Article Type : Research Article
Authors : Cesar Salazar O
Keywords : Credit Decision making; Human behaviour; Eye tracking; Pupillometry
Studying human behaviour when making economic decisions is
important because they are biased by emotions and instincts. Our approach
consists of contributing to the demonstration that non-rational brain processes
do influence decision-making and that the study of human behaviour and its
decisions should be involved with new technologies such as the eye tracker.
This research focuses exclusively on the use and relevance of eye tracking
within selective visual attention processes, in response to any marketing
stimuli, such as exposure to brands of products known or unknown to the medical
doctor. There is ample literature that supports that gaze fixations and pupil
diameter are associated with cognitive and emotional processes, showing that
large dilations in the diameter of the pupil are associated with positive
choices, compared to small dilations in the diameter of the pupil that suggest
negative choices. It was possible to establish that the diameter of the pupil
acted as an emotional indicator before the presentation of a promotional video
of a cream for the healing and treatment of skin wounds. Additionally, it was
determined that gender is an explanatory variable for the difference in pupil
diameters and that gaze travel and fixations in certain areas of interest,
allow us to discover two important conclusions. The first being that the count
of fixations together with their duration, are indicators of attention
generated in that particular area, the second is related to the importance of
the correct measurement of the diameter of the pupil that allows to establish
if that fixation that produced attention, is due to an emotional response of
acceptance or rejection.
Human behavior and its complexity
go beyond traditional observation techniques. Studying human behavior at the
time of making economic decisions is important to adjust the offer of companies
to the real needs and desires of customers [1,2]. But from the perspective of
behavioural economics, human decisions are emotionally biased, how can you
determine if the decision made is emotional, rational or a mixture of the two,
and in what proportion? With which it is necessary to establish research
methodologies consistent with this assumption. The purpose of this article is
to determine whether the analysis of pupillometry is applicable to establish
whether the decision made by individuals obeys an emotional or rational
process. For this purpose, the eye-tracker technology, was applied in the
laboratory work [3].
Based on the
foundations of the school led by Kahneman and Tversky, where the decisions of
human beings are biased by emotions and instincts, our approach consists of
contributing to the demonstration that non-rational brain processes do
influence decision-making. Of decisions and that the study of human behaviour
and its decisions should be involved with new technologies such as the eye
tracker. For this purpose, there is abundant literature that mainly raises how
the dilation of the pupil in humans is associated with brain processes of
approach / flight to external stimuli. Specifically, the diameter of the pupil
is an indicator of measurement of the attention processes of the human being
[4-8]. Additionally, there is ample literature that supports that gaze
fixations and pupil diameter are associated with cognitive and emotional
processes [9]. On the other hand, for they also suggest that the increase in
the diameter of the pupil is involved in the processes of visual attention.
Therefore, large dilations in the diameter of the pupil are associated with
positive choices, compared to small dilations in the diameter of the pupil that
suggest negative choices.
The regulation of
emotions is essential for adaptive behaviour and mental health. He investigates
the effects of different emotion regulation strategies on pupil dilation, skin
conductance responses, and subjective emotional responses. These results
indicate that the pupil diameter is modulated by emotional arousal, but is
initially related to the extent of mental effort required to regulate automatic
emotional responses [10].
that the diameter of the eye's pupil indexes the
modulation of the state of arousal and responds to a wide variety of cognitive
processes, including mental effort, attention, surprise, decision processes,
decision biases, value beliefs. , uncertainty, volatility, exploitation /
exploration, compensation or learning rate [11]. In the context of decision
making, pupil dilation varies in contexts of uncertainty / certainty. When
decisions are made in the absence of uncertainty, as in simple
stimulus-response association tasks, the relationship between response and
information gain is straightforward. In conditions of uncertainty, the situation
is a bit more complex. Zenon highlights this observed relationship between
reaction time and pupillary dilation, they are best modelled by repressors that
span the entire reaction time period of the tests, rather than limited short
pulses at the start of the stimulus. These findings suggest that the process
from which pupillary dilation originates is maintained throughout the decision
process. The pupil dilates when we make decisions and these fluctuations in
pupil size reflect decision-making calculations during and after an election.
Like most decisions in real life, they are guided by the results of previous
elections. These results show that fluctuations in pupil size can provide
detailed information on the calculations underlying value-based decisions and
subsequent updating of value beliefs [12].
The rise of new
technologies that allow human-computer interaction, known by its acronym in
English as HCI -Human Computer Interaction, are becoming achievable for
marketing, helping to control mismatches or errors in the results of consumer
studies, which can also be substantially minimized with the application of
these new technologies [13]. Within the great boom of new technologies and
since the early eighties, the eye tracking process or eye tracking has been on
the rise. Various fields of study cover the application of this tool, from
neurological research to usability or experience research. This research
focuses exclusively on the use and relevance of eye tracking within selective
visual attention processes in response to any marketing stimuli, such as
exposure to brands of products known or unknown to the doctor. Taking into
account the above, it is pertinent to clarify how HCI helps us to capture data
on eye movements. In this case, these are two simple things, but they are
usually confused [14]. The Eye-tracking is the software with which the data
obtained by the Eye-tracker are processed, which is the hardware in charge of
directing recording the eyes. The Eye-tracker's job is to record these types of
movements at actual speeds and accelerations. The human eye is a large
detection sensor that moves at almost imperceptible speeds [15,16]. Its field
of view is not that extensive, it is a relatively small field composed of an
ellipse one hundred and eighty degrees (horizontal), versus 130 degrees (vertical).
The precision of the human visual field is less than two degrees, which has
been called the foveal focus area and is the one that concentrates or focuses
the main visual field, followed by the parafoveally area that concentrates
between two and five degrees of visual acuity and it ends with an area greater
than five degrees that is called the peripheral area. In this sense, it can be
said that when human beings want to see or focus with precision, they do so in
the foveal area and if they want less precision, they move to the parafoveally
area and then to the peripheral area. In order to focus our vision on something
of our interest, two types of eye movements come into action, saccades and
fixations [17,18]. The fixations are responsible for placing the retina on a
stationary target to process it and the saccadic is the movement whose main
task is to stabilize the eye in order to focus or “stabilize” the retina. They
are extremely fast movements measured in milliseconds.
Two eye-tracking measures that can be used to
study cognitive development and plasticity: pupil dilation and spontaneous
blink rate [19]. Gaze analysis, which can reveal the current focus of attention
as well as cognitive strategies, pupil dilation is modulated by the
locuscoeruleus-norepinephrine system of the brain, which controls arousal and
physiological attention, and has been used as a measure of subjective
difficulty of homework, mental effort, and neural gain [20]. The spontaneous
blink rate correlates with dopamine levels in the central nervous system, and
may reveal goal-directed behaviour and learning processes. Eye tracking has
largely occupied brain imaging research as a way to study the mechanisms
underlying behaviour. How visual monitoring has been and could be extended to study
cognitive development. For visual monitoring, the use of eye-trackers are
necessary. The eye-tracker is the Hardware in charge of directing recording the
eyes. Your job is to record these types of movements at real speeds and with
realistic accelerations. Visual tracking, sampling rates ranging from 25 to
2000 measurements per second, meaning that the fastest trackers achieve
temporal resolution of less than milliseconds, similar to the Davidson (1988)
EEG. It is here where the use of high-tech devices (Eye-trackers) becomes
relevant, which really record the movements of the eyes and do not de-calibrate
in the face of sudden or involuntary movements of the head. After the data is
captured by the eye-tracker, these data are analysed with a software called
eye-tracking, which in synthesis what allows at least is to group and summarize
the data of the fixations, saccades, coordinates and diameters of the pupils
and blinking. Creating in the light of these data the traces or paths travelled
by the eyes (Gaze Maps) or the areas where the vision or heat maps (Heat Maps)
were mostly fixed. The relevance of using pupillometry as an indicator of
emotional processes can be seen in the previous paragraphs. But for this
purpose, it is also necessary to apply robust and precise eye-tracker
technologies, with which the measurement of the pupil diameter is made easier,
since it is a non-invasive technique, and more precise, with the configuration
of the appropriate eye-trackers.
It is important that
the lighting conditions remain constant, because the pupil reacts to changes in
lighting, and that this condition must be taken into account for the
measurement of changes in it and to be able to explain the capture of the data
as a reflection of the pupillary response to information processing. The
present study is exploratory of a descriptive nature, due to the difficulty
presented for the recruitment of study subjects, who in this case were medical
professionals. 15 appointments were made by medical doctors who will work in
areas where they have permanent contact with patients. Of the 15 doctors to
study, eight were men and seven women. Seven were under 29 years of age, six
were between 46 and 55 years old, one was between 29 and 39 years old, and
finally a doctor over 55 years of age. The pupillometry test was carried out on
an individual, closed space of four square meters of surface, with dim and
constant light through two eco-halogen lamps. Additionally, the walls of the
space used are opaque white, which prevents glare and glare. A 17 '' monitor
and the Eyeteeth company eye-tracker were used (60 Hz), was located 70
centimetres away from each of the participants. Each participant obtained a
calibration of both eyes greater than 80%. The resolution of the monitor used
was 1366 x 768 pixels. At one minute and six-second-long video was projected on
this monitor, showing the presence of a topical cream for the treatment of
burns and skin ulcers. The software used to analyse the data obtained with the
eye-tracker was the Mangold Vision 4.0 eye-tracking. Each one of the
participants was informed before the test what it consisted of and a brief
explanation of the eye-tracker, which is a non-invasive technology without any
risk to the individual's visual health. For the analysis of the data obtained
by the eye-tracker, the attachment aggregation method was used. This method
creates groups around anchor points. For this, at least 3 coordinates must be
found that were recorded within the defined distance from each other (Cluster
size) and have at least a minimum fixation at the moment these three points are
found, all points within this area they are added to a single group. Those
results can be read as real life fixation points and all values outside of
those areas will be ignored. With this method, 1032 aggregated records were
finally obtained.
He was told that a video of one minute and six
seconds would appear. At the end of it, a questionnaire automatically appeared
on the screen, with questions regarding the content of the video. The
questionnaire data was captured through the CAWI application of the company
Tesi [21-23].
Figure 1 shows the AOIs or predefined areas of interest for data analysis. There are five areas of interest determined a priori: those located in the center of the screen that go from attribute one to attribute three, in the area of ??attribute number one appeared the brand of the product and the following sentence: "Bioengineering applied to the healing of wounds”. For attribute number two, images of cells, chains of cells, and the like were constantly appearing. For attribute number three it appeared: "Greater granulation tissue, appearance, color and elasticity similar to normal" and "Wounds, Burns, Skin Ulcers". The fourth area of ??interest was located in the upper left corner of the screen, where the product brand would continually appear, and the area of ??interest located in the lower right corner of the screen, where the laboratory logo always appeared. Pharmaceutical manufacturer of the product to be evaluated. The product evaluated was a topical cream for the healing of skin wounds due to burns. In this article, neither the brand nor the name of the laboratory is presented, for reasons of privacy of this data (Figure 1).
Figure 1: Areas of Interest.
Figure 2 shows the four heat maps or heat maps as a result of the eye tracking tests applied to the men participating in the study. These heat maps reflect the specific moments where the greatest focus of attention was generated in the video. In maps number one and two it is observed that the highest count of gaze fixations occurred for attribute number one and to a lesser extent for attribute number three. At these specific moments in the video there were no reinforcement phrases towards the product. In map number three, the fixation trend towards attributing one continues, and for attribute three, attention is already focused on the explanatory text or argumentation about the healing benefits of the product. In map number four, attention is strongly focused on the healing effect of the cream, that is, an attribute number one (Figure 2).
Figure 2: Heat maps for men.
Figure 3 shows the four heat maps or heat maps made to the women participating in the study. These maps reflect the specific moments where the greatest focus of attention was generated in the video. The resulting heat maps are similar to the behaviour of the gazing for the group of men, with a similar tendency towards great attention to the demonstration of the effectiveness of healing cream, chosen in attribute number one of map number four (Figure 3).
Figure 3: Heat mas for female.
Now, observing in figure 4 the distribution of absolute frequencies by gender, of each one of the AOIs from the heat maps or heat maps of figures one and two, attribute number one obtained the highest number of aggregated fixations, followed by attribute three and two. The brand and logo areas present low counts of added fixations, and of the latter two, the pharmaceutical company logo did not attract the attention of the study participants (Figure 4).
Figure 4: Frecuencies by gender.
Descriptively, some incidence of the gender variable can be seen in each of the areas of interest under study. In this sense, we proceeded with the ANOVA analysis to clarify whether both the gender variable and the area of interest variable, influenced the behaviour of the pupil diameter of the participants. In order to complement this analysis, Figure 5 provides the aggregate distribution of fixations in relation to the diameter of the pupil observed in men and women. The diameter of the pupil for men and women ranges from approximately three to seven millimetres. The data distribution for both men and women does not fit the normal curve. With an average diameter of the pupil smaller in men than in women. This suggests that the process of fixation and attention was greater in women than in men, because the diameter of the pupil in women is greater than in men (Figure 5).
Figure 5: Normal test.
Continuing the analysis
of the incidence of the gender variable and particularly the category of women,
it is pertinent prior to any confirmatory analysis, to apply the normality test
for the pupil diameter through the Shapiro Wilks contrast, this contrast
yielded a significance of 0. 000 with 1032 degrees of freedom and its statistic
of 0.982. Checking what was stated in the previous paragraph of the absence of
normality in the data of the variable pupil diameter.
Having determined the
absence of normality, the analysis of variance (ANOVA) was applied, in order to
determine if there were significant differences in two groups of independent
variables. These groups were the gender variable and the areas of interest
AOIs. Duration time (Fixation) and pupil diameter were chosen for the variables
to be contrasted. Before proceeding with the analysis of variance, the Levine
test of equality of variances was requested for the two variables to be
contrasted. In relation to the independent variable Gender and the variable
time of duration, the statistic was 14,565 with significance 0.000. In turn,
Table 1: ANOVA.
Tests of
Normality |
||||||
Kolmogorov-Smirnov |
Shapiro-wilk |
|||||
Statistic |
df |
sig. |
Statistic |
df |
sig. |
|
Duration time |
0.162 |
1032 |
0 |
0.784 |
1032 |
0 |
Pupil Diameter |
0.048 |
1032 |
0 |
0.982 |
1032 |
0 |
Table 1 shows the results of the ANOVA tests for the independent variables gender and AOIs. The gender variable affects the two contrasted variables (Time of duration of fixation and pupil diameter), with level p <0.05, however, for the independent variable AOIs, it only affects the differences in the variances of the variable “Pupil diameter”. Additionally, the Games-Howell test (Partial eta squared column of table number one), of non-equality of variances being very low in the score, only allows us to clarify that the independent variable gender affects the two independent variables more strongly than the variable AOIs. Having determined that the gender variable affects the behaviour of the pupil diameter and the areas of interest (AOIs), and as a descriptive complement to the ANOVA analysis, the correspondence analysis is presented in order to identify the closest associations between the categories. Of the AOI variable and the categories of the variable "Pupil diameter". Given that the variable "Pupil diameter" has the characteristics of a continuous variable, it was transformed into an ordinal variable, converting these values to "z" values with mean 0 and deviation 1. Values greater than one deviation were named as "Dilated Pupil", values between 0 and 1, were assigned the category "Normal pupil" and values lower than 0, it was called with the category "Pupil in contraction". Through this method a new ordinal variable with three categories was created. For the correspondence analysis procedure, SPSS 22 was used. Assigning the independent variable "AOIs" in the column categories and the new variable "Pupil Dilation" in the row categories. Table 2 shows the results of the correspondence analysis, observing that the dimension number one represents 95.5% of the total variation and the dimension number two represents only 4.5% (Table 2).
Table 2: Dimensions of the correspondence analysis.
Summary |
||||||||
Dimension |
Singular value |
Inertia |
Chi square |
Sig. |
proportion of Inertia |
Confidence singular value |
||
Accounted for |
Cummulative |
Standard deviation |
Correlation 2 |
|||||
1 |
0.158 |
0.025 |
0.995 |
0.995 |
0.03 |
-0.028 |
||
2 |
0.34 |
0.001 |
0.045 |
1 |
0.03 |
|||
Total |
0.026 |
27.052 |
0.001a |
1 |
1 |
|||
a. 8 degrees
of freedom |
In table number three, we can see the relative contributions of each of the categories of the column and row variables to the two dimensions that represent the associations or correspondences between them. Dimension one that collects or represents 95.5% of the total variation, is loaded first by the category “Dilated Pupil” followed by the category “Contracted Pupil”. In the categories of the variable AOIs, attributes three and two, respectively, contribute the most to this dimension, followed by attribute one and the product brand. For dimension number two, which represents 4.5% of the total variance, it corresponds to the category "Normal pupil" and the logo of the pharmaceutical laboratory (Table 3).
Table 3: Relative Contributions.
Variable |
Code |
D1 |
D2 |
Pupil |
Contraction |
93.62611 |
6.37359 |
Diameter |
Normal |
11.5468 |
88.45359 |
Dilated |
99.7319 |
0.26587 |
|
Areas of Interest AOIs |
Attribute 1 |
89.5288 |
10.4489 |
Attribute 2 |
96.88145 |
3.11754 |
|
Attribute 3 |
99.3691 |
0.63136 |
|
Brand |
91.38765 |
8.61399 |
|
Logo |
63.34464 |
36.65255 |
Observing the correspondence map of figure number six and taking into account the relative correspondence of dimension one (D1), we see the closeness of attribute three and the brand of the product, with the "Dilated Pupil" and attribute two moves away of attribute three and without correspondence with the size of the pupil. This could mean that the brand and attributes featured in focus area number three generated more attention than attributes one and two, along with the pharmaceutical company logo (Figure 6).
Figure 6: Correspondences map.
Next, the analysis of the verbal responses to the questions asked after watching the video is described, it is convenient to remember that asking a person to answer some questions together with the evaluation of each of the response items, is a process that activates system two or the rational brain proposed by Kahneman and Tversky. Here, the importance of comparing or descriptively comparing the responses obtained in the eye-tracking tests is highlighted against the verbal responses that are presented below. Table number four collects the answers to the question about the general opinion about the video, the data represented includes the mean and the coefficient of variation, obtaining the highest value for the category "Credible" and the one with the lowest coefficient of variation (Table 4).
Table 4: Descriptive video analysis.
|
Mean |
Coefficient of
Variation
|
Shocking |
5.267
|
24.3 |
Differentiated |
5.333
|
25.2 |
New |
5.6
|
28.5 |
Credible |
6.2
|
13.9 |
In table number five, the averages and coefficients of variation of the responses obtained two sentences that collected the attributes represented within the areas of interest are presented. Being able to establish the interesting coincidence between the verbal responses and the fixations with their respective diameters of the pupil. The highest average is collected by one of the attributes located in area of interest number one, followed by attribute number one "Accelerates the healing process” (Table 5).
Table 5: Attribute statistics.
Frase |
Mean |
Coefficient of Variation |
It is
effective in managing wounds and skin ulcers. |
5.993 |
23.4 |
It has
bacteriostatic and bactericidal activity, protecting the affected area from
possible infection. |
6.8 |
6.1 |
Reduces wound evolution time. |
5.867 |
30.8 |
Accelerates the healing process. |
6.257 |
19.5 |
Provides a
greater granulation fabric leaving skin with similar appearance, color and
elasticity. |
6.267 |
22.1 |
Finally, in table number
six the doctors decided whether they would prescribe the product or not. 60%
say they know the product, but had not prescribed it and would start to
prescribe it.
Through the work
carried out in this article whose main purpose was to determine if the analysis
of pupillometry (Measurement of the pupil diameter) is applicable to establish
if the decision made by individuals obeys an emotional or rational process.
Throughout the present investigation, the reasons why different authors have used
the measurement of the pupil diameter (Pupillometry) in various areas of
medical and psychological research were exposed, and its importance in
establishing emotional reactions without the mediation of verbal responses. In
this research, eye-tracking technology was methodologically combined with CAWI
(Computer Assisted Web Interviewing) technology, which was in charge of
capturing the verbal responses of the study participants. I was able to
establish that the diameter of the pupil acted as an emotional indicator before
the presentation of a promotional video of a cream for the healing and
treatment of skin wounds. Additionally, it was determined that gender is an
explanatory variable for the difference in pupil diameters and that gaze travel
and fixations in certain areas of interest, allow us to discover two important
conclusions, the first being that the count of fixations along with their
duration, are indicators of attention generated in that particular area, the
second is related to the importance of the correct measurement of the pupil
diameter that allows to establish if that fixation that produced attention, is
due to an emotional response of acceptance or rejection. With which it was also
possible to contrast in this study that the verbal responses of the evaluated
medical doctors did coincide with the findings of the pupillometry. At this
point, it is of special interest that in future research the Human Computer
Interface (HCI) tools can be combined with the traditional survey-type
instruments. This combination allows enriching study findings with a broad
application landscape. What is important is creativity in methodological
designs for consumer research as opposed to applying a single data collection
technique. With advances in computer storage and processing speed, such
creativity and innovation is very achievable.