DOI: 10.31038/ASMHS.2019313

Abstract

This study assessed the variation of concerns about expressing emotions in social relations, emerging from the increasing use of smartphones. Fifty respondents from the continental U.S.A participated through an application of Mind-Genomics Science. Two mind-set segments emerged. People in the first mind-set were concerned about the increasingly use of smartphones for social interaction, and the need for instant feedback. People in this segment stressed the need to put on a mask when presenting only a happy successful face. This segment was also preoccupied with our changing language skills, and with the increasing lack of privacy because everything one with smartphones is trackable. People in the second mind-set segment expressed concerns regarding the negative social effects of using smartphones to express emotions: less interaction at meals, isolation from personal relationships, fewer expressions of feelings, losing patience more quickly, and emerging health issues. A PVI (personal viewpoint identifier) is presented to allow discovery of these two mind-sets among new individuals, enabling a deeper understanding of the mind-sets by future researchers.

Introduction

An increasing array of studies have begun to document both the positive and the negative contributions to health and well-being benefits associated when people expression their feelings through this modern device, a multi-purpose mobile computing device, inexpensive, and widely used. The smartphone has become ubiquitous, and now a necessity in people’s lives offering, among others, new opportunities to interact in social relations. This study assesses concerns with the relation between use of smartphones and the changing patterns and abilities in a person’s expression of emotions.

Smartphone use produces both positives and negatives. One is example is self-esteem. Communication through smartphones did not predict self-esteem. The smartphone also has an effect on interpersonal relations. The smartphone is a way to express caring feelings. People wanted to use smartphones to bond with each other, thus reducing depression [1].

Interesting patterns emerge when the research focuses on the use of smartphones in the evening, rather than during the day. Those who work in the evening felt that using the smartphone to express emotions diminished a feeling of well-being, suggesting that smartphones used in the worktime (here evening) to express emotion, was simply not effective. There should be a time when smartphones are used to express emotions, and that time is not when one works. Furthermore, using smartphones in the evening for work increased psychological detachment [2].

When a person uses the smartphone compulsively, an increasingly behavior, certain psychological correlates emerge. In a study of links between psychological traits and the behaviors of smartphone users, Lee, Chang, Lin, and Cheng [3] reported that compulsive use of smartphone for expressing emotions was positively related to: locus of control, anxiety regarding social interaction, and finally a need for touching.

The growing evidence on negative effects of using smartphones raises the question about attitudes of people towards this use, which is the focus of this study. Compulsive use of smartphones provides a wonderful psychological ‘petri dish,’ emerging out of an increasingly used technology. The issues involve social relations, as well as well as characteristics of various types: Personal, psychological, emotional, and social-environmental, respectively [4–6].

1. Social isolation, family discord, divorce, academic failure, job loss and debt [7].
2. People with depression, loneliness, social anxiety, impulsivity and distraction may easily become compulsive users in expressing emotions through smartphones [8]. Fifty-four percent of compulsive users reported a prior history of depression; 34% had an anxiety disorder; and 52% had a history of alcohol and drug abuse.
3. The place offering internet access, the degree of time to use the internet, peer relationships, parenting types were also linked to compulsive use [5].
4. Most of compulsive users did not have proper school, work or interpersonal relationship, respectively [5]. They felt anxious and lonely without their smartphones [9].

Defining The Problem – What Concerns about Emotions and Empathy in this New World

The preponderance of social research about trends comes either from observing trends of behavior, and/or asking people about their feelings, either through surveys or discussions, such as focus groups with people discussing a topic, or in-depth interviews with one or two people to probe deeply. From these investigations one learns what is happening, and why it is happening. The two questions provide a good idea of the nature of the trend and may even suggest what will be the trajectory of the trend.

Mind Genomics approaches the problem of what and why through a different approach, one grounded in experimental psychology, and based in the notion of experiments to understand causality. Mind Genomics approaches the issue by presenting the respondent with a variety of vignettes, descriptions of a situation, a feeling, an observation. These vignettes are created by a Socratic technique, involving asking four questions about the topic, questions which tell a story, providing four answers to each question, mixing the answers together by a systematic approach, presenting these mixtures or vignettes to respondents, obtaining ratings, and then determining which element or answer in the vignette is a ‘driver’ of the rating.

The foregoing approach sounds very circuitous to answer a problem of ‘what is the trend,’ or ‘how do you feel?’ Yet, as the data will show below, this presentation of vignettes, getting responses, extracting the contributions of the elements, and uncovering the pattern provides a richness of insight that could otherwise not be obtained.

Approaching the problem by Mind Genomics

Mind Genomics begins with the raw material, ideas, and messages. These are called silos (questions) and elements (answers to the questions.) The actual study will involve mixing these elements into combinations, but before the mixing can occur, one must assemble the raw material, relevant statements about the issue, and sort them into questions and answers.

The actual work of creating silos and elements, questions and answers, is not presented here. Only the final set of test material is presented, as shown in Table 1. The task of developing the questions and answers is separate, and appropriate for any effort which wants to ‘sharpen’ a person’s mind. By expanding a topic such as loss of empathy into questions and answers, silos and elements, we force the researcher to ‘rewire’ thinking, proceeding in a more structure, and more methodical, more inclusive manner than the superficial thinking which accompanies the activities of ordinary daily life.

The four questions and the four answers to each question represent just a small sample of the many different ‘ideas’ involved in the possible loss of empathy through the overuse of today’s smartphone and similar devices. It is important to emphasize the that Mind Genomics does not try to answer the ‘big question’ by one study, but rather builds up a picture of the topic through many small studies whose combined information reveals an underlying pattern. The study reported here deals with only a limited number of meaningful aspects of the smartphone experience as it drives and affects interpersonal interactions and empathy. One could easily repeat this study, with new questions, new answers, and in doing so obtain another answer, or more correctly another ‘slice’ of the answer. The metaphor with the MRI should become clear. The MRI takes different pictures, from different angles. Mind Genomics does so as well, but pictures of big topics, not pictures of tissues [10].

Creating Combinations of Elements or Answers by Experimental Design

Today’s course of science education teachers that the scientific method works by isolating variables and studying the variables in detail. In such a way, one obtains a sense of ‘how nature works.’ The notion of studying interactions among variables is, of course, part of the basic topic. Interactions are, for the main part, studied with the belief that one studies a single variable, in the presence of other variables which affect that single, studied variable. That is, the focus is still on the single variable. The interaction of two variables is presented as how one or several variables ‘affect’ the response to the variable being studies.

Mind Genomics works in a different way, perhaps more in the spirit of everyday life. The focus of Mind Genomics is how bigger ideas emerge from the composition of smaller ideas, with these smaller ideas evaluated in combination with each other. Mind Genomics might be the study of art, where the focus is on the study of an emergent experience that experience coming from the combination of different independent variables. For Mind Genomics, it is the compound idea which is important, that compound idea coming from mixing together the different elements or answers.

Mind Genomics is taken from mathematical psychology, as adapted by the late Professor Paul Green and his colleagues at the Wharton School of the University of Pennsylvania [11].

Table 2 shows an example of six vignettes, i.e., combinations of elements or answers. Each combination is a defined mix of elements or answers, the exact composition coming from a prescription of combination to be followed. The table shows the element identifying codes for each vignette (top) and the binary expansion of the design, as a preparatory step for data analysis by OLS (ordinary least-squares) regression.

Table 2. Experimental design underlying six vignettes, i.e., combinations of elements.

It is important to note that the combinations, the vignettes, are not complete. That is, some vignettes comprise four elements, one element or answer from each silo or questions. Examples include Vig1 and Vig6. The remaining four show only three elements in a vignette. There are also others which show only two elements in the vignette.

The rationale for incomplete vignettes is that the elements must be statistically independent of each other. The only way to do that is to ensure that they are uncorrelated. In turn, uncorrelated means that knowing the components three of the components of the vignette should not affect the fourth at all. If, for example, we set up the rule that the vignette must have exactly one element from each silo, i.e., one answer from each of the four questions, then knowing three of the vignettes automatically tells us e which specific set of elements will constitute the source of the fourth element in the vignette. In such a situation, the elements are not truly independent of each other in a statistical sense. The regression analysis will fail because of so-called multi-collinearity.

The experimental design itself prescribes 24 vignettes or combinations of elements. Although the combinations might seem to be created in a haphazard fashion, the reality is quite the opposite. There is one basic design which is ‘efficient.’ By the word ‘efficient’ we mean that the respondent is required to evaluate a minimal number of vignettes (here 24), that all 16 elements appear equally often, that a vignette has at most one element from a silo (i.e., one answer from a question), and that the combinations can change from person to person, but the basic design remains the same. The latter, so-called permuted designs, enables Mind Genomics to test many different combinations of elements, with each respondent evaluating a unique set of 24 combinations [12]. Once again, the analogy here is the MRI, which takes many pictures, many ‘slices’ of the tissue, and combines these pictures to get a more complete picture.

Figure 1 shows an example of a vignette as the computer presents the vignette to the respondent. The figure is presented in the way the vignette would appear on the screen of a smartphone.

Figure 1. Example of a vignette as it appears on the screen of a smartphone.

Figure 1 shows a very simple format, easy to use. The introduction to the topic appears on the top of the figure. This introduction never changes from screen to screen. The only change is the content of the vignette, information that the respondent is instructed to consider as one, and only one idea. The respondent reads, or more realistically skims, the set of vignettes, and selects an answer. The screen immediately changes to the next vignette, reducing the onerous nature of the interview.

The respondent also provides answers to age (year born), gender, and a third question dealing with attitudes or behaviors towards the topic. For this study, the third question was phrased as follows:

Choose who you want to be in one year

1 = smartphone for fun and calling

2 = smartphone for calls only

3 = smartphone primarily for fun

4 = no addiction to a smartphone

5 = Not applicable

The information collected about the respondent, age, gender, and the third question (Choose who you want to be) allows the researcher to divide the respondents by who they say they ARE, what they say they DO, or what they say they BELIEVE, respectively.

The actual experience and the response measures

The study was set up on an APP (BimiLeap), which required the construction of the study in a simple format, beginning with study name, then the selection of four questions which ‘tell a story’ (see Table 1), creation of four answers for each question, and a rating scale. The information, once entered in the APP was sent to the e-panel recruiters, specializing in these studies, and affiliated with the APP (Luc.id, Inc.) There were no respondent qualifications, other than an approximate equal distribution of genders.

The respondents were selected by Luc.id, and then invited to participate. Working with Luc.id ensured that the entire study with 25 respondents could be done within the space of less than one hour, with a full report three minutes after the close of the study.

The respondents participated, the study was closed, and the data were analyzed.

The computer program measured two things:

1. Rating the vignette – The direct cognitive response. The rating reveals the conscious degree of concern with what was being read on the screen [11].
2. Response time – The cognitive load. The response time covaries with the effort was being expended to ‘process’ the information, with response time [13].

The computer program first measured the response time, in seconds, between the time that the vignette appeared on the screen, and the time that the respondent rated the vignette. Response times of 10 seconds or longer were converted to 10 seconds, based upon previous experience with response times, showing that only a small proportion of response times were longer than 9 seconds, and of these, most were 15 seconds or longer. We surmise that the respondent was otherwise engaged for a moment while participating, and thus we truncated the range of response times to 0–9 seconds.

The ratings on the 9-point scale were converted to a binary scaling by bisecting the scale into two regions. Ratings of 1–6 were converted to 0, and ratings of 7–9 were converted to 100, respectively. A small random number (<10^–5) was added to each converted number. The rationale for the binary conversion is that for most studies of this kind, it is easier to understand binary data (no/yes) than to understand scalar data. One does not know what the scale points mean. This analysis produces only a slight loss of information at the top and the bottom of the rating scale. Figure 2 shows a scatterplot of average ratings from each of the 50 respondents. Each filled circle corresponds to the average from one respondent, the average based on the ratings of the 24 vignettes (abscissa), or the average based on the binary transformed value for the 24 vignettes (ordinate.)

Figure 2. Scatterplot of average ratings from each of the 50 respondents, based on the 9-point rating (abscissa) or the binary transformed rating (ordinate). Each filled circle corresponds to one respondent.

Morphology – Patterns of Responses

The first analysis concerns the distribution of responses for the ratings of concern, based on the 1–9 scale, and the response times. The results appear in Figure 3. The density plot combines the data from all 50 respondents, each respondent evaluating 24 vignettes, bringing the total number of data points to 50 × 24 or 1,200.

Figure 3. Distribution of ratings of concern, and distribution of response times. The distribution emerges from responses to the 1200 vignettes.

Figure 3 suggests a range of levels of concern across the individual vignettes, as well as a range of response times. As noted above, response times greater than 9 seconds were automatically transformed to 9 seconds because most of the responses take no more than 7–8 seconds. Longer responses MAY indicate deeper thought, but it is more likely that these longer times signal that the respondent was in some way interrupted.

An analysis of average ratings across the 50 respondents shows dramatic person-to-person differences. There are respondents who, on the average are only modestly concerned with the loss of empathy, whether the concern is measured on the original 9-point scale, or on the binary transformed scale. There are also respondents who rate the vignettes very quickly, faster than three seconds on average, and in contrast, respondents who rate the vignettes very slowly, on average taking six seconds or longer to rate a vignette. Figure 4 shows the distribution of these averages.

Figure 4. Distribution of average patterns of responses. Each filled circle corresponds to the average response of one of the 50 respondents, each respondent evaluating 24 vignettes.

The final analysis of the morphology of responses concerns the questions whether those who are more concerned, on average, respond more quickly. The answer to this question is NO, at least for the data on empathy. Figure 5A shows a scatterplot, with each filled circle corresponding to one of the 50 respondents. The plot suggests a random relation between the degree of concern manifested by the average rating (after transformation to binary) shown on the abscissa, and the average response time in seconds shown on the ordinate.

It is important to note that up to now we have looked at the data in terms of what economists called a ‘cross-sectional’ analysis. That is, we have not looked deeply at the composition of the vignettes. Rather, we have searched for emergent patterns from ‘different but complete’ representatives of a domain. What we mean by ‘different but complete’ is that each of the respondents is a separate, measurable entity. We do not know what that entity comprises. We are simply interested in discovering some type of explainable regularity, a pattern, emerging when we plot different measures taken on the same set of entities. We are searching for an unplanned, possibly happenstance regularity of nature, without necessarily doing the experiment to create the possibility of discovering that regularity.

Figure 5A. Relation between average response time (ordinate) and average degree of concern (binary transform, abscissa). The straight line is the estimated best fit. Each filled circle corresponds to one of the 50 respondents.

Deep Analysis – Relating the elements (answers) to the binary ratings

The basic ‘project’ of Mind Genomics is to discover the ‘algebra of the mind.’ Mind Genomics does so by experiments. Rather than relying on cross-sectional analysis of already-completed test stimuli, with the hope that a pattern emerges, and the further hope that the pattern can be explained, Mind Genomics creates the conditions for finding a meaningful pattern. Mind Genomics does so by systematically creating combinations of ideas (the answers), presenting these combinations of known composition to respondents, obtaining ratings, and deconstructing the response to the part-worth contribution of the components.

The contribution of experimental design cannot be sufficiently lauded. Experimental design allows us to create MANY possible combinations, test each one, and combine the data into one analysis. The strategy, as explained below, forces the emergence of a meaningful pattern just by the very nature of how the elements are combined. Mind Genomics does not look for patterns as much as finds the patterns, and perhaps even cavalierly expressed, ‘trips over the abundant patterns.’ It is the task of Mind Genomics to record these patterns which so easily emerge, and then to label the patterns, and move on to understanding more about these discoveries.

The data from the experimentally designed vignettes allows for analysis by standard statistical methods, specifically OLS (ordinary least-squares) regression and then cluster analyses. The former, OLS regression, relates the presence/absence of the 16 elements to the rating, or more correctly to the binary transformed rating (0/100). The latter, cluster analysis, allows the discovery of groups of respondents showing similar patterns of coefficients from the OLS regression. These groups, called clusters or segments, represent like-minded individuals, who show similar patterns of responses to the test elements. They become Mind-Sets in the language of Mind Genomics.

We begin with the OLS regression. The inputs are the independent variables and the dependent variable. There are 16 independent variables, one for each of the 16 elements shown in Table 1.
Table 2, top, shows the composition of six of the vignettes, in terms of the specific elements. The regression analysis cannot deal with this type of data. We transform the data to binary, creating first a set of 16 new variables (A1-D4). The variables take on the value 0 when the element is absent from the vignette, and they take on the value 1 when the element is present in the vignette. These are called ‘dummy variables,’ because they have only two values, 0 or 1. They convey no other information other than absent or present, respectively.

The dependent variable for the OLS regression is the binary transformed rating, namely 0 or 100. We add a small random number during the transformation to ensure that the 0 or 100 become a more variable set of numbers. This strategy of adding a very small random number ensures that we can use OLS regression for respondents who limit their ratings to the lower part of the scale (1–6, all transformed to 0), or who limit their ratings to the higher part of the scale (7–9, all transformed to 100.) The very slight variation in the ratings suffices to protect against a ‘crash’ of the regression program due to the problem of ‘no variation in the dependent variable.’

When we run the OLS regression we obtain output as shown in Table 3. The OLS regression uses all 1200 observations or cases as input. The number 1200 comes from the 50 respondents, each of whom evaluated 24 vignettes, totally 1200. Although the experimental design allows us to run OLS on the data of each respondent, we choose to combine all the relevant data together, and run one OLS model, the so-called Grand Model.

Table 3. Performance of the elements, based on the total panel, and the binary transformed rating. Each number shows the contribution to the likelihood of saying concerned (rating 7–9.) The elements are sorted in descending order of coefficient.

The Grand Model is expressed by the simple linear equation: Binary Rating = k₀ + k₁(A1) … k₁₆ (D4)

Table 3 shows the following parameters

1. The additive constant, k₀, which is the estimate value of the binary rating in the absence of elements. It can be interpreted as a baseline, namely the likelihood or probability that a response will be ‘YES’ (rating 7–9), in the absence of elements. Of course, all the vignettes comprised 2–4 elements, by design, so the additive constant is a purely computed parameter.
2. The coefficients, k₁-k₁₆ value in the model. A coefficient tells us the incremental percent (positive coefficient) or the decremental percent (negative coefficient) of responses being ‘YES’ when the element is incorporated into the vignette. We look for reasonably high positive elements, those around 7.51 or higher, which, from many studies, appears to correspond to meaningful behavior of other sorts, such as buying an item when the rating scale is likelihood to buy.
3. The t-stat or t-statistic, showing the ratio of the coefficient to the standard error of the coefficient. We want a high ratio, to indicate to us that the coefficient differs from 0. The t-stat may be likened to a measure of signal/noise. The t-stats for the total panel are relatively low, suggesting that the coefficients, even the high ones, are based on results with a great deal of noise or intrinsic variability. For example, element A4, ‘We are isolated from personal expressions of feeling because of smartphones,’ with a coefficient of 6.51 may really result from dramatically different points of view, some strongly positive, and others strongly negative. We would like to see t-statistics which are very high, suggesting that they are based on a lot of agreement, not just the result of a ‘tug of war’ between dramatically different points of view.
4. The p-value is the probability that the coefficient comes from a sampling distribution with a true value of 0, rather than what we observe. We always look for low p-values. That is, we always look for low probabilities. When we have a high probability, it means that the coefficient may look different from 0 (e.g., be 3 or 4, or -1 or -5), but the reality could be that the true value of the coefficient is closer to 0.
5. The data we see in Table 3 do not simply provide us with numbers. They allow us to get a sense of the underlying structure of the mind as the mind comes to grips with these statements about the smartphone and empathy.
6. We begin with the additive constant, which tells us the expected percent of times that we will observe a rating of 7–9 when we talk about ‘concern’ but don’t talk about anything specifically other than the general introduction to the problem. Our coefficient is 54.02, meaning that in the absence of elements, a purely theoretical situation but a good baseline, about half the responses will be ‘I am concerned,’ i.e., a rating of 7–9.
7. Each element either adds or subtracts a percent of responses about concern. For example, when we incorporate element A4 (We are isolated from personal expressions of feeling because of smartphones), our coefficient is 6.51. This means that an additional 6.51% of the responses will turn from indifferent/unconcerned (rating of 1–6) to concern (rating 7–9.) A vignette with this one element is expected to generate a percent of ‘concerned’ responses equal to the sum of the additive constant and this single element, or 54.02 + 6.51 = 60.53.
8. Not every element drives concern. Some elements drive no concern, and some even reduce concern. Here are the elements which actually have little impact.

   Our language skills are changing

   Lose patience with others if they are not ALWAYS ON

   We feel alone because others seem so happy and successful

   We present only our happy successful face

9. We can compose new combinations, and estimate the reactions to these combinations, by summing the additive constant and the individual coefficients of the elements being incorporated. We must be careful to limit the number of elements to a maximum of four, and preferably combine elements from the different silos or questions in Table 1, not from the same silo.

Building a Model for Response-Time

When a respondent evaluates a test vignette, the respondent must read the vignette, whether slowly or quickly, following which the respondent presses one key to assign the rating. The time between the appearance of the vignette and the response can be deconstructed by OLS regression into the contributions of the component elements. The independent variables are the presence/absence of the 16 elements, and the dependent variable is the response-time, the time between appearance of the vignette and the rating of that vignette.

We follow the same procedure as we did for the ratings, namely put all the data together into one database, and build a single model. The model is written similarly to the equation above, relating the binary transformed response to the presence/absence of elements. The only difference is that there is no additive constant. The ingoing hypothesis is that in the absence of elements the response time is defined to be 0. The following equation expresses the model:

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

Table 4 suggests a small range of responses for the individual elements, with the fastest response given to A1, because of the lowest coefficient, and the slowest response given to D4, because of the highest coefficient. A1 is a simple fact. D4 is a more frightening proposition, forcing people to stop a bit, if only to think of the implications of being tracked.

Table 4. Response times to the elements based on the total panel.  Each number (coeff) shows the estimated number of seconds required to read and process the information in the vignette.

Do we respond faster or slower to elements which concern us?

Now that we have deconstructed the vignettes into the contribution of the elements towards making the respondent concerned, as well as the time need to process the elements (response time), we can begin to understand the dynamics of concern. The first question is whether there is a clear relation between response time for the element and concern about the element? We look at the relation based upon responses to the elements, rather than response patterns by different individuals, as we had done in Figure 4.

Figure 5B shows the scatterplot. Each of the filled circles corresponds to one of the 16 coefficients. It is clear from Figure 5B that there is virtually no relation between response time and concern, when we look at the pattern generated by the total panel for the 16 individual elements.

Figure 5B. Relation between the response time (ordinate) and concern (abscissa) for the total panel. Each filled circle corresponds to one of the 16 elements.

Patterns Emerging from Subgroups

The initial analysis of the results suggested both that the respondents could differentiate the elements when we consider their coefficients for both transformed binary ratings (concern), and for response time. Some elements drive concern, some do not. Yet, there is a sense that combining the respondents into one group and creating a model for that group may mask differences among the elements in terms of which drive concern, and which are responded to slowly versus quickly.

As an example of the group-to-group differences that one can find, consider two elements, the first being the strongest performing for the total panel (A4: We are isolated from personal expressions of feelings because of smartphones), and the second being virtually irrelevant for the total panel (C2: We talk less and text a lot.) The results of a group-to-group analysis appear in Table 5.

Table 5. Example of how the same element may be judged dramatically differently by respondents in different subgroups.

Our groups are the following:

1. Total
2. Gender – Male vs Female
3. Age – 18–25; 26–39, 40+
4. Two mind-sets based upon patterns of binary coefficients (2A, 2B)
5. Two mind-sets based upon patterns of response time (2C, 2D)

The coefficients for an element different by group, sometimes only by a little, sometimes by a lot. Just because an element performs well for the total panel does not mean that it will always perform well when we look at subgroups. This is especially the case for Mind-Set segments based upon similar patterns of coefficients, with the coefficients coming either from the binary transform of concern (2A vs 2B) or coming from the response time (2C vs 2D.)

Table 6. Key differences by gender

Table 7. Key differences by age group

We now turn to listing the key differences by complementary subgroups. We look only at those elements which generate a coefficient of +8 or more for concern based upon the binary transform, or those elements which generate a response time less than 1.0 seconds.

Segmenting the Respondents on The Basis of the Pattern of Responses (Concern Versus Response-Time)

One of the hallmark features of Mind Genomics is the focus on Mind-Sets. A Mind-Set is defined as a way of thinking about a topic. Operationally the Mind-Set is defined as a set of mutually-consistent and compatible ideas which are held by an individual. Although the statement incorporates people, the Mind-Set is really a set of ideas, not the person. It is the person, the physical individual, who responds and manifests the Mind-Set. In this way Mind Genomics looks at the ideas first, and then who holds these ideas.

The Mind-Sets are developed by the statistical method of clustering [14] the underlying idea is that the thinking pattern of each person can be represented numerically by the pattern of the coefficients, whether the coefficients relate to the binary-transformed ratings of concern or relate to response time, respectively.

When we cluster the respondents based upon the binary-transformed ratings of concern, we look for a small number of mutually complementary groups of individuals who show different and interpretable patterns. That is, the strongest performing elements for each Mind-Set should ‘tell a story.’ This is the first criterion, ‘interpretability.’ The second criterion, parsimony, requires that we create as few Mind-Sets as possible. It is better to emerge with fewer, somewhat less precise, Mind-Sets, than more Mind-Sets, albeit one which are each more precise.

Table 8 shows us the two Mind-Sets emerging from segmenting or clustering the respondents using coefficients from the binary-transformed rating.

Table 8. The two Mind-Sets emerging from segmenting or clustering the respondents using coefficients from the binary-transformed rating.

Table 9 shows the two Mind-Sets emerging from clustering respondents using the coefficients from response time.

Table 9. The two Mind-Sets emerging from clustering respondents using the coefficients from response time.

Do we Get Faster as we have More Experience with the Vignettes?

As respondents go through the experiment, looking at 24 vignettes and rating them, we can measure the average response time. For ease of analysis, we have broken up the data into the vignettes appearing in the first third, the second third, and the final third of the experiment, i.e., in sets of eight vignettes. Table 10 shows clearly that there is a large reduction in response time between the first eight vignettes, and the remaining vignettes.

The response time does not, however, quicken in a monotonic way, as the respondent goes through the experiment. There are three elements which show erratic behavior, with the response time actually increasing as we go from the middle of the experiment (vignettes 9–16) to the end of the experiment (vignettes 17–24). The common factor is the word ‘other.’

We feel alone because others seem so happy and successfulTexting causes misinterpretation of many feelings by others Far less talking with each other at meals

Finding Mind-Sets 2A and 2B (Concern) In the Population

The essence of Mind Genomics is the discovery of how people think about a topic, and, of course, the emergence of different ways of thinking about the same topic. Now that we have demonstrated at least two different mind-sets, the next issue is to explore how these mind-sets distribute in the population, and then discover co-variations of these mind-sets, whether in terms of who the people ARE, how the people THINK, and/or what the people DO. Can we discover new knowledge regarding these mind-sets, and if so how, when our basic science need be developed with only 25–50 people?

As noted before, discovering mind-sets is straightforward, and can be done easily and quickly with a small sample of people. That discovery is akin to discovering the primary colors. We do not need to sample thousands of objects to discover the primary colors. On the other hand, to relate membership in a mind-set to other aspects of the person (ARE, THINK, DO) requires that we assign people to one of the mind-sets, and then look for relations between these people whose mind-sets have been established and other aspects of the people.

We create a PVI, a personal viewpoint identifier, using the data in Table 9, but with the table expanded to show the coefficients of all 16 elements emerging from Mind-set 2A and Mind-set 2B, respectively. The PVI requires the respondent to rate six different ‘questions’ emerging from the data, with each question corresponding to one of the 16 elements. The questions are chosen from those elements which best differentiate between two mind-sets, or in the case of three or four mind-sets, the elements which best differentiate among the three or four mind-sets, respectively. In turn, the questions are answered on a binary scale. The underlying algorithm then assigns the respondent to one of the two (or when appropriate three/four) mind-sets. Figure 6 shows a worked example which is available at http://162.243.165.37:3838/TT08/.

Table 10. Change in the response time as the experiment proceeds.

Figure 6. The PVI (Personal Viewpoint Identifier) and the feedback information about the respondent. The upper part shows the Welcome screen of the PVI with the five binary questions, while the bottom part presents the feedback to the two mind-sets.

Discussion

Looking at the two mind-set segments we learn that people in the first mind-set segment were concerned with the need to use smartphones for social interaction and for instant feedback from others. People in this segment stressed their concerns about one’s need to put on a carnival mask when presenting only a happy successful face [15]. This segment was also preoccupied with increasing trends of changing language skills, and the increasing lack of privacy because everything on the web is trackable.

People in the second mind-set segment expressed concerns regarding the effects of using smartphones to express emotions: less talking at meals, isolation from personal relationships, fewer expressions of feelings, losing patience more quickly, and creating or aggravating health issues. Even with the stated benefits of expressing more with less energy could not counteract the concern regarding the negative health effects was raised.

The most important outcome of this study is the support for previously raised concerns and a recommendation. This study supports previous studies which warned against compulsive use of smartphones and the difficulty to treat it [4,5]. We call work organizations to take responsibility and limit the use of smartphones in the evening for work purposes.

Acknowledgment

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

References

1. Park N, Lee H (2012) Social implications of smartphone use: Korean college students’ smartphone use and psychological well-being. Cyberpsychology, Behavior, and Social Networking 15: 491–497.
2. Ohly S, Latour A (2014) Work-related smartphone use and well-being in the evening. Journal of Personnel Psychology 13: 174–183.
3. Lee YK, Chang CT, Lin Y, Cheng ZH (2014) The dark side of smartphone usage: Psychological traits, compulsive behavior and technostress. Computers in human behavior 31: 373–383.
4. Choi NY, Han EG (2006) Predictors of children’s and adolescents’ game addiction: Impulsivity, communication with parents and expectation about the internet games. Journal of Korean Home Management Association 24: 209–219.
5. Kim H (2013) Exercise rehabilitation for smartphone addiction. Journal of exercise rehabilitation 9: 500.
6. Kim K, Ryu E, Chon MY, Yeun EJ, Choi SY, et al. (2006) Internet addiction in Korean adolescents and its relation to depression and suicidal ideation: a questionnaire survey. International journal of nursing studies 43: 185–192.
7. Young K, Pistner M, O’MARA JAMES, Buchanan J (1999) Cyber disorders: The mental health concern for the new millennium. CyberPsychology & Behavior 2: 475–479.
8. Kim MO (2001) A study on the effects of family resilience of adoption of family of children with disabilities. Korean J Family Social Work 8: 9–39.
9. Online news. Youth serious smartphone addiction. 2013. Available from http://www.kyeonggi.com/news/articleView.html?idxno=675154.
10. Moskowitz HR, Gofman A, Beckley J, Ashman H (2006) Founding a new science: Mind genomics. Journal of sensory studies 21: 266–307.
11. Green PE, Krieger AM, Wind Y (2001) Thirty years of conjoint analysis: Reflections and prospects. Interfaces 31: S56-S73.
12. Gofman A, Moskowitz H (2010) Isomorphic permuted experimental designs and their application in conjoint analysis. Journal of Sensory Studies 25: 127–145.
13. Brunken R, Plass JL, Leutner D (2003) Direct measurement of cognitive load in multimedia learning. Educational psychologist 38: 53–61.
14. Jain AK, Murty MN, Flynn P (1999) Data clustering: a review. ACM computing surveys (CSUR) 31: 264–323.
15. Bakhtin M (1984) Rabelais and his world (Vol. 341). Indiana University Press.
16. Van Deursen AJ, Bolle CL, Hegner SM, Kommers PA (2015) Modeling habitual and addictive smartphone behavior: The role of smartphone usage types, emotional intelligence, social stress, self-regulation, age, and gender. Computers in human behavior 45: 411–420.

DOI: 10.31038/ASMHS.2019312

Abstract

We present an exploratory study to understand the mind of people with respect to what they want from caregivers. Using experimental design of ideas, we present typical respondents with mixtures of ideas about caregivers, obtain a response, and deconstruct the response to the contribution of the component ideas. The decomposition revealed three mind-set segments: Devotion to the job and to the well-being of the patient; Treat them like family; Focus on empathy, respect, and competence. The mind-sets can be found in the population at large through the use of a PVI (personal viewpoint identifier). The study using Mind Genomics opens up the possibility of better understanding the mind and inner world of professional caregiving, and suggest the different psychological needs and wants of the professional caregiver, a long-neglected group.

Introduction

As chronic illness is expanding and life expectancy is growing the focus on ideals for professional caregivers for the aging ill is introduced. Ideals are translated into daily conduct and living circumstances that are vitally linked to well-being and life quality. Ideals held by nursing homes often do not represent ideals held by professional caregivers [1].

Since agitated patients create a heavy burden on professional caregivers, qualified professional caregivers perceived subjective values such as alleviating anxiety and striving for a peaceful and calm environment as very important [2]. Objective values that are linked to contributions of professional caregivers to a high quality of life were: empowering patients to choose for themselves; going outside; offering pleasant experiences, recognizing their mental world as being authentic; encouraging being active and; facilitating the freedom of movement [2].  There may be conflicting ideals in some cases between safety and being active or empowerment and provoking anxiety.

Daily caregivers, however, were mostly concerned with promoting pleasant experiences for patients [1]. Qualified professional caregivers were characterized by being attentive, empathetic, understanding; listening and assuring Safety. There is a paucity of research on the public expectations from caregivers. This study stimulates a discussion as to attributes of professional caregivers in the framework of a good life quality.

Since care and assistance for the ill are no longer limited to compensating for the functional consequences of the chronic illness, caregiving today aims at preserving quality in life of people with chronic illness, particularly in psychogeriatric. Thus, professional caregivers aspire to create the best possible quality of life to clients they care for.

In the literature dimensions of quality of life are: safety, privacy, self-determination and freedom, being useful/giving meaning to life and spirituality [3]. But the strongest effects on quality of life as shared by chronically ill people with professional caregivers are health and illness, mobility, deafness, being able to do less and less, not knowing the way anymore and forgetfulness [4].

Interviewed people with caregivers who fully understood questions outlined various aspects of caregivers that contribute to their life quality: cheerfulness, happiness, being happy with life, humor, tranquility, being allowed to express positive feelings and/or being approached by others in a positive manner. Chronically ill people at home mentioned nature, good and bad weather, and listening to classical music as contributing to life quality [5]

Nursing home resident with caregivers added self-esteem, self-image, being involved in the things around you, living in the midst of your family, feeling attached, being understood and being accepted as positively influenced their quality of life [6]. People who live in nursing homes for long derive much support from relationships and for them seeing the grandchildren, the partner, knowing that they are doing well and that the contact with them is good greatly affects their life quality [7].  Residents of nursing home perceived attention they are get, making friends, feeling loved, communication, one-on-one contact, and contacts with professional caregivers as mostly influencing their quality of life [8,9].

In addition, they also mention hobbies such as reading, watching television, watching movies, taking walks and going on vacation. The absence of favorite activities decreases life quality.

This study examines perceptions of people as to what makes a good professional caregiver.  This study aims at developing the mind-genomics of caregiving attributes that affect quality of life. The science of Mind-Genomics will provide preconditions for making professional caregiving more effective, customer-driven, customer-oriented to enhance quality of life.

Issues with Knowing what is Important

Discovering what is important to people has become increasingly important in our service-oriented economy. As individuals live longer, have more disposable income, and more choices, a knowledge about what people ‘want’ becomes a strategic advantage both for driving choice, and for driving satisfaction.

There is no shortage today of studies on what is important. In virtually every sphere of human endeavor, researchers, business people, and even those who provide the specific service want to need, and in fact need to know ‘how am I doing?’ and ‘what is important to you?’ The enthronement of such information is such that there is a whole field of science and application called ‘customer satisfaction,’ featuring questionnaires, scoring methods, and so forth.  On the simplest side, consultant Fred Reicheld, for example, has stated that there is only one question that one needs to answer about satisfaction, and that is ‘would you recommend this to your friends?’ [10] On the other, more complex side, are various handbooks of marketing and services scales [11].

Simply having a plethora of questions and scales does not tell us what is important. We know that the scales cover different aspects of service, but there is also the ongoing realization that people differ from each other. Marketers have long since recognized that these person-to-person differences are not random, but may come from fundamentally different groups in the populations, psychographic segments [12].

The Contribution of Mind Genomics

Mind Genomics is a recently emerging science which deals with the way people make decisions. The fundamental notion is that individuals have frames of reference for the products and experiences of everyday life. These frames of references are unknown but can be uncovered experimentally by presenting respondents with different combinations of statements about a situation of experience, obtain ratings, and deconstruct the response to the contribution of each statement. When the statements with specifics of an experience, one sees quickly which elements are ‘important’ and which are either only modestly relevant or even complete irrelevant.

Through experiments, Mind Genomics continues to the reveal that for almost all situations encountered in normal life, there are a variety of different frames of reference, or mind-sets. These mind-sets can be uncovered through experimentation, and specifically by clustering the pattern of respondents for different individuals.  Mind Genomics reveals that for most situations, there are a limited number of mind-sets or basic frames of references, usually two or three, occasionally one or two more.

For this study, the Mind Genomics experiment involves a set of steps, beginning with a topic (what makes a good caregiver), proceeding to four questions, and then providing four answers to each question, or a total of 16 answers. The four questions ‘tell a story,’ and are used to elicit the four answers. It is the answers which provide specific information.

Table 1 presents the four questions and the four sets of four answers. By the nature of caregiving, there are many more facets to explore. The objective of Mind Genomics is to provide cartography of the situation, and not an exhaustive, complete answer. Mind Genomics has been designed to understand limited parts of an experience in a way that is easy, quick, inexpensive, and instructive.  Thus, in the world of Mind Genomics, it is not one long, expensive, comprehensive study which provides the answer, but rather a series of short, simple, focused studies, whose data provide, in combination, much of the information needed to understand the topic.

Table 1. The four questions, and four answers for each question regarding what makes a good caregiver

The Test Stimuli – Vignettes

In most survey research the respondent is presented with one question at a time, and instructed to answer that question. This type of research relies on the memory of the respondent, as well as on the attitude of the respondent towards the specific topic. There is always the problem that the respondent will answer the question in the way that the respondent feels the question ‘should be answered.’  This ‘interviewer’ bias occurs because the respondent wants to please the interviewer, as well as be considered to be appropriate and ‘politically correct.’

Mind Genomics gets around the problem of interviewer bias and political correctness, giving the appropriate response, by present the stimuli in the form of combination statements (so-called vignettes), getting the respondent to rate the entire vignette as a single entity, and then deconstructing the response to a set of vignettes into the part-worth contribution of the component statements using OLS, ordinary least squares regression.

The intellectual origins of the Mind Genomics approach, working with combinations and then deconstructing the response, come from the field of mathematical psychology known as conjoint measurement [13] The underlying notion is that people respond to combinations of messages in everyday life, knowing almost intuitively what they like, and what they do not like. In daily life the norm is to be exposed to these combinations, and for decisions to emerge rapidly, without conscious deconstruction into the components. Of course, the respondent can always justify the response to these combinations, but the judgment is automatic. And thus, Mind Genomics mirrors the behavior of ordinary life.

The test vignettes are created by experimental design. For this particular version of Mind Genomics, comprising four questions or silos, and four answers or elements, the experimental design dictates 24 combinations, with each of the combinations of vignettes comprising 2–4 answers or elements. Each vignette comprises at most one element or answer.  The design ensures that most of the combinations are incomplete.

Table 2 shows the basic experimental design for one respondent. Each respondent, in turn, evaluates a unique set of 24 vignettes, created by different versions of the experimental design. The structure of the underlying experimental design is maintained. The only difference is the specific combinations, which differ from respondent to respondent. This strategy ensures that Mind Genomics covers a wide number of alternative combinations in the study, a strategy similar to the MRI for studying the brain, which takes many ‘pictures’, combining them to form a picture of the brain.

The Respondent Experience

Table 2. Experimental design underlying the vignettes

Mind Genomics studies are executed on the web. The respondent is invited through a panel provider, here Luc.id, Inc. The panel provider specializes in the recruitment of respondents for these types of short studies. For this study we simply requested an approximately even break in terms of the proportion of male versus female respondents, and an approximately even distribution across ages.

The actual experience, taking a total of 4–5 minutes, began with the invitation. The respondents, motivated to participate by their membership in the Luc.id panels, responded to the email invitation by clicking an embedded link. The respondents were led to an orientation page, explaining the study in general terms.  The respondents then completed a classification page, requesting information about age, gender, and a third classification about whether the respondent was a caregiver. About half the respondents participate as caregivers in one way or another (26 of 56 respondents.)

Figure 1 shows an example of the vignette. The vignette presents the introduction, the 2–4 elements or answers centered and stacked on atop the other, and then the rating question at the bottom. The format makes it easy for the respondent to examine the vignette and assign a rating. The information presented is most important. Of far less importance is the way the vignette ‘looks.’ As long as the respondent can locate the relevant information, the format does its job.

Figure 1. A typical vignette as it appears on a respondent’s smartphone.

When the respondent selects a rating, there is no need to press ‘next.’ The study automatically progresses to the next vignette, a feature which makes the study less onerous to the respondent, who needs to do far less work.  The vignette is written in the language that most people find is comfortable with smartphones, namely with abbreviations (u for the word ‘you.’) As society progresses increasingly towards smartphones, such changes in usage, and even hitherto ‘incorrect spellings and diction’ are becoming the norm for regular conversation and texting. We felt it important to adopt the language of the everyday, rather than to make the interview more formal in its language.

Preparing Mind Genomics Data for Analysis through the Binary Transform of the Rating Scale

Although the use of category (Likert) scales is widespread, all-too-often it is unclear to the user of the scale what the scale means. A great deal of effort may be expended on assigning names to the scale points in order to make the scale meaningful to those who must make practical decisions with the results. A good example of this effort is the work on assigning the proper names to the nine scale points of the so-called Hedonic Scale [14].

An alternative to the nine-point scale used here is to convert the scale to binary, with a convention established by author Moskowitz for 35 years, since 1984. The convention is to transform the ratings of 1–6 to 0, and the ratings of 7–9 to 100, and then add a very small random number to each transformed value, the small random number being in the vicinity of (10–5.)  This transformation makes the results ‘binary,’ no or yes, with the subsequent property that anyone can now understand the meaning of the response. The transformation may be stricter (1–7 transformed to 0) or less strict (1–5 transformed to 0), but the effect is the same. The results are easier to understand by scientists, managers, and the general readership, who are accustomed to issues with a binary choice, no versus yes, respectively. The transformation reduces some of the metric information, but the interpretability of the results more than makes up for that loss of metric information and ‘discriminatory fineness.’

What Resonates in The Mind of the Respondent Regarding a Professional Caregiver?

The first analysis uses OLS (ordinary least-squares) regression. The data from all 56 respondents are included in the analysis, with the total number of cases or observations totaling 1,344 (56×24.) The regression modeling does not pay any attention to which the respondent IS, nor the order of testing. All the observations are treated equally.  The regression model tried to fit a linear equation of the form:

Binary Rating = k₀ + k₁(A₁) + k₂(A₂)…k₁₆(D₄)    The parameters of the model for the total panel is shown in Table 3.

Table 3. The performance of the elements on Question 1: How much do you like the caregiver

The foregoing equation states that the binary rating (our transformed variable from the original scale) is equal to the additive constant (k₀), and 16 individual weights or coefficients, one for each of the 16 elements.

The additive constant tells us the conditional probability or percent of responses expected to be 7–9 in the absence of elements. Of course, all vignettes comprised a minimum of two and a maximum of four elements, respectively, meaning that the additive is simply an estimated parameter. Nonetheless, it is a useful indicator of the degree of predisposition to like a professional caregiver. For our data the additive constant in Table 3 is 74.14, meaning that even without elements, the odds of a person liking a professional caregiver is 74%.  It will be the elements which do the work.

Looking down the column labelled ‘Coefficient,’ we have sorted the 16 elements from high to low, Despite the very high constant, there is one strong performing element, D2, ‘show them love .. Gratitude, let them feel at home like family.’ This element has a coefficient of 7.73.  There is one other strong element, A4, ‘be understanding, self-sufficient, and very patient.’

Not every element drives liking. Some elements, those having to do with the problems or issues encountered with residents, those who are being taken care of, generated negative coefficients. These are elements that are not liked. They are from Question or Silo B, ‘How do you know that a resident needs help?’

B3      most residents are unable to speak so it’s also good to pay attention

B1      sometimes they tend to pace back and forth which can be very unusual

B2      they make uncomfortable noise which can be a sign for medical treatment or simply need to be taken to the bathroom

The coefficients do not describe the data perfectly. The model shown in Table 3 is known as a cross-sectional model, which uses the raw data. We are interested, of course, in the coefficients, but also in the degree to which we can believe that the coefficients represent a ‘real contribution’ to the rating.  The degree to which the coefficients represent a departure from 0, the 0 signifying no contribution, comes from the t-statistic and the p-value, shown in the right-most two columns.

Every one of the coefficients comes from what is known as a sampling distribution. What we observe in Table 3 in terms of the value of the coefficient is only one value from many values that the coefficient could take on.  Were we to repeat the experiment or study 100 times, would we get a coefficient that is not closer to 0, or even 0 itself? That question is answered by the t-statistic, which is the ratio of the coefficient that we observe to the standard error of the coefficient, i.e., to the standard deviation of the coefficient were we to do the study 100 times, or so. We look for high t-statistics, preferably 2.00 or more, but at least 1.6 or more. That ratio tells us that the ratio of the signal to the noise, i.e., the ratio of the coefficient to the standard error of the coefficient, is reasonably high.  In turn, when we work with a t-statistic around 1.64 or so, we have a 10% probability that the ‘real’ coefficient is 0. We accept the coefficient as significant, as important.

As a matter of experience, the coefficients are typically not particularly significant for the total, but tend to become more significant with subgroups, especially mind-set segments. As a rule of thumb, we look at coefficients of 7.51 or higher (8 in rounded format) as being important, signaling that the element plays an important role to people. Previous observations in many studies by author Moskowitz suggest that coefficients around 8 correspond to elements which are relevant to people who make decisions based on these elements.

Gender Differences

Men and women are identical in their basic liking of a professional caregiver. Their coefficients are  76 for males, and 73 for females, respectively.  It is in the specific elements which drive liking that we see the differences (Table 4). Men want the interaction to be efficient, speaking with respect so that the person cared for is heard. It is on the activity itself. In contrast, women want to see that there is an emotional connection, patience, and love. It is not so much on being efficient as in bonding.  The same difference between efficient/effective activity and emotional understanding/bonding occurs for those elements which are not strong for either gender, but elements showing large differences in the coefficient. A good example of this is D3: show acknowledgment instead of letting them feel invisible.

Table 4. Gender differences for the performance of the elements on Question 1: How much do you like the caregiver

Age Differences

We see strong effects due to age. Table 5 shows that the additive constant is high across the four age groups, ranging from 17 to 80. Table 5 is arranged in descending order to highlight the remarkable differences.  All additive constants are high, 70 and above, but the additive constants of the youngest respondents, ages 17–25 is remarkably high, 90. This very high additive constant suggests that the younger respondents are prepared to like virtually any description of a professional caregiver.

Table 5. Age differences for the performance of the elements on Question 1: How much do you like the caregiver

The age differences emerge quite strongly when we look at the different ages:

Age 61–80 respond to statements about respect and warmth

Age 41–60 respond to statements about patience

Age 26–40 respond to statements about being dedicated to the job

Age 17–25 find nothing important in a positive sense, but don’t seem to want to hear about practical issues with which caregivers must deal

Emergent Mind-Sets

One of the hallmarks of Mind Genomics is its focus on underlying mind-sets or ways of looking at the world, as a key way to understand a topic, and differences in judgments. What may seem to some to be ‘irrational behavior’ exhibited by some individuals may, in fact, simply be the fact that the individual has a different frame of reference, and different weights in the criteria.

Tables 4 and 5 show differences in the importance between the genders (Table 4) and across ages (Table 5.) It may well be that the differences among people are deeper, comparable to differences among people in terms of genes. The name ‘Mind Genomics,’ in fact, is taken from the metaphor of genetic differences, not applied to the physical chromosomes of people, but to the way different people focus on the same situation, but different in what is important.

In order to uncover the mind-genomes, or ‘mind-sets,’ we simply cluster the 16 coefficients generated for each respondent. Mind Genomics allows us to create an individual-level model for each respondent relating the presence/absence of the 16 elements to the binary rating of ‘Like the caregiver.’ Each individual is one object among a set of 56 objects. Clustering, a well-accepted statistical method, divides the 56 objects, our respondents, into a small set of non-overlapping groups. The criteria for division is the minimization of ‘distance’ between pairs of our 56 respondents, with the property that the emergent clusters or mind-sets be both parsimonious (the fewer the better), and interpretable (the clusters tell meaningful stories)

Our data suggest three separate clusters or mind-sets, shown in Table 6. The division into the clusters or mind-sets

Table 6. Emergent mind-sets for the performance of the elements on Question 1: How much do you like the caregiver

Mind-Set 1 – Be devoted to the job, and to the well-being of the patient. This mind-set shows the lowest additive constant (63), but strongly to a few of the elements, and not to the others. These respondents react strongly to devotion and professionally competent behavior.

Mind-Set 2 – Respond to caregivers who treat their patients as family. They show a higher additive constant (75).

Mind-Set 3 – Respond to a focus on empathy, respect and competence. This mind-set shows the highest additive constant (82.)

There are six of the 16 elements which do not drive a strong response by any of the three mind-sets.

Beyond Liking to Decision Making

One needs to observe everyday life for just a moment to realize that most of the decision is either unconscious, or virtually automatic. One can always inquire ‘why’ a specific decision or action was done, but the reality of life is that without the automatic behaviors that we exhibit we simply could not function the environment around us is simply too complicated, with too many conflicting cues

The previous tables showed us that there is meaningful order in a person’s reaction to what is considered a good professional caregiver. That is, there is a story which seems to emerge. The differences between the genders, among the ages, and among the mind-set segments make sense.

What we don’t know is the speed with which the decision is made, and whether or not there is a relation between how much someone ‘likes’ a message and how ‘fast’ the person processes that message. Liking something may be entirely different from processing. Valences, emotions, may have something or perhaps nothing with how quickly we react, although when it comes to food quite often we will observe a disgust reaction quite quickly.

The data that were collected from this study, ratings, were accompanied by another, parallel type of information, response times.  Each vignette appeared on the screen and was rated. We have already dealt with the relation between the elements in the vignette and the rating, or more correctly, the binary transform of the rating. We now turn to the response time.

How Fast People Respond to these Vignettes – a ‘Morphological Analysis’ of the Data Patterns

As in everyday life, people do not focus on what they read, but simply pay attention, and make a judgment. Our previous data looking at the ‘liking’ rating suggest that the data makes ‘sense.’ Some of the 56 respondents take a long time to respond, others take a short time to respond.’  Figure 2 shows that there is a distribution of average response times across respondents (ordinate) as well as a distribution of average rating times across the same respondents. There is no simple relation between the two. People who take longer to rate the vignettes are neither more accepting nor more rejecting of what they read. The pattern appears random.

Figure 2. Scatterplot plot showing the average rating of liking on the 9-point scale (abscissa) versus the average response time (ordinate.) Each filled circle corresponds to one of the 56 respondents.

A clearer picture of group to group differences emerges when we compute the averages, specifically of the rating question (liking), the binary transformed rating, and the response time.  Table 7 shows that the groups do not differ very much in terms of the 9-point rating (range of 7.1 to 7.7), somewhat more for the binary transformed rating (69 to 85), and most dramatically in terms of response times.   For response times, males respond more quickly than do females, younger participants respond more quickly than do older respondents, and the mind-set segments respond at different speeds, with Mind-Set 1 (devotion to the job and to the well-being of the patient) responding most quickly and Mind-Set 3 (focus on empathy, respect and competence) responding most slowly

Table 7. Average ratings, binary transformed ratings, and response times, for total and subgroups.

Response times tell about how long a respondent requires to ‘process’ the information. We do not know the neurophysiological correlates, but we can surmise that those respondents requiring a longer processing time as somehow considering the message in a different way, especially when we have response times of over a second or two.

In previous studies (unpublished observations) it appears that the first position may be very ‘noisy’ with respect to response time. We can eliminate the ‘noise’ by considering the response without the measures in position #1, the first position in the set of 24. Table 7 shows that there about a change of 0.3 seconds, three tenths of a second in the average response.

Deconstructing Response Time

The experimental design allows us to deconstruct the measured response time into the contribution of the individual response times. Since we were not able to monitor the respondents, we could not determine whether the respondent was multi-tasking, a behavior which would lead to very long response times. In order to remove the biases due to multitasking, which could make the response time go from a few seconds to a few hundred seconds, we arbitrarily set a cut-off of 15 seconds. Any response time of 15 seconds or higher was brought down to 15 seconds.  A few respondents generated response times in the hundreds of seconds, but the majority of response times were far shorter, as Figure 3 shows.

Figure 3. Distribution of response time for the 56 respondents, and the vignettes in position 2–24.

A key emerging aspect of Mind Genomics is the focus on possible neurophysiological correlates of ratings. One of the most popular of these neurophysiological measures is ‘response-time,’ which is presumed to reflect underlying processes.  Table 8 shows the results from OLS regression, for the total panel. All of the vignettes were included in the regression modeling, which related the presence/absence of the 16 elements to the response times.  The model is expressed as a simple linear function, without an additive constant.

Rating Time (Seconds) = k₁(A1) + k₂(A2) … k₁₆(D4)

The pattern in Table 8 suggests a ratio of about 2.5/0.8, or 3/1 in terms of the number of seconds required to process the different messages. By inspection, it appears that the shorter elements take less time to process. The very longest is a sentence which requires additional cognitive processing, first because it is long, and second because it has a second and third clause, respectively. Each clause must be processed to understand the full meaning of the sentence.

Table 8. Response time deconstructed into the contribution of the individual elements (coefficient), as well as the statistical significance of the coefficient (t-statistic)

First clause: they make uncomfortable noise

Second clause:  which can be a sign for medical treatment

Third clause:  or simply need to be taken to the bathroom

It may be that each clause takes roughly about 0.8 seconds or so to process if it has a major idea. A minor idea might add another 0.4 seconds, rather than another 0.8 seconds.

Does response time co-vary with liking?

Do people respond more quickly to what they like?  If they do in the aggregate, does this co-variation remain when we look at different subgroups, such as gender, age, and mind-set, respectively?  The answer to this question requires that we deconstruct the binary-transformed ratings into the contribution of the 16 elements, and the response times into the contributions of the 16 elements, do the two analyses separately. Each analysis generates a coefficient for each of the 16 elements. We plot the coefficient for response time against the corresponding coefficient for binary-transform rating.

The results appear in Figure 4A for total panel, Figure 4B for gender, Figure 4C for age, and Figure 4D for mind-set.  The results tell a story, but one whose character changes with the r group being analyzed. There is no suggestion of total randomness, however, although there are plots showing a great deal of noise.

Figure 4A. How response time covaries with liking. The darkened circles correspond to the coefficients of the 16 elements.

Figure 4B. How response time covaries with liking. The darkened circles correspond to the coefficients of the 16 elements. The plot is by gender

Figure 4C. How response time covaries with liking. The darkened circles correspond to the coefficients of the 16 elements. The plot is by gender

Figure 4D. How response time covaries with liking. The darkened circles correspond to the coefficients of the 16 elements. The plot is by segment or mind-set

Total Panel (Figure 4A) – As an element is liked more, it is responded to faster.

Males (Figure 4B) – As an element is liked more, it is responded to faster.

Females (Figure 4B) – Response time and liking covary randomly;

Age 17–25 (Figure 4C) – Response time and likin covary randomly.

Age 26–40 (Figure 4C) – As an element is liked more, it is responded to more slowly, suggesting the item is being considered, perhaps for a relative.

Age 41–60 (Figure 4C) – As an element is liked more, it is responded to more slowly, suggesting the item is being considered, perhaps for a relative.

Age 61–80 (Figure 4C) – As an element is liked more, it appears to be responded to more quickly, but the relation is quite ‘noisy.’

Seg1 (Figure 4D) – Devotion to the job and to the well-being of the patient. The relation appears to be almost random, certainly very noisy.

Seg2 (Figure 4D) – Treat them like family. The relation appears to be almost random, as well.

Seg3 (Figure 4D) – Focus on empathy, respect, and competence.  The relation appears to be almost random as well.

It appears that the difference in processing speed is not a function of the respondent’s mind-set, but rather a function of WHO THE RESPONDENT IS.

Finding Mind-Sets from Mind-Genomics in the Population

This first paper in the Mind Genomics effort to understand and improve professional caregiving has easily revealed at least three mind-sets in the population of individuals who may some day be responsible for hiring caregivers for themselves or for an ill relative.  Mind Genomics suggests that although almost all of the idea or messages about caregivers are either modestly or strongly positive, that surface positivity is not the case when we dig into the details, and uncover mind-sets. The mind-sets, pervasive in the population, respond to different aspects of the caregiver’s attitude and job. What pleases one mind set may not please another mind-set. We do not see radical opposite points of view, as we might for foods of various types (e.g., spicy versus bland foods), but we see the opportunity to optimize the fit of a caregiver with the person who hires that caregiver.

How can a person find out who she or he IS if a caregiver, and whom she or he WANTS when hiring a caregiver?  One way is to assign a new person, either client or caregiver, to the mind-set, through a short questionnaire, the PVI, the personal viewpoint identifier. The identifier, created by author Gere, uses the 16 coefficients for the three mind-sets respectively (Table 6), and creates a set of questions to be answered NO or YES. The pattern of responses to the questions assigns a person, either client or caregiver, to one of the three mind-sets.  Figure 5 shows the PVI, with the top part showing the test itself, and the bottom showing the three outputs, which can be either given to the caregiver and/or to the person hiring the caregiver.  The objective in future Mind-Genomics studies on caregiving is to expand the scope of the different aspects of caregiving, and for each probe far more deeply. The present study is thus the ‘first salvo’ in that effort.

Figure 5. The PVI (Personal Viewpoint Identifier) for Professional Caregiving

Discussion and Conclusion

In this exploratory study of responses to professional caregivers, we have opened a new area in the emerging science of Mind Genomics. The topic is the response of potential clients of professional caregivers to what the caregiver should do, and what constitutes preferred behavior and conduct versus behavior and conduct that are disliked. The elements chosen for this initial study were all positive, developed by the senior author, Ellis, a professional caregiver, is author Frazier.

With the rapid aging of the population, the increase in dementia and other debilitating illnesses, studies of this type are called for in the world of caregiving. With caregivers currently at a financial disadvantage, it is hoped that this first paper on Mind Genomics and what is desired by the population of a good caregiver can become a stimulus for recognition of the very valuable service, and a tool for continual improvement and increased professional and public recognition of their efforts.

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

References

1. Kalis A, Schermer MH, van Delden JJ (2005). Ideals regarding a good life for nursing home residents with dementia: views of professional caregivers. Nursing ethics 12: 30–42.
2. Zingmark K, Sandman PO, Norberg A (2002) Promoting a good life among people with Alzheimer’s disease. J Adv Nurs 38: 50–58. [crossref]
3. Ettema TP, Dröes RM, Lange J de, Ooms ME, Mellenbergh GJ, et al (2005). The concept of quality of life in dementia in the different stages of the disease. International Psychogeriatrics 17: 353–370.
4. Thorgrimsen L, Selwood A, Spector A, Royan L, De Madariaga Lopez M,Woods RT, Orrell M (2003) Whose quality of life is it anyway? The validity and reliability of the Quality of Life-Alzheimer’s Disease (QOL-AD) scale. Alzheimer Disease and Associated Disorders 17: 199–200
5. Dröes, R. M., Boelens-Van Der Knoop, E. C., Bos, J., Meihuizen, L., Ettema, T. P., Gerritsen, D. L., & SchöLzel-Dorenbos, C. J. (2006). Quality of life in dementia in perspective: An explorative study of variations in opinions among people with dementia and their professional caregivers, and in literature. Dementia, 5, 533–558.
6. Dröes RM, Meiland FJM, Schmitz M, & Tilburg W van (2004). Effect of combined support for people with dementia and carers versus regular day care on behaviour and mood of persons with dementia: Results from a multi-centre implementation study. International Journal of Geriatric Psychiatry  19: 1–12
7. Lyketsos CG, Gonzales-Salvador T, Chin JJ, Baker A, Black B, & Rabins P (2003) A follow-up study of change in quality of life among persons with dementia residing in a long-term care facility. International Journal of Geriatric Psychiatry 18: 275–281.
8. Ready RE, Ott BR (2003) Quality of Life measures for dementia. Health Qual Life Outcomes 1: 11. [crossref]
9. Ready RE, Ott BR & Grace J (2004). Patient versus informant perspectives of quality of life in mild cognitive impairment and Alzheimer’s disease. International Journal of Geriatric Psychiatry 19: 256–265.
10. Reichheld FF (2003) The one number you need to grow. Harv Bus Rev 81: 46–54. [crossref]
11. Gilmore A and McMullan R (2009) Scales in services marketing research: a critique and way forward. European Journal of Marketing 43: 640–651.
12. Wells WD  ed.,( 2011). Life style and psychographics. Marketing Classics Press.
13. Green PE, Krieger AM and Wind Y, (2001) Thirty years of conjoint analysis: Reflections and prospects. Interfaces, 31(3_supplement), S56-S73.
14. Peryam DR and Pilgrim FJ (1957) January The hedonic scale method of measuring food preference. In Food technology 11, 32–32).