g., a brokerage. We look at how people respond to descriptions of investor behavior, to say ‘fits me’ or ‘doesn’t fit me.’ We are not looking for theoretical structures, but simply for a way to understand the way people think of themselves. We are acting as a psychologist, a doctor, a diagnostician of the mind, and not presenting a deep approach of what we believe are the theoretical bases for underlying behavior’.

Method

The research used the approach developed for the emerging psychological science of Mind Genomics, with origins in experimental psychology, experimental design, and marketing. The objective of Mind Genomics is to map out the decision rules for a topic, with that topic being familiar, such as eating a food, choosing a physician, selecting a product in a store, or in this study, investing in the stock market. In all of these topics, one either asks the respondent to describe her or his behavior through qualitative research (focus groups; discussions), observes the behavior, uses surveys, or following Mind Genomics, presenting the respondent with simulated situations, and observing the behavior. Mind Genomics produces a cartography of ideas or perhaps more appropriately, a listing of the relevant idea in a topic, and a metric associated with these ideas. The metric may be the linkage of the idea to oneself (fits me) or the degree to which the idea drives an expected action, such as choose to invest or choose to buy.

The origins of Mind Genomics in terms of psychology and philosophy comes from the method of induction, offered by philosopher Francis Bacon, combined with the Socratic Method for asking questions. In the simplest terms, a Mind Genomics exploration or cartography of a topic comprises the definition of the topic, the asking of four ideas, the creation of four answers for each question, the combination of these answers into vignettes, and finally obtaining respondent ratings of these combinations, followed by a statistical analysis of the ratings of combinations to show the ‘effect’ of each individual idea. The scientific history of Mind Genomics has its origins in the merger of experimental psychology to understand the ‘mind’ statistics using experimental design to create the necessary stimuli, and marketing research which focuses on the daily life of people. The necessary papers constituting the background can be found by looking at the introduction and references provided by [9, 10, 11] psychology, market research and Box [12] experimental design in the field of statistics. As described above, the Mind Genomics method may seem to be one of the many different forms of surveys, and one would not be totally wrong to conclude so. Yet, there is a difference between a survey and a Mind Genomics cartography. With a survey, one asks the respondent a question, and obtains one of several answers. The analysis shows which answers are linked with a question/ The analysis provides mind-sets as well, different patterns of answers to the same question. The analysis does not show causality, however, as might be the conclusion if one could link a response (invest) to a set of messages. In contrast to the intellectual history and applications of the survey method, Mind Genomics can be said to be an experiment. Mind Genomics creates a set of systematically created combinations with known elements in each combination, presents these combinations to the respondent, who integrates the information in each combination, and rates the combination on a scale defined by the researcher. The analysis links the response to the individual element, revealing the decision criteria of the respondent. The respondent need not explicate the criterion for decision; they emerge from the regression analysis.

For the world of investment, Mind Genomics works at the level of the conversation, not the level of technical. That is, the test elements, the stimuli, are mixed into vignettes, combinations of simple phrases describing the nature of the information that a person might obtain from everyday life. These elements are presented as key sources of information. The respondent is then instructed to read each vignette, i.e., combination of ‘information’ and say whether paying attention to that combination of information ‘describes me’ or does not describe me.

Explicating a Mind Genomics Experiment – What information is Perceived to be ‘Relevant’

A great deal of the informal talk about investment deals with the source of the information, and perhaps some surface specifics. This information can be gleaned from participating in the myriad conversations which occur in the morning in offices, at breakfasts among friends, and so forth. Mind Genomics captures this information through a Socratic process, comprising the requirement to generate four questions which tell a story, and then four answers to each question. The task is not particularly challenging but does require people to think in a critical manner. (Table 1) presents the four questions and the four answers to each question. If this study were to be a simple questionnaire, then the researcher would list the answers in some randomized order, present the 16 answers as 16 actual ‘questions’ and instruct the respondent to scale the importance of the answer in terms of how affects the respondent’s decision when thinking about an investment, whether the respondent is an actual investor or instructed to think in the way an investor would think. The important thing to keep in mind is that the survey forces an intellectual consideration of each answer ‘separately,’ and out of context. The task is difficult primarily because it is hard to think of just one idea at a time. Often the respondent attempts to give the researcher answers deemed to be politically correct and appropriate. Mind Genomics works in a different way.

Table 1. The four questions and the four answers to each question.

The underlying premise is that people may not be able to tell an interviewer what they want, how they make a decision, and what they will do. At a conscious level the respondent may not even know the answers to the questions, but the social constraints of interviewing require an answer or permission to not guess. Yet, for the most part, people behave in a consistent manner, even though they might not be aware of just how consistent their behavior might be. The question is how to determine what people think when they cannot or will not reveal this in a directed questionnaire, despite their real-world consistent behavior. In the Mind Genomics experiment the researcher identifies messages which may be relevant. These are the 16 answers to the four questions in (Table 1). The role of the question is simply to set the story, and to elicit the answers. It is the answers which are relevant. Mind Genomics then combines these answers (elements) into small, easy to read vignettes, concepts, according to an experimental design. The design ensures that each of the elements appears an equal number of times in different vignettes and constitutes only part of the vignette. Other elements are present in the same vignette. The respondent’s task is to read the entire vignette as one idea and assign a rating. The respondent rates 24 such vignettes. The combination of different elements makes it impossible for the respondent to focus on one element. Rather, as in real life, the respondent must respond quickly, almost intuitively, to the combination. What emerges is a more valid response to the test stimuli, the vignettes. As will be explicated below, the experimental design allows the researcher to deconstruct the response to the mixture, the vignette, into the separate contributions of the 16 elements.

Creating the Test Combinations (Vignettes) by Experimental Design

People are accustomed to reading paragraphs or collections of disparate information, and making a decision on the combination. It is intellectual challenging when reading a combination to assign the relative importance of each piece of information in the combination, although when pressed to do so, the respondent can probably come up with a criterion. There are a number of methods which try to estimate the likely performance of a combination by the evaluation of single elements (self-explication of components), or the choice between pairs of elements [13]. These approaches are cognitively challenging, making the respondent move beyond sampling and rapidly evaluating the combination, but rather focus on the components of a mixture, an atypical situation.

In the Mind Genomics system, the 16 answers or elements are combined into a set of 24 vignettes, combinations, with the experimental design valid for a single individual. That is, all of the 16 elements are statistically independent of each other, each vignette comprises at most only one element or answer from any question, and there are vignettes in which answers from some questions are missing. Each vignette comprises 2–4 elements. This approach, the within-subjects design, is very powerful because it allows the data to be analyzed at the level of the individual respondent, a property very critical for mind-set segmentation. One other feature deserves mention. When only one set of 24 vignettes were to be tested, the researcher would have to be very careful about selecting the specific vignettes. Rather than forcing a lot of thinking about just what combinations to develop by the experimental design, Mind Genomics makes use of a patented technology, permuted experimental design [14] to systematically vary the specific elements that are combined in the vignettes. The mathematics of the design are maintained, but the combinations different for each person. Thus, it becomes possible to explore a topic area quickly, reduce most of the intellectual bias, and uncover the ‘mind of the respondent’. The set-up of the study has been automated (www.BimiLeap.com). The app allows the researcher to type in the questions, then the answers, and a rating scale. Afterwards, the researcher selects the panel, generally using an on-line panel provider. (Figure 1) shows the set of set up screens that the researcher would use to create the experiment. After the researcher has set up the study, the researcher launches the study. Usually the on-line panel providers generate the necessary, qualified respondents in about 2–3 hours. The data are analyzed automatically to generate summary models for total panel, gender, age, a third question (type of information), and emergent mind-sets (discussed later on.)

Figure 1. The set-up of the study in the BimiLeap app prior to the experiment. The left panel shows the four questions, the middle panel shows the four answers to the first question, and the right panel shows the respondent orientation and scale.

The analytic approach for each subgroup (e.g., total panel, gender, age, third question, mind-set) is the same, comprising the assemblage of all the data for the relevant respondents for that subgroup (24 records for each respondent), and a linear regression model relating the presence/absence of the elements in a vignette (code 0=absent, 1=present) to the rating (1–5), or more generally to a binary transformation of the rating (1–3 ® 0 ; 4–5 ® 100). The binary transformation reflects legacy from consumer research and management needs. Consumer researchers often use category or Likert Scales, anchored at both ends (e.g., our 1–5 scale). Managers often prefer data presented in binary form, no/yes, no go/go, reject/accept, etc.

Building Models to Show How The Elements Drive The Binary Response (Top2Box) and response time (RT)

The experimental design created for each individual respondent produce 1200 individual vignettes or combinations of elements, most of which differ from each other. Thus, the Mind Genomics experiment covers a wide range of alternative combinations of elements. The experimental design created for each indivual, based as it is on a ‘kernel’ experimental design permuted for the respondents, ensures that the set of 16 elements are statistically independent of each other.

The data matrix comprises 1200 rows 24 rows for each respondent. Each respondent has a unique identification number, and a set of 16 columns to store the the independent variables. The independent variables, the 16 answers or elements, are coded 0 or 1, respectively. A 0 represents the fact that the element is absent from the vignette, whereas a 1 represents the fact that the element is present in the vignette. The three dependent variables are the original rating on a 5-point scale, the binary equivalent (Top2Box, 0/100) emerging from a recoding of the ratings, and finally the number of seconds showing the ‘response time’ (or reaction time), defined as the time between the appearance of the vignette and the respondent rating. (Table 2) shows part of the data set that will be used in the regression modeling (Table 2) shows the data structure. The experimental design comprises simply the listing of the different elements. The design must be translated into a series of 0’s (element absent from the vignette), and 1’s (the element present in the vignette.) The statistical modeling, OLS (ordinary least-squares) regression analysis, assigns a weight to each of the 16 predictor variables (A1-D4) so that by knowing what elements are in the vignette one can estimate the likely rating by simply summing up the coefficient [12].

Table 2. Structure of the first eight vignettes, and the ratings.

The actual study was run in the middle of August, 2019, using an on-line panel provide, Luc.id. The actual process of developing the study, running the respondents, and then analyzing the data, took approximately three hours, from start to finish, using the above-mentioned program www.BimiLeap.com. The program itself guides the user, from the start (specifying the topic, posing the questions, requiring four answers to each question) on through instructing the respondents, asking other questions beyond age and gender, and then requiring the researcher to write a short paragraph of WHY the study is being done. The latter requirement, a short paragraph about WHY, comes from the major use of the BimiLeap program to ‘teach critical thinking,’ and not just to be a survey tool. A great deal of consumer research can be made tortuous by forcing the researcher to create a questionnaire, call a panel service, and run the study. The approach of BimiLeap and other modern platforms is to dispense with the back and forth process of dealing directly with an on-line field service. The approach requires the research to specify the specifics of the respondent, assuming they are not overly specific, and then launch the study with a credit card. Everything else is automated. The process returns with a report in PowerPoint® and well as the raw data and relevant tabulations in Excel®. This automated set-up allows the entire process to be completed in 2–3 hours, with the set up of the study, e.g., thinking of the questions and answers, the critical thinking, coming to the fore as the rate-limiting step.

Modeling

For the OLS regression we use ALL the data from all relevant respondents in a defined subgroup. For example, with 50 respondents, and with 24 vignettes for each respondent the relevant data for the total panel is 50×24 or 1200 rows of data, also called ‘observations.’ The independent variables are the 16 elements, coded 1 or 0. The dependent variable is the so-called Top2Box (a term from consumer research). The Top2Box becomes 100 (plus a small random number) when the original rating is 4 or 5. The Top2Box becomes 0 (plus a small random number) when the original rating is 1,2 or 3, respedtively. In a second analysis, looking at response time as the dependent variable, the same structure of analysis holds, except that the dependent variable is simply the response time from the appearance of the vignette on the respondent’s screen until the rating. The response time is recorded to the nearest 10^th of a second.

We inteprret the coefficient as the, probability that the respondent will assign a rating of 4–5 to a vignette when the eleent appears vignette. When we switch to the response time, the we intepret the coefficient as the number of seconds neede to process the element, i.e., to read it and move on. A benefit of the modeling is its ability to deconstruct the compound stimulus, the vignette, into the part-worth contributions of the individual elements. Researchers often believe that this decomposition can be done easily and with only a bit of attention. A strategy occasionally used is to circle the relevant elements if a vignette. This ‘strategu of highlight what seems to be important’ seems at first quite reasonable, but in light of the power of the regression analysis enabled by the experimental design, such manual circling appears to be inefficient, primitive, and unable to deal at all with response time and processing speed. (Table 3) presents the output of the regression analysis. Most regression packages present the regression results in the same fashion. The elements are presented at the left of the table, with abbreviation first, and then the text of the element. The elements are presented in descending order of magnitude, as shown by the column marked ‘Coeff.’

Table 3. Statistical output of the regression model for the Total Panel. The dependent variable is the binary transformation ‘describes me.’  The independent variables are the 16 answers or elements in the vignettes.

The regression model estimates the coefficients, k1-k16, for the simple linear equation:

Top2 or Number of Seconds = k₁(A1) + k₂(A2) … k₁₆(D4)

For the Top2Box model, the coefficient (coeff), shows the expected number of binary points that would be added to a vignette if the element were inserted into the vignette. This can be interpreted as the incremental percent of respondents who would rate the vignette as 4 or 5 when the element is inserted into the vignette.

The element most describing the respondents in their own opinion, at least on average, is:

C1: Follow the vertical because I’m heavily knowledgeable in it and feel it’s home (coefficient of 11.06.)

The element least describing the respondents in their own opinion, at least on average, is:

D3 Everything I hear tells me no one knows exactly … what a chance for ME (coefficient of 2.75)

Next to the coeff (coefficient) is the column label SE. SE is the standard error of the coefficient, or the expected variability of the coefficient if the study were repeated. The coefficient is an estimated parameter for real data. As such, the variation in the data upon repeated studies may be expected to affect the estimated value of the coefficient. The lower the variation in the coefficient, i.e., the lower the value of SE, the more likely it is that we have value of the coefficient which is not truly 0. The likelihood of having a coefficient truly different from 0 is given by the t-stat (t-statistic, a measure of the signal/noise ratio), and the p-value, the probability that the t-statistic is really 0. We should look at all of the coefficients above 0.11 as being truly different from 0. Our data suggests most of the elements are really truly different from 0, i.e., probably are somewhat truly descriptive of the respondents as a group. The only elements which are probably 0 are these four, with coefficients lower than 5.

D4  My experience tells me there’s money to be made here … I just know it4.15

B2   Story: Structural problems in the vertical promise downstream trouble 3.49

C3  The mathematical models I use say ‘this vertical is hot’ 3.24

D3  Everything I hear tells me no one knows exactly … what a chance for Me 2.75

The user interpretation of the coefficients is different from the statistical interpretation. The coefficient shows the linkage between the person and the element. High coefficients mean that there is a strong linkage. In percentage terms, where 0 is no linkage and 100 is perfect linkages, i.e., 0=does not describe me … 100=describes me, the coefficient gives the additive percent towards the complete linkage. The elements are additive. That is, one can put up to four elements together, when they are answers to different questions and estimate the total degree of linkage.

The regression model does not have an additive constant for this study. The rationale for omitting the additive constant, i.e., forcing the regression model through 0, is that in the absence of elements there is no meaning to the additive constant. In contrast, were the rating scale to be ‘likelihood to purchase,’ and a 1–5 scale transformed in the same way, the additive constant would be meaningful. It would be the likely interest in purchasing the product or service in the absence of any information. That likelihood value is both relevant and truly informative.

Results From the Regression Analysis

The regression modeling was done with a variety of subgroups, as the respondents defined themselves. (Table 4) shows us the coefficients from the total panel, from genders, and from the two ages. It is clear from (Table 4) that many numbers are linked to the way the respondent feels about herself or himself. These are the elements which generate coefficients of 10 or higher, twice the standard error, and thus a coefficient whose t-statistic approaches or exceeding 2.0.

Table 4. Performance of the elements by key subgroups, defined by who the respondents say they ARE

One way to look at these results might be to sort the elements by the number of subgroups which find the element to be important (i.e., the coefficient of 10 or higher). Whe we do this sort, we find that there are just two extraordinarily strong elements, elements whose strong performance is not surprising.

B3       Story: Investors seem to be taking a big position in a vertical

A2       The news appears in Bloomberg

Despite the presence of strong performing elements in each group, there is no easy story to be gleaned from the data. The data are statistically strong but suggest that either there is no pattern, or more likely, the pattern has little to do with who the respondents ARE in a geo-demographic sense. When we look at the respondents by their self-described attitudes towards investing, the picture becomes much clearer, as (Table 5) shows. Those who do not like investing feel that they are best described by elements which convey a ‘general feeling.’ Those who invest with advice feel they are best described by messages about the source of the information and described by their own research. Those who are active investors feel that they are described both by the source and by the ‘story.’

Table 5. Performance of the elements by key subgroups, defined by how the respondent describes her or his attitude towards investing

Mind Sets

One of the tenets of the emerging psychological science of Mind Genomics is that for each topic of everyday experience (e.g., shopping, consulting a doctor, etc.) or even thinking (e.g., about issues of ethics and morality) there exist small, specific domains of the topic. Rather than the grand top of investing, for example, the domain might be limited to the ‘nature of the information to which I am exposed.’ That topic is the subject of this Mind Genomics cartography. Mind Genomics posits that in every domain of the topic, small or large, there may be several different patterns of information to which an individual might respond, rather than only one pattern. That is, individuals differ from each other in the nature of the information to which they respond, the messages which ‘inform their decision.’ Beyond simply positing these different groups of individuals, Mind Genomics provides the computational machinery both to identify these different groups, so-called Mind-Sets for a topic and then a way to assign any new person to one of the Mind-Sets, the method being called the PVI, the personal viewpoint identifier. Discovering Mind-Sets is a statistical process, objective in nature for its computations, but subjective in terms of decision-making about the nature of the revealed Mind-Sets. The approach is quite straightforward, following a statistical path comprising four steps:

Step 1 Select the basic data from which the Mind-Sets will be uncovered

The basic data comprises the 16 coefficients for each respondent. Each respondent generates 16 coefficients from the regression model relating the transformed binary response to the presence/absence of the element.

Step 2 Estimate the ‘distance’ between each pair of respondents

The distance may be defined in any number of ways, ranging from the Minkowski R metric (e.g., the mean squared differences along each of the 16 pairs of coefficients, for R = 2, the familiar Euclidean measure of distance), to the Pearson distance, a metric which looks at the similarity of patterns. The Pearson distance between any two objects e.g., people, is value (1-R), where R is the Pearson correlation between the two respondents, based upon the values of the 16 coefficients. The Pearson R has maximum of +1 when two variables are perfectly related to each other in a linear fashion, and thus show the same pattern. The distance is thus 0, because the two patterns are perfectly related. The value (1-R) is 0, when R = 1. The Pearson R has a minimum of -1 when two patterns are perfectly inversely related to each other. The distance is thus 2 (1- – 1 = 2).

Step 3 Place the respondents into either two or three mutually exclusive and exhaustive groups

This is called clustering [15] The objective of clustering is to minimize the distances within a Clustering this requires computation. The composition of the clusters, the emergent mind-sets, will vary somewhat depending upon the way the ‘distance’ is defined. This should not be worrisome since the clustering is simply meant to be a heuristic, to divide the respondents in a way that may be useful for other analyses.

Step 4 Interpret the clusters or mind-sets

The mind-sets, mutually exclusive and exhaustive, should ‘make sense’ (interpretable), and should be as few as possible (parsimonious.)

(Table 6) suggests three different mind-sets, as follows

Table 6. Performance of the elements by three mind-sets, subgroups, defined by similar patterns in the way people describe their attitudes towards investing. Mind-Set is abbreviated MS

Mind-Set 1 – Responds to where the news appears,

The news appears in the Wall Street Journal

The news appears in Bloomberg

Mind-Set 2– Responds to the story

Story: Structural problems in the vertical promise downstream trouble

Story: Investors seem to be taking a big position in a vertical

Mind-Set 3 Responds to recommendations of friends and appears to be intuition-driven

Just got into the vertical because of recommendations from people I trust

Finding the Mind-Sets In the Population

Respondents who differ in their attitudes in terms of the nature of information to which they respond may or may not realize that there are different groups, different mind-sets. (Tables 5, 6) show clear differences among the groups in terms of their responses to different types of information, which covary with the group to which they belong. Yet, if one subscribes to the belief that people want to hear messages which resonate with them, it might be a better strategy to work with mind-sets of investors, rather than relying upon how the investor defines herself or himself. What might happen if one were to know at the start of the conversation the mind-set to which a prospect belongs? One could then be more comfortable choosing the messages because many of these messages linked with the mind-sets show very high coefficients, 15 or higher. (Table 7) shows the distribution of the three mind-sets by total panel, gender, age group, and self-stated preferences for the type of information. There is no clear pattern.

Table 7. Distribution of the respondents by total and the three mind-sets.

An alternative way uses an algorithm known as the PVI, the personal viewpoint identifier. The PVI asks the respond six questions derived from the experiment, and from the pattern of answers the PVI assigns the new person to the most likely mind-set. The PVI algorithm uses the coefficients to identify which combination of elements, posed as questions and answered with a 2-point scale (Not ME: Me) (Figure 2) shows the PVI questionnaire as presented to the respondent. The order of questions varies across the respondents. The PVI also allows the researcher to ask specialty questions as well, in order to gain more knowledge. The PVI takes about a minute to complete

Figure 2. The PVI (personal viewpoint identifier) for the investing experiment.

Response Time and Engagement

Experimental psychology began with the systematic study of reaction time, the time between the presentation of a stimulus (e.g., our vignette), and the time when the respondent assigned a rating, or simply noted that she or he ‘detected’ the stimulus. There is the abiding, probably correct, belief that longer reaction times correspond to ‘more things going on in the mind.’ Shorter reaction times, therefore, suggest fewer things going on in the mind, or the fact that the respond ‘sees’ the message and discards it without consideration [16]. The Mind Genomics experiment itself lasts 3–4 minutes in the 5-minute interview. During that time the respondent is presented with 24 vignettes, and required to read the vignette (more likely glancing through it, grazing for information), and then responds. There is little time to read the entire vignette. The reaction must be almost automatic, namely see, rate, see, rate, etc.

(Table 8) presents the estimated response times for the 16 elements, by key self-defined group (gender, age, respectively). The respondents answer quickly, and are not at all aware of how much time they spend on each element. The OLS regression estimates the likely number of seconds required for each element to be read and processed. Those elements which generate coefficients of 2.0 (two seconds or longer) are shown in in shaded cells, and bold type. These are the elements to which the respondent attends, whether the attention reflects an emotional reaction, or an attempt to comprehend the meaning of the element. (Table 8) suggests that older respondents typically take longer to process then information than do younger respondents. Those over 60 show higher coefficients for response time than those respondents under 60. There is also the suggestion that the genders differ in what engages them. Female’s attention is engaged by other people (The news is given by your colleagues at your office), whereas male’s attention is engaged by technology (The mathematical models I use say ‘this vertical is hot’).

Table 8. Response time in seconds for each element. Data for total panel, gender, and age, respectively.

When we move to self-defined groups in terms of how one invests (e.g., invests with advice, etc.), (Table 9) suggests clearly different patterns of engagement. Those who say that they do not like investment pay a great deal of attention to the story. Those who say that they invest with advice also pay attention to the elements dealing with the story, as well as pay attention to ‘clues’ about performance, typically given by others. Those who say that they are active investors pay attention to one element, ‘Story: Investors seem to be taking a big position in a vertical.’ When we move to the mind-sets defined by the pattern of coefficients, we see that there are differences as well, albeit not the strong differences that we saw for those who self-define themselves in different ways in terms of attitudes toward investing.

Table 9. Response time in seconds for each element. Data shown for three different self-descriptions of the respondent’s attitude toward investing.

Table 10. Response time in seconds for each element. Data shown for three Mind-Sets.

Those who fall into Mind-Set1, paying attention to news from different sources, pay attention to messages which promote some type of discovery, either from the news, from listening to friends and colleagues, or intuition.

Those who fall into Mind-Set 2, paying attention to the story, pay attention to the source (Wall Street Journal, Bloomberg) and to news about sell-offs, and to a sense of finding out something that no one else knows.

Those who fall into Mind-Set 3, paying attention to friends and their own intuition pay attention to the Wall Street Journal, to Bloomberg, to their own unique discovery, and to colleagues.

What is surprising here is that the response times, a measure of engagement, does not covary strongly with who the respondents are. That is, a respondent who feels strongly about something which defines her or him may not be engaged with that message if engagement is measured by response time.

Discussion

As we saw in the introduction, a literature search on investing behavior uncovers a vast range of issues, ranging from behaviors used, strategies adopted, and the inner mind of the investor. To a great degree studies about the psychology of investing have emerged, not unexpectedly, from the marriage of finance and psychology. The emphasis of these studies is on the financial implications of psychological states of mind and its co-variation with strategies. There is relatively little published dealing with the discourse between the investment specialist and the customer. The sheer issue of gaining versus losing deflects the focus from the purely psychological ‘attitude’ to the attitude of investing as an economic behavior What is missing is the person-to-person behavior, the social aspect of the investor, not the economic aspect. Knowledge of the mind of the investor provides us with a new avenue of understanding finance. The field of behavioral economics focuses on the nature of people’s decision making under uncertainty. Investing is in that class. We are often treated to interesting experiments on how people make their investment choices, on the ratiableonal approach to investment. We are less exposed to issues about the nature of information. The approach presented here provides a simple, easy-to-execute foundational study on the ‘mind of the investor,’ not so much dealing with rationality or irrationality, but rather dealing with the way the investigator navigates through opportunities, through information, through communication with others, respectively. Through Mind Genomics we use the economic aspects of the investing simply as a set of test stimuli, ‘assayed’ by the human mind. Our focus is on the mind anticipating economics-relevant behavior [17, 18] and not on the marriage of the mind and the theoretical underpinnings of economics [8].

Acknowledgement

Attila Gere thanks the support of Premium Postdoctoral Research Program of the Hungarian Academy of Sciences.

References

1. Gervais S and Odean T (2001) Learning to be overconfident. Review of Finance Studies 14: 1–27.
2. Mezick EM (2002) Investing wisely: Individual investors as information seekers. Journal of Business & Finance Librarianship 7: 3–23.
3. Escrig-Olmedo E, Munoz-Torres MJ, Fernandez-Izquierdo MA (2010) Socially responsible investing: sustainability indices, ESG rating and information provider agencies. International Journal of Sustainable economy 2: 442–461.
4. Barber BM and Odean T (2001) Boys will be boys: Gender, overconfidence, and common stock investment. Quarterly Journal of Economics 116: 261–292.
5. Bengtsson C, Persson M, Willenhag P (2005) Gender and overconfidence. Economics Letters 86: 9–203.
6. Barber BM and Odean T (2000) Too many cooks spoil the profits: Investment club performance. Financial Analysis Journal 56: 17–25.
7. Hirshleifer D and Subahmanyam A (1998) Investor psychology and security market under- and overreactions. Journal of Finance 53: 1839–1886.
8. Tversky A and Kahneman D (1974) Judgment under uncertainty: Heuristics and biases. Science. 185: 1124–1131.
9. Moskowitz HR and Gofman A (2007) Selling blue elephants: How to make great products that people want before they even know they want them. Pearson Education.
10. Green PE and Rao VR (1971) Conjoint measurement for quantifying judgmental data. Journal of Marketing Research 8: 355–363.
11. Green PE and Srinivasan V (1990) Conjoint analysis in marketing: new developments with implications for research and practice. The Journal of Marketing 54: 3–19.
12. Box GE, Hunter WG, Hunter JS (1978) Statistics for experimenters, New York, John Wiley.
13. Louviere JJ, Flynn TN, Marley AAJ (2015) Best-worst scaling: Theory, methods and applications. Cambridge University Press.
14. Gofman A and Moskowitz H (2010) Isomorphic permuted experimental designs and their application in conjoint analysis. Journal of Sensory Studies 25: 127–145.
15. Dubes R and Jain AK (1980) Clustering methodologies in exploratory data analysis. Advances in Computers 19: 113–238.
16. Boring EG (1929) A History of Experimental Psychology. The Century Company, NY.
17. Cross F (1973) The behavior of stock prices on Fridays and Mondays. Financial Analysis Journal 29: 67–69.
18. Menkhoff L, Schmidt U, Brozynski T (2006) The impact of experience on risk taking, overconfidence, and herding of fund managers: Complementary survey evidence. European Economic Review 50: 1753–1766.

DOI: 10.31038/ASMHS.2019353

Abstract

The objective of this study was to understand how ordinary people react to information about conferences. The study is part of the Mind Genomics exploration of the world of the everyday. Respondents evaluated systematically created vignettes about conferences, with the elements of the vignettes presenting information about the topic of the conference, the way the material is presented to excite emotions, the way the facts are presented, and the after-conference activities, respectively. The study introduced the assessment of interactions between ideas in a vignette (scenario analysis). The results suggest three clearly different mind-sets; those who focus on the topic, those who focus on the nature of the presenter, and those who focus on the after-conference activities. These three mind-sets distribute similarly in the population. The study presents a PVI, personal viewpoint identifier, allowing a conference planner to understand the mind-set to which members of the prospective audience may belong, which knowledge may produce a more impactful conference.

Introduction

One need only look at the proliferation of non-governmental organization intent on solving key issues in the world to get a sense of an increasing social awareness. Beyond the world of the organization is the world of the meeting, where experts and others in the field come together, under one or another aegis or directorate, to discuss the problems, to formulate solutions, or simply to meet. The meetings are in the thousands, often by invitation, and limited, presumably to those attendees whose interest is established in the topic. Every organization attempts to validate its meaning, its raison d’etre, either by publications which communicate important information to the world, by publications of an academic nature which dissect the problem, or more typically in today’s world, by a ‘meeting.’ The meeting, formally titled ‘conference’ , assembles those who are involved in the topic. The conference may turn into a standard, periodic meeting, or become a one-off attempt to ‘solve a problem’ or at least to discuss how various experts would approach the problem. One needs only look at the announcements of such meetings to get a sense of how popular it is for people to get together for short, concentrated periods of time, be available to the public as the public face of those concerned, come up with recommendations, and then scatter back to their regular jobs. Our focus in this paper is to understand how the average person reacts to these types of conferences. To be sure, the topic of ‘meeting’ is not one of high interest to people, unless they are somehow involved. Despite the special nature of conferences, the notion of conferences is by now well known, especially due to the high-profile nature of conferences dealing with important issues. The emerging science of Mind Genomics, the study of the everyday, provides a perfect tool to understand how ordinary people respond to descriptions of these ‘meetings.’ We are not taking the pulse of people towards meetings in general, the points of satisfaction and dissatisfaction, but rather trying to understand the mind of the typical person confronted with special topic, issue-related conferences.

The world of public conferences

The topics of conferences vary dramatically. Most conferences are of minor importance, dealing with specific issues and relevant to a limited number of people, the organizers and the attendees. On the other hand, there are major conferences, often sponsored by world organizations such as the United Nations or by NGO’s (non-government organizations.) The participation of NGO’s continues to interest researchers [1–4], perhaps because the NGO’s are involved with high-profile topics. Conferences are also venues for professionals to meet, and especially for graduate students to introduce themselves to their colleagues, and present papers about their work [5,6] For the seasoned professional, conferences are a venue for promoting one’s work, and for developing a support system [7,8]. For the undergraduate student, an often-overlooked ground of nascent professionals, conferences can provide a launching pad to create a life-long professional [9] The focus of published research literature dealing with conferences is the topic itself, and secondarily a venue in which people interact [10]. In spirit the literature is sociology, the interact of people, and not psychology, the individual’s needs, wants, feelings, and behavior. There is some literature on the desires of conference attendees in terms of they want [11,12], and even papers dealing with people’s behaviors in conferences, such as tweeting [13]. The result of such investigations reveals the nature of people’s participation, and even prescriptions about creating good conferences [14]. They do not tell us about the inner feelings of people towards a conference as a part of their daily life.

This paper moves from the conferences as a topic of sociology, looking from the outside, to a topic of psychology, looking at the conference from the inside, focusing on the reactions of a respondent presented with vignettes, small descriptions, about conferences. Each description comprises different features of the conference; topic, speaker, nature of information, follow up activities. The objective of the Mind Genomics effort in general, and this study in particular, is to create the science of the everyday, using one’s knowledge of quotidian events like conferences, to understand the way people think about and make decisions about the ordinary events of their lives.

The Mind Genomics approach

When confronted with everyday situations, especially those which do not require much thinking, and where there is no real ‘risk,’ the typical person acts automatically, what Nobel Laureate Daniel Kahneman calls involving the System 1 mode of thought. System 1 is the emotion-driven, automatic system upon which virtually people rely most of the time. System 1rapidly processes the information, incorporates emotions, and drives a response, often on what seems to be ‘auto-pilot.’ Such a system is necessary to navigate a life in which many choices are required to be made each few minutes, ranging from where to move when walking, to how to eat a meal, and so forth [15]. The automatic responses and behaviors need to be limited to the very ordinary. For example, people who attend the meetings often form opinions in what seems to be an automatic fashion, talking freely about the different aspects of the meeting without much rehearsal. That is, people respond to the meetings, can dissect the different aspects of the meetings with great ease after the fact, and in the case of boring meetings, during the meeting as people talk with each other while ignoring the presenter. The set of tools to investigate the aspects of the everyday are housed in Mind Genomics. Mind Genomics is the emerging science of the everyday, created to look at the nature of the different patterns of reactions that people exhibit to descriptions of situations [16–18]. The intellectual history of Mind Genomics emerges from mathematical psychology [19], and the adaptation to market research with consumers [20, 21]. When applied to social issues such as meetings sponsored by NGO’s, Mind Genomics reveals what is important about the meeting versus what is unimportant, or how the presenter affects the credibility of the information being presented.

Raw materials

The input to Mind Genomics comprises a set of questions, and alternative answers to each question. Table 1 presents the four questions and the set of 16 answers. Typically, these questions ‘tell a story’, with the different answers providing the necessary material to ‘flesh out’ the story. The questions never appear in the actual experiment with respondents. Rather, the questions are used to elicit the answers, which will appear in combination.

Table 1 presents the information in a way which enables the respondent to ‘graze the vignette,’ and extract the relevant ideas. Each answer is prefaced by an introductory phrase, such as ‘conference topic,’ ‘presenter,’ ‘expert’ and ‘follow-up’ respectively. Although this format does not lead to grammatically elegant vignettes, the format makes it easy to present the respondent with the information necessary to make a decision. In studies using experimentally designed combinations of ideas, with individuals exposed to many combinations, it is becoming increasingly vital to shorten the interview. The time for elegantly written but dense paragraphs has passed, as researchers are forced to do increasingly shorter interviews and experiments. The design of the individual elements in this study (Table 1), and the design of the vignettes are done with the recognition that the entire Mind Genomics experiment should last no more than five minutes.

Table 1. The questions and answers about conferences.

The answers are combined according to an underlying experimental design, a prescription of what answers or ‘elements’ should be combined [22] The experimental design prescribes a set of 24 combinations for the array of four questions and four answers for each question. The underlying experimental design ensures that all 16 answers are statistically independent, permitting the analysis by OLS (ordinary least-squares) regression. The underlying experimental design ensures that some of the test combinations, called ‘vignettes’, are incomplete, lacking either one or two answers. This deliberate creation of incomplete vignettes is done so that the coefficients from the OLS regression have ‘absolute value,’ allowing them to be compared from study to study, even when the studies comprise other types of messages.

Each respondent evaluated a totally unique set of combinations of vignettes, created according to the same underlying experimental design, but ‘permuted’. This strategy enables the researcher to assess many different combinations of answers or elements [23] it is important to emphasize that this strategy of testing many different vignettes, with 1–2 evaluations of each vignette, means that we look for patterns by looking at the entire space of alternatives, rather than looking for patterns by canceling out the variability. Most research looks for patterns by suppressing the noise through replication. Mind Genomics does the exact opposite, discovering the pattern by looking at a lot of the space, even if the individual measures are ‘noisy.’

The rating 5-pooint rating scale combining understanding and action

Traditionally, Mind Genomics has worked with bipolar Likert scales, anchored at the top and at the bottom. The scales usually have focused on one dimension, whether that be ‘do not understand versus understand’ (what is), or ‘not motivated to do something versus motivated to do something’ (intended action.). When the two topics of ‘what is’ versus’ intended are investigated in the same study, they often have been separate questions, answered quickly in succession. The Mind Genomics experiment presented here represents the next generation, in which a single rating question is created to encompass two dimensions.

For this study, the five rating points appear below:

Here are conferences dealing with major problems. How do you feel about this specific conference as described? Choose one of the following five answers

1=tuned out immediately. ..waste of time

2=don’t understand facts & not motivated to solve problem

3=understand facts but not motivated to solve problem

4=don’t understand facts but motivated to solve problem

5=understand facts & motivated to solve problem

Analyzing the results at a surface level

The easiest way to understand the data, and to compare groups looks at averages. We create the following key dependent variables, and then compare them by group:

1. Response time
2. Rating 1 converted to binary (tuned out)
3. Rating 5 converted to binary (understand facts & motivated to solve problems),
4. Ratings 3&5 converted to binary, a so-called ‘netted variable’ that we call UNDERSTAND.
5. Ratings 4&5 converted to binary, another ‘netted variable’ that we call MOTIVATED TO SOLVE PROBLEMS

Table 2 shows the average value for each of these five variables, for total panel, and key subgroups (gender, age, self-defined focus on the topic of these conferences, and finally two groups of mind-sets emerging from dividing the 50 respondents into complementary groups, based upon the pattern how motivated they are to solve problems

Table 2. Average ratings for response time (seconds), and binary variables, based upon the analysis of subgroups.

The important lesson from Table 2 is that there are differences which manifest themselves in the response time, and in the pattern of ratings. We see the expected differences between the respondents who say that they are ‘tuned out’ versus those say they are passionate. Furthermore, some age differences emerge, few gender differences emerge, and so forth.

Table 2 lacks the cognitive dimension of the results. We see behaviors, but the averages have only meaning in a numerical way, telling us an external measure, a measure that we attempt to use as we search for an underlying pattern. The pattern lies within the mind of the researcher, not in the data. There is no cognitive richness in the data presented by Table 2, but only patterns, the meaning of which must be imposed on the data, and with any luck, will be perceived as appropriate for the data, not as imposed on the data.

Finally, in (Table 2) there is the story, but the story is general, not specific, not rich, and certain does not tell us of the inner workings of the mind of the respondent.

Linking messages to judgments

We undertook this Mind Genomic study to understand how people react to the different aspects of these NGO-sponsored conferences. We focused on five different responses that people might have, including absolutely no interest (Rating 1, R1), do not understand and not motivated to solve the problem (Rating 2, R2), understand but not motivated to solve the problem (Rating 3), do not understand but motivated to solve the problem (Rating 4, R4), and finally understand and motivated to solve the problem (Rating3, R3) or motivated to solve the problem (Rating5, R5). We also presented two net key variables; understand (Rating3 + Rating4, R3+R4), and motivated to solve the problem (Rating4 + Rating5, R4+R5).

The underlying experimental design enables us to relate the presence/absence of the 16 elements to either one of the responses (1–5), or any subset of the responses (e.g., those of men versus those of women). The approach used is known generically as regression analysis, occasionally referred to as ‘curve fitting.’ The objective is to deconstruct the dependent variable to the contribution of the 16 contributing elements the answers provided the four questions. The cases or observations for the regression analysis comprise the full set of 1200 vignettes, wherein one knows both the composition of the vignette, e.g., which of the answers were present,’ and the reaction, e.g., which rating or net rating was selected, and thus converted to 100. All rating variables will be presented after transformation to the binary values 0 (not chosen) or 100 (chosen) when the respondent evaluated the vignette. (Table 3) shows us the coefficients for six dependent variables; R1, R5, Net Not Understand, Net Not Motivated, Net Understand, Net Motivated. The model begins with the additive constant. The additive constant is the estimated value of the dependent variable when there are not elements or answers in the vignette, a hypothetical situation since all vignettes comprised at least two elements, and at most four elements. Nonetheless, the additive constant is a useful number, behaving as a baseline. When we look at the coefficients, we should keep in mind that the underlying statistics of the regression enable us to estimate the likelihood that the coefficient that we observe is not just a random occurrence from an underlying distribution of coefficients with a real average or mean of 0. That critical value is 8 or higher, or -8 and lower. Knowledge of that range (beyond +/- 8) helps us focus on those answers or elements which drive a strong positive or negative response. These strong performers are shown as number in bold font, and in shaded cells. Visual inspection suggests possible patterns in this otherwise daunting ‘wall of numbers.’

Table 3. Links between elements (answers, messages) in the vignettes and six dependent variables. The strong linkages are shown in bold font, and shaded cells.

We look now in a rapid fashion at the six response variables:

R1 – Tuned out: This response has a low additive constant, 24, meaning that in the absence of elements in the vignette, approximately one quarter of the responses will be ‘tuned out;’ There are no key drivers of ‘tuning out’, at least with the total panel.

R5 – Understand and motivated to make a change: This variable has the lowest additive constant, 20, meaning that in the absence of elements in the vignette, approximately one fifth of the responses will be this positive. It is the job of the elements, the answers, to drive understanding and motivation. Only one element is sufficiently powerful to drive this response, A3, conference topic: loss of respect and empathy of people towards each other. The strong response to A3 emerges because of the strong ‘pull’ of this idea.

Net Not Understand: This variable is constructed from the two response variables which feature ‘Do Not Understand’. They are rating choices R2 and R4. When either is selected, the newly constructed variable, Net Not Understand, is given the value of 100. When neither is selected, Net Not Understand is given the value of 0. Ironically, the only group which promotes a possible misunderstanding is the presenter being from a critical NGO (non-governmental organization.) This suggest that the role of NGO is not perceived as very instructive, at least by the average American respondent.

NET NOT MOTIVATED (to solve problems): This variable takes on the value 100 when the respondent chooses R2 or R3, both involving no motivation to solve problems. Otherwise, this variable takes on the value 0. The key destroyer of motivation to these respondents is the presenter being the well-known author on the topic. It is as if having the well-known author is a symbolic fulfillment of what has to be done. Metaphorically, the author is the ‘priest’ who atones for the congregation. An analogy may be made to modern corporations which send their employees to conferences on innovation, have walls of awards and certificates in their lobbies, but are prisoners to outdated processes, and believe that despite innovation, ‘process is king.’ As long as the employees listen to experts, the corporation may be said to fulfill its role to embrace innovation.

NET UNDERSTAND: This variable is constructed from R3 and R5. Both responses talk about understanding the facts. The additive constant is high. The key element driving this response is the conference dealing with teaching students to think critically. The respondents believe that they will understand the issues involved.

NET MOTIVATED: This variable is constructed from R4 and R5. Both responses talk about being motivated to change. The elements appear to connect with the human experience:

conference topic: loss of respect and empathy of people towards each other

conference topic: problems teaching students to think critically

conference topic: government actions and quality of life

presenter: talks about personal experiences and suffering

presenter: well-known social activist

presenter: critical NGO (non-governmental organization)

Does the topic of the conference affect how people judge the different vignettes?

The features or messages to which one responds are not independent of each other. That is, depending upon one part of the message, another part of the message may either make sense or not make sense. A good example is the price. For example, we can lay out prices for an object, and ask people to rate the degree to which the price is fair. Yet, none of the pricing data makes sense unless we know the object or service for which the price is designed. A $2.00 price for a loaf of bread is meaningful. A $2.00 price for an automobile makes no sense whatsoever.

The Mind Genomics system enables us to assess pairwise interactions answers from different questions (or elements from different silos.) This ability to address the issue of interactions emerges as a happy byproduct of the nature of the underlying experimental design, a structure which specifies the test combinations. The design remains the same for all respondents, but the actual combinations change from one respondent to another. The ensures a statistically robust set of combinations, with all the answers from one question appearing with all the answers in the other questions. In other words, the final set of combinations is sufficiently robust to allow us to pull out pairwise combinations.

The strategy to uncover pairwise interactions is straightforward both in computation and in meaning, respectively. We divide the set of 1200 vignettes into strata, based upon the answer in one question. In the analysis presented here we divide the 1200 vignettes into five strata, depending upon the specific question. We will focus on Question A, the topic of the conference. Our focus now is how the different elements or answers perform when the conference topic is held constant, focusing on problems in world environment, problems in teaching students to think critically, and so forth.

We first sort the data, creating five strata, depending upon the particular topic. We then run the OLS regression once again, this time running the data separately in each stratum. The OLS regression is run on 12 predictors, the four answers from Question B, the four answers from Question C, and the four answers from Question D. Thus, we have five parallel analyses, one for each topic, and one analysis where no topic is specified. Thus, the focus is not on the topics as separate, but the topics as guiding the performance of the remaining elements. This approach has been coined ‘scenario analysis’ [24]. The reason for the term ‘scenario’ is that the analysis operates with a specific type of meeting, the ‘scenario’ in which everything is judged.

What makes a respondent feel that he or she would tune-out, i.e., reject the conference

We begin our analysis of interactions by identifying those elements or answers which lead to the respondent ‘tuning out.’

1. No topic (all vignettes lack the presence of A1-A4): These vignettes generate a fair amount of tune-out responses. The additive constant is 26, meaning that in the absolute of a topic, and just information about the presenter and the follow-up,, about one out of four responses will be ‘tuned out’ (R1). What really bores people, however, is a presenter with a video narration. The coefficient is +12, a really boring strategy.
2. Problems in the world’s environment: This topic is also slightly boring, with an additive constant of 20. That 20 means that in the absence of answers or specific elements, just knowing that the conference is about problems in the world’s environment will generate about 20% responses of ‘tuned out.’ However, there are no elements which drive ‘tuned out.’ The elements themselves are interesting.
3. Problems in teaching students to think critically: This topic of a conference is more interesting. The additive constant is 15, meaning only 15% of the responses are expected to be ‘tuned-out’ when the specific elements or messages are missing. Once again, we see that no elements or answers drive boredom and tune-out. The specifics are interesting.
4. The loss of respect and empathy of people towards each other: This is also a fundamentally more interesting topic, with the additive constant of 14. There are three elements which drive the respondent feel that he or she would tune out, despite the fundamentally interesting nature of the topic:

   follow-up: create workshops in schools

   presenter: critical NGO (non-governmental organization)

   presenter: well-known social activist

5. Government actions and the quality of life: This is perhaps the most likely topic to drive the response of ‘tuned out.’ The additive constant is 25. Beyond that, however, we find no elements which are turn-offs.
6. We conclude from this analysis that there are interactions between the topic of the conference and the elements which can be found boring. Some interactions are dramatic. A narrated video presentation might be a turn off by itself when there is no topic specified, and a turn-off when the topic is loss of respect and empathy of people towards each other (coefficients of +12 and +6, respectively), but will be not a turn off when the topic is government actions and the quality of life (Table 4).

Table 4. Scenario analysis. How the nature of the conference (top row) drives the response R1 (tuned out).

Getting the message across – what drives the response of ‘I understand the facts’?

Our second analysis looks at the drivers of ‘I understand the facts), which comprises responses R3 and R5, together. When a respondent selected R3 or R5, this new ‘net variable’ of ‘understand’ was assigned the value 100. When a respondent selected R1, R2 or R4, respectively, this new net variable ‘understand’ was assigned the value 0. The analysis then proceeded as did the previous analysis, considering five strata, based upon the topic of the conference. (Table 5) shows the detailed results.

Table 5. Scenario analysis. How the nature of the conference (top row) drives the ‘net response’ of understand the facts (combined Rating3 and Rating5).

1. The additive constants suggest that even without elements or answers, at least half of the responses are going to encompass some understanding. The most likely understanding will come from conferences dealing with problems in the world’s environment. The least likely understand will come from conferences dealing with loss of respect and empathy of people towards each other.
2. No topic specified – strongest contribution to understanding comes from creating free groups to give meaning and motivation. Follow up here is important.
3. Problems in world environment – basic understanding is very high (additive constant = 67), but no elements or answers increase understanding.
4. Problems teaching students to think critically’

   expert: well-known author on topic

   presenter: video presentation narrated

5. Loss of respect and empathy of people towards each other – expert: well-known university professor with to-do list
6. Government actions and quality of life –

   Follow up: create free groups to give meaning and motivation

   expert: high government official in topic area

What makes the respondent feel that she or is motivated to make changes?

Our third analysis looks at the net rating of ‘yes, motivated to make changes.’ This net variable comprises the selection of rating 4 (do not understand the facts, motivated to make changes) or the selection of rating 5 (understand the facts, motivated to make changes).

The actual analysis is identical. The only difference is the choice of the dependent variable. (Table 6) presents the detailed results regarding what motivates the reader to believe that she or he will take action.

Table 6: Scenario analysis. How the nature of the conference (top row) drives the ‘net response’ of motivated to make changes (combined Rating4 and Rating 5).

When we look at the additive constants, showing the expected likelihood of people saying ‘I am motivated to make changes,’ we find that the highest motivation emerges with conferences on loss of respect and empathy of people towards each other (additive constant = 59.) The lowest likelihood emerges with conferences regarding problems in the world’s environment (additive constant = 31), and problems teaching students to thinking critically (additive constant = 36).

The elements or answers which drive the motivation tell their own stories. The operating elements which ‘work’ must have a topic of the conference

Problems in the world’s environment – best to have a video presentation and follow-up groups.

Problems teaching students to think critically – best to have a social activist presenting, or a narrated video presentation, and then follow-up groups

Loss of respect and empathy of people towards each other – best to have a person with experience talking about the experience

Government actions and quality of life – best to have an NGO presenter

Gender differences

Often, genders do not differ dramatically from each other, except in topics that are gender-relevant, such as cosmetics. The data from the total panel (Table 3) can be deconstructed into the responses by gender (Table 7). When we look at males versus females for the net response of ‘understand’ (ratings 3 and 5 combined), we see that women are more likely to say that they ‘understand the fact’s (additive constant 55 for women, 43 for men), and that they are also more ‘motivated’ (additive constant 41 for women, 28 for men). Thus, the first observation is that women will be more likely to say that they are affected by the conference.

Table 7. Comparison of male versus female in their ratings of ‘understand the facts’ and ‘motivated to make a change.’ The numbers in the body of the table are the coefficients from the ‘net’ models (understand = R3 & R5; motivated = R4 & R5).

The real gender differences emerge when we look at the answers or elements. In terms of helping the person to understand the facts, men strongly feel that the conference will help them to understand two topics, problems teaching students to think critically and loss of respect and empathy of people towards each other. It may be that these are the only topics that men feel they will learn something new.

When it comes to the topic of motivation, (Table 7) shows dramatic differences by gender. Men are convinced by presenters from NGO’s and by social activists. Men start at a lower level (additive constant = 28), feeling that it will be harder to motivate them, and in turn feel that NGO’s and social activists will be effective. Women, in contrast, are far more likely to say that they will be motivated (additive constant = 41). The truly dramatic topics, those which women think will motivate them those dealing with loss of respective and empathy (coefficient = 21), teaching students to think critically (coefficient = 17), and government actions and the quality of life (coefficient = 13).

In search of different mind-sets

It has become increasingly clear during the past decades that people differ dramatically in what they find interesting. This variation across people in liking is not a new discovery. The old adage holds increasingly today: Of taste one does not dispute. Each person has his or her own pattern of preferences, these preferences ranging from the sensory experience one enjoys (e.g., different flavors), but moving on to experiences themselves (ways of being treated; activities to do on vacations.)

The notion of differences across people is obvious. One important question is to develop a way to measure the pattern of preferences, which has been done by Mind Genomics, and just demonstrated for data from the total panel versus from males versus females. The next question is to determine whether there are fundamental groups of people, so that the patterns of preference are similar within a group, but the patterns of the groups differ dramatically from each other?

Discovering different groups, mind-sets, can be formulated in terms of a statistical problem answerable by the technique of clustering [25]. Each respondent in this study generated a set of 16 coefficients, one coefficient for each of the 16 phrases. The coefficients we choose are those emerging out of Ratings 4 and 5, motivated to change. Clustering divides the set of 50 respondents into mutually exclusive groups, with the property that the patterns within a group are similar to each other, whereas the patterns of the averages of the groups are very different from each other. These groups, statistically developed, are called Mind-Sets in the parlance of Mind Genomics.

The clustering procedures works with a measure of ‘distance’ between pairs of respondents. The distance is defined as the quantity (1-R), where R is the Pearson correlation coefficient. Thus, the distance measure looks at how well the two patterns correlate. When the patterns of coefficients from two respondents correlate perfectly, they are really reacting in the same way to the answers or elements. The Pearson correlation is 1.0, and the distance should be minimal, which it is. The distance is (1-R), i.e., (1–1), or 0. In contrast, when the two respondents react in opposite ways, they are maximally different from each other. The Pearson correlation is -1, and the distance is maximal (1 – – 1 = 2.)

The clustering procedure is agnostic, not concerned with the meaning of the clusters, focusing only on satisfying the mathematical criteria of maximal distance between the averages of the two clusters on the 16 answers, and minimal distance between pairs of respondents within a cluster. The clustering must be augmented by some researcher input, specifically:

Parsimony – fewer clusters or mind-sets are better than more, both from an aesthetic point of view in research, as well as from an actionability point of view when the data are put to use.

Interpretability – the cluster must ‘make sense,’ i.e., tell a story

The two clusters based upon ‘Motivated’ (R4 and R5; converted to binary) were not interpretable. Too many different ‘stories’ emerged. The three clusters which emerged based upon ‘Motivated’ tell a more coherent story, and so we settle on the three clusters

The three-cluster solution is remarkably simple to interpret, suggesting three different ways to motivate the audience. There are those who are motivated by the topic, those who are motivated by the presenter, and those who are motivated by the after-conference opportunities to share ideas (Table 8).

Table 8. Comparison of three emergent Mind-Sets based on clustering the coefficients from ‘Motivated’ (R4 & R5). The numbers in the body of the table are the coefficients from the ‘net’ models (motivated R4 & R5; understand the facts R3 & R5.

Finding mind-sets in the general population for better conference design and effective messaging

Mind Genomics as we have just demonstrated begins to provide a corpus of information about the aspects of daily life. The issue beyond science and discovery is application. How can one apply these results in a way which makes the discoveries more than simply part of the knowledge base of sociology and human behavior? Can we learn more if we can expand the discovery of these three mind-sets beyond the limited confines of this study of 50 respondents? In other words, can we apply this information to create better conferences, or at least better understand the audience’s predispositions towards what they want in a conference?

It will be difficult, if not impossible, to assign a person to the proper mind-set simply by knowing who the person IS. (Table 9) shows the distribution of the three mind-sets by gender, by age, and by how the person describes herself or himself when it comes to issues about the world. The distribution is fairly flat, so any opportunity to find a specific group of people with a designated mind-set if probably going to end up in failure.

Table 9. Distribution of the three mind-sets across gender, age, and self-report attitude towards conferences.

An alternative way to identify people comes from reducing the large-scale experiment to a set of questions, the PVI, the personal viewpoint identifier. The questions emerge from the actual experiment, the study described here. The PVI as currently designed, comprises a fixed number of six questions, with the questions themselves taken from the actual study, and thus varying from study to study. The respondent reads each question and chooses one of two answers. The total set of 64 patterns is mapped to the assignment to a mind-set. Thus, each of the possible patterns corresponds to the likely membership in one of the three mind-sets. The approach is empirical, based upon the actual study, with the PVI created shortly after the experiment.

(Figure 1) shows the PVI as the respondent see it. It takes approximately 30–45 seconds to complete the PVI. The appropriate mind-set may either be returned in a report to the respondent as a motivating device to make the PVI fun, and in turn, the data may be store in a digital record. That record, obtained from thousands of people, may be used for marketing in the case of commercial events, and follow-up for other uses, e.g., for health when the topic is not conferences, but health-issues and concerns.

A parenthetical note: Without the knowledge of mind-sets, and the disturbing reality that these mind-sets distribute without any noticeable skew towards a specific group in the population, marketers, event planners and others continue to believe that who a person is co-varies with how a person thinks. That is, in the absence of such knowledge, one must use demographics and other variables. Rather than doing the simple Mind-Genomics experiment followed by a PVI for the topic, the strategy has evolved to using Big Data of low information density, coupled with very high-powered analytics. The metaphor is needing to rely upon powerful, expensive equipment in a mine where the gold is rare, rather than using simple equipment or even one’s own hands in a mine where the gold is abundant.

Figure 1. The PVI for conferences as the respondent would see it.

Response time

Beyond the aspect of what persuades at a cognitive level, aspects captured in the rating, lies a whole world of ‘processing,’ of psychological aspects to which the cognitive mind may not be privy. Experimental psychologists almost a century and a half ago recognized that beneath the surface responses to test stimuli lie many factors, such as attitudes, norms, and so forth. These factors govern the response but cannot be articulated.

Recent developments in Mind Genomics have focused on capturing the response time to vignettes, defined as the time between the appearance of the test stimulus and the response to the test stimulus. With the advent of today’s computer technology this information is readily available. The response times (often referred to in the literature as ‘reaction times’) become meaningful when they can be paired with specific stimuli, as they are in the Mind Genomics paradigm. That is, when the researcher can estimate the number of seconds that can be linked with each answer, it becomes possible to learn more about what engages the respondent. We don’t know what is happening, but we do know that some answers are processed more slowly (longer response times), and some elements are processed more quickly (shorter response times.)

The analysis of response times requires a slightly modified equation. The equation incorporates all the answers or elements as predictors, but there is no additive constant. The rationale is that in the absence of answers or elements in the vignette the response time is 0. We write the equation as follows:

Response time = k₁(A1) + k₂(A2) … k₁₆(D4)

(Table 10) shows the response times for the three mind-sets generated from the mind-sets based upon ‘motivation’. The table shows the longest response times as shaded cells with bold font. What surprises in a delightful way is the observation that the different mind-sets pay attention in accordance with their mind-sets. The correspondence is not perfect, but there is a clear connection between what persuades/motivates and what people attend to. This is an area worth exploring in more detail.

Table 10. Estimated response times attributes to the different answers/elements, from each of the three mind-sets.

Discussion and conclusion

The literature of conferences is a growing one. The focus, however, is the nature of the specific conferences, from the point of view of the topic, and the influence of the topic. Few papers, if any, focus on the psychology of the listener, other than perhaps papers dealing with the role of conferences in the development of a person’s professional career. This paper introduces a new world of understanding conferences, not so much from the topic and the importance of the topic to the world, but rather conferences as a part of a person’s quotidian, daily life. As noted in the presentation of Mind Genomics, the world of the everyday presents us with a way to understand people. With the tools of Mind Genomics, we begin a new psychology of people, the psychology of the ordinary, of which conferences as a topic constitute one facet.

The results from the data should not surprise, although the reality is that were one to be asked about ‘what makes a good conference,’ one might not emerge with answers as clear as those provided by Mind Genomics. Nor, in fact, would there be the specifics provided by the cognitively rich stimuli used in Mind Genomics studies, specific, meaningful statements. The initial foray into ‘what interests a person in a conference’ shows the simplicity by which one can begin to create a detailed understanding of a person’s mind with regard to a topic. The call now should be for systematics, namely structured investigations. Should the topic of such investigations be ‘conferences,’ and the current study comprises, the next steps would be the way the conferences are organized, the nature of the material presented, the tonalities of the presentation, the venues, and so forth. Following this structured approach, it is likely that an entire “foundational knowledge infrastructure” (FKI) about conferences might be constructed within the period of a year, providing insight to the specific topic of conferences, but potentially greater insight into the nature of social interactions of a formal nature. The potential of a set of FKI’s, updated each year, and done cross-sectionally within a topic, across topics, within a country, and across countries, beckons, almost a ‘Wiki of the Mind.’

Acknowledgement

Attila Gere thanks the support of Premium Postdoctoral Research Program of the Hungarian Academy of Sciences.

References

1. Clark AM, Friedman EJ, Hochstetler K (1998) The sovereign limits of global civil society: a comparison of NGO participation in UN world conferences on the environment, human rights, and women. World Politics 51: 1–35.
2. Gordenker L (1995) NGO participation in the international policy process. Third World Quarterly 16: 543–556.
3. Ritchie C (1995) Coordinate? Cooperate? Harmonise? NGO policy and operational coalitions. Third World Quarterly 16: 513–524.
4. Van Rooy A (1997) The frontiers of influence: NGO lobbying at the 1974 World Food Conference, the1992 Earth Summit and beyond. World Development 25: 93–114.
5. Gardner SK, Barnes BJ (2007) Graduate student involvement: Socialization for the professional role. Journal of College Student Development 48: 369–387.
6. Jacobs N, McFarlane A (2005) Conferences as learning communities: some early lessons in using ‘back-channel’technologies at an academic conference–distributed intelligence or divided attention? Journal of Computer Assisted Learning 21: 317–329.
7. Kowalsky M (2012) Networking at conferences: Developing your professional support system. Knowledge Quest 41: 60.
8. Shoaff-Ballanger SM, Davis DJ (1992) Professional conferences for art educators: A pilgrimage to excellence. Art Education 45: 37–44.
9. Carsrud AL, Palladino JJ, Tanke ED, Aubrecht L, Huber RJ (1984) Undergraduate psychology research conferences: Goals, policies, and procedures. Teaching of Psychology 11: 141–145.
10. Gunawardena CN (1995) Social presence theory and implications for interaction and collaborative learning in computer conferences. International Journal of Educational Telecommunications 1: 147–166.
11. Hines SS (2014) What do library workers want from professional conferences? In: Revolutionizing the development of library and information professionals: Planning for the future (1–16). IGI Global.
12. López-Pastor VM, Monjas R, Manrique JC (2011) Fifteen years of action research as professional development: Seeking more collaborative, useful and democratic systems for teachers. Educational Action Research 19: 153–170.
13. Reinhardt W, Ebner M, Beham G, Costa C (2009) How people are using Twitter during conferences. Creativity and Innovation Competencies on the Web. Proceedings of the 5^th EduMedia Pg No: 145–156.
14. Werner SE, Kenefick C (2005) A Primer for effective organization of professional conferences. Medical Reference Services Quarterly 24: 39–54.
15. Kahneman D, Egan P (2011) Thinking, fast and slow. New York: Farrar, Straus and Giroux.
16. Moskowitz HR, Gofman A, Itty B, Katz R, Manchaiah M, et al. (2001) Rapid, inexpensive, actionable concept generation and optimization: the use and promise of self-authoring conjoint analysis for the food service industry. Food Service Technology 1: 149–167.
17. Moskowitz H, Gofman A, I novation Inc (2003) System and method for content optimization. U.S. Patent 6: 662,215.
18. Moskowitz HR, Gofman A, Beckley J, Ashman H (2006) Founding a new science: Mind genomics. Journal of sensory studies 21: 266–307.
19. Luce RD, Tukey JW (1964) Simultaneous conjoint measurement: A new type of fundamental measurement. Journal of mathematical psychology 1: 1–27.
20. Green PE, Rao VR (1971) Conjoint measurement for quantifying judgmental data. Journal of marketing research 8: 355–363.
21. Green PE, Srinivasan V (1990) Conjoint analysis in marketing: new developments with implications for research and practice. The journal of marketing 54: 3–19.
22. Box GE, Hunter WG, Hunter JS (1978) Statistics for experimenters, New York, John Wiley.
23. Gofman A, Moskowitz H (2010) Isomorphic permuted experimental designs and their application in conjoint analysis. Journal of Sensory Studies 25: 127–145.
24. 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.
25. Dubes R, Jain AK (1980) Clustering methodologies in exploratory data analysis. Advances in Computers 19: 113–238.