INVESTIGATING VISUAL ATTENTION & COMPREHENSION WITH DRUG LABELS
My Graduate thesis study interested in understanding how information salience and presentation of information has an effect on the way cannabis users interpret and pay attention to potency and warning label information.
MY ROLE
Principal Investigator
THE TEAM
Personal Thesis
TOOLS & METHODS
Eye-tracker, Facial Expression Analysis, Quantitative Research, Qualitative Research, SPSS,
BACKGROUND
But first… why is this important to study?
Cannabis packaging labels are legally required to include a vast amount of information across all US states. Given the potential for negative effects from cannabis consumption, it is vital that appropriate packaging and labeling regulations are developed that will encourage consumers to notice important information pertaining to potency and warnings, and also ensure that they are be able to understand the information the label is trying to convey. Thus, this study was interested in understanding how varying the presentation of information on cannabis labels can impact attention and comprehension.
More specifically, this study aimed to investigate the effects of varying the presentation of the potency of cannabis and the salience of cannabis warning labels (CWLs) which are required by California law to be included on packaging
Methods and Procedures
Recruitment and Participants
Participants were recruited using the San Jose State University Research Pool (SONA).
To determine an appropriate sample size, an a priori sample size computations were conducted using GPower, with an effect size of 0.5, alpha = 0.05, and a power of 0.8. The results indicated a sample size of 27.
Prior to selecting participants, recruited individuals who signed up for the study via SONA were asked to complete a screener survey via text message to determine whether they meet the requirements of the study.
To qualify, individuals had to be:
At least 18 years old
Fluent in English to allow for them to comprehend labels
Able to read text without glasses from a 27 inch monitor while seated 1 foot away from a screen
Active / past cannabis users OR be open to consuming cannabis in the future.
Participants were given 1.0 course credit for their participation.
Participants were asked to sign a consent form upon arriving at the lab for the study. Each participant was emailed an electronic copy of a consent form on Docusign that detailed any risks, benefits and proceedings of the study, and their right to withdraw from the study at any time they wish without any negative consequences.
Apparatus
The stimuli images were presented on an 28” monitor with a screen resolution of 1920 x 1980 pixels and a refresh rate of 60 Hz. Eye tracking responses were measured using a Gazepoint GP3HD V2 Eye tracker. The facial responses were measured using a Cisco Precision HD F1.7/50 Degree View Camera. iMotions eye tracking software was used to present and record the eye-movement.
Phase 1: Eye-tracking & Facial Expression Analysis
Phase 1 of the study was a 3 x 3 factorial within-subjects full factorial design study where each participant experienced 9 conditions. Each condition involved a participant viewing a specific combination of
a) Presentation of Potency information
- (i) MG, (ii) THC %, (iii) Traffic light indication;
b) Presentation of Cannabis Warning Label (CWL)
- (i) Bullet Points (ii) Original %, (iii) Symbol imagery.
The dependent variables (DV’s) in Phase 1 of the eye-tracking study were eye tracking metrics - specifically saccade count, dwell time, and average fixation duration. Additional dependent variables were facial expressions and emotions which were captured by the camera.
Example of sample Cannabis Package labels viewed by participants
Example of sample Cannabis Package labels viewed by participants
Phase 2: Comprehension Accuracy Test
Phase 2 of the study was a 3 x 3 x 2 factorial within-subjects full factorial design study where each participant experienced 18 conditions each. The accuracy of the responses collected from the participants in the quiz section were coded with either a ‘1’ which indicated a correct response or ‘0’ which indicated an incorrect response. The IV were
Cannabis label warning presentation (Original, bullet points, symbols)
Potency presentation (Traffic light, THC %, or MG)
Potency indicated on the label (high or low)
The dependent variables (DV’s) in Phase 1 of the eye-tracking study were eye tracking metrics - specifically saccade count, dwell time, and average fixation duration. Additional dependent variables were facial expressions and emotions which were captured by the camera.
Each image of a label they viewed was followed by these two questions:
“Is this a low, medium, or high dose?” with radio button options of “low”, “medium”, and “high”. 5% THC and 5mg will be meant to signify ‘low’ potency and 25% THC and 25mg will be used to signify ‘high’ potency as used in the study by Leos-Toro et. al (2020) who used Canadian regulations and the products on the market as a guide.
The second question was: “The Cannabis Government Warning Label (the bottom section of the packaging) is easy to understand” with the options “strongly agree”, “agree”, “neutral”, “disagree” and “strongly disagree”.
Analysis
Facial Expression Processing
The following emotions and facial expressions were investigated and were flagged for every instance of any of the emotions being detected while participants were looking at the cannabis warning label: confusion, contempt, engagement, negative, brow furrow, inner brow raise, neutral.
Using iMotions, a threshold cutoff of 50 was set for each emotion, if the value ever exceeded 50 the algorithm classified the presentation of an emotion/facial expression.
Eye Tracking Analysis
Two separate Areas of Interest (AOI’s) were manually defined by using the iMotions tool to draw a rectangle around the desired area. The AOI’s were defined as the Potency AOI and the CWL AOI. The following eye-tracking metrics were compared across each of the 9 different conditions for each of the 29 participants for each of the two AOI’s: saccade count, dwell count, average saccade duration, dwell time (fixation %), dwell time (gaze %). The Eye-tracking metrics were exported via iMotions into an .xls format after which a repeated measures ANOVA was conducted.
Quiz Accuracy
A three way repeated measures ANOVA was performed to compare the effects that different variations of the Potency presentations and the CWL presentations had on participants' accuracy when determining the potency indicated on cannabis labels. The binary responses were coded whereby a ‘1’ indicated that participants had correctly identified the potency and ‘0’ indicated that they were unable to correctly identify the potency.
Ratings on CWL
Participants were shown each of the 18 labels and were shown the following statement: “The Cannabis warning label is easy to understand”. Respondents were given the following radio options to respond with: “Strongly Agree”, “Agree”, “Neutral”, “Disagree”, and “Strongly Disagree”. Responses were coded whereby a ‘2’ corresponded to “Strongly Agree”, ‘1’ corresponded to 'Agree’, 0 corresponded to ‘Neutral’, ‘-1’ corresponded to Disagree, and ‘-2’ corresponded to Strongly Disagree.
A repeated measures ANOVA was performed to compare the effects that different variations of the CWL presentations had on participants' subjective ratings of the CWL.
Results
Subjective data
Verbal Opinions on CWL Presentation
Responses were grouped together based on preference between the symbol representation and the text based representation of the CWL.
13 / 24
PREFERRED TEXT OVER SYMBOL
11/ 24
PREFERRED SYMBOL OVER TEXT
Why Participants Prefer Symbols over Text
Symbols are more visual and easier to see
People won't have to spend time reading text
Symbols feel easier to understand.
Why Participants Prefer Text over Symbols
Symbols can be easily misinterpreted
Novices may not understand what the symbols mean
Symbols feel lazy
Symbols are not not easy to understand.
10 mentioned that they preferred the bullet point form of the text as it felt easier to read compared to the original paragraph
“[Symbols are] easier and faster to comprehend instead of having to read everything out. I think others would agree with me because most of the time we consume cannabis, we don't often look at the warning labels. Having symbols/signs on the warning labels encourages me to actually look at them more.”
Verbal Opinions on Potency Presentation
14 / 21 responses preferred the MG presentation of potency, whereas 5 participants preferred potency to be presented in THC % and the remaining 5 preferred the traffic light indication.
Participants who were explicitly against the Traffic light presentation and % THC presentation felt that they were confusing to understand and felt more comfortable with the MG presentation.
Another idea that appeared with the traffic light indication is the idea that it is a subjective grading system as one person’s low potency could be another person’s high potency.
14 / 21
PREFERRED MG
OVER THC %
(make it look nicer on Figma)
5 / 21
PREFERRED THC %
OVER MG
Why Participants Prefer MG
Symbols are more visual and easier to see
People won't have to spend time reading text
Symbols feel easier to understand.
Why Participants Prefer THC%
Symbols can be easily misinterpreted
Novices may not understand what the symbols mean
Symbols feel lazy
Symbols are not not easy to understand.
10 mentioned that they preferred the bullet point form of the text as it felt easier to read compared to the original paragraph
Discussion
Potency Presentation
The comprehension phase of the study demonstrated that presentation styles can have an effect on the way cannabis users interpret potency.
Participants were more accurate when judging potencies when the potency was shown with a Traffic light presentation compared to when it was shown with the THC %.
This is an important finding as current California requirements ask potencies to be indicated in THC % which was least preferred by participants alongside the symbol representation.
CWL Presentation
Participants preferred the Bullet Point presentation significantly more than both the Original Presentation and the Symbol presentation.
Subjective opinions from participants from the open text response question also pointed towards this finding, whereby 10 participants specifically mentioned the bullet points being easier to read compared to the Original and Symbol presentations.
This is another important finding as current California cannabis label requirements involve displaying the Cannabis Warning Label as a paragraph (the Original condition) which participants felt was difficult to read and made it less likely for them to want to read it.
Making a warning label that includes pertinent information regarding cannabis consumption more salient and easy readable and accessible is vital and is something that policy makers need to prioritize in order to help decrease the negative risks that are associated with cannabis consumption (maybe put this in a box)
Limitations
A limitation in this study is the sample size. Although the sample size was large enough (N = 29) to gain significant results and reach a power of 0.8, a larger sample size may have allowed for better triangulation of results across the multiple subjective and objective measures with more confidence.
All participants were recruited from the SJSU Research Pool (SONA) pool to receive undergraduate Psychology credits, which is not a representative population of the average cannabis user in California.
Future Studies
Future studies should involve using additional objective measures which could help better determine and measure cognitive load and visual attention than eye-tracking data which when alone objectively may not be able to point towards much.
A lot of valuable information can also be obtained from subjective measures as well such as focus groups, interviews, and surveys.
The eye-tracking algorithm used to process facial expressions with iMotions was AFFDEX, however iMotions FACET appears to perform better across all measures according to validation research by Stöckli, Schulte-Mecklenbeck, Borer, & Samson (2017).
More research can also be done on understanding why participants specifically prefer one presentation type over the other. Additionally, studies can also be done in more realistic scenarios where reading cannabis labels would be more likely to improve external validity.