🧠Stress vs. Scores💯

Stress in high-stakes situations like exams is a common experience, yet most studies on the body's stress response rely on artificial lab stimuli rather than real-world scenarios.

A study conducted by Md Rafiul Amin, Dilranjan S. Wickramasuriya, and Rose T. Faghih^[Link] aimed to investigate physiological stress responses in real-world academic settings. Unlike traditional stress studies that rely on artificial stimuli, this research focused on capturing real-time physiological data from students as they underwent midterm and final exams.

Students were asked to wear the Empatica E4 wristband while they take their midterm 1&2 and final exams. The study recorded 6 health features during 3 exams for each student (10 students). We focused on the following 4 features:

• Skin Surface Temperature (TEMP): Measures skin temperature using a built-in temperature sensor in degrees Celsius (°C).
• Electrodermal Activity (EDA): EDA reflects sweat gland activity, which is controlled by the autonomic nervous system. Higher values indicate increased stress.
• Heart Rate (HR): Number of times a heart beats in a minute (BPM). Elevated heart rate is commonly associated with anxiety and cognitive load.
• Blood Volume Pulse (BVP): Captures blood volume changes in the microvascular tissue using a photoplethysmograph (PPG) sensor. If high BVP is accompanied by high heart rate (HR) and increased electrodermal activity (EDA), it may indicate high cognitive stress.

Along with these 4 features, a new feature, 🧠Stress, was created that would track overall student stress felt during an exam.This was created by using the heart rate, blood volume pulse, and skin surface temperature features. Using the individual average student data per min datasets, each feature value was normalized using the global max/min values:

We first wanted to show how the health features changed over the course of an exam. This visualization below shows the trends of given health features for students over the course of any given exam. Try selecting different exams, health variables, and students using the buttons on the left and right and see if you see any patterns/trends!

Interesting Takeaways: The heart rates of students for all tests seems to steadily decrease as an exam goes on. For Midterm 2 and the Final exam, students experienced a slight drop in skin surface temperature, which then increased significantly and stayed high until the end of the exam. In contrast, for Midterm 1, the students didn't experience notable decreases in skin surface temperature, and TEMP only began significantly increased after the testing duration was already halfway over. Blood volume pulse shows little correlation with exam duration, but all tests have momentary spikes for certain students. The same could also be said for electrodermal activity, although there seems to be a slight trend of EDA increasing in the second half of an exam for most students.

We then wanted to show the correlation between average health feature values and how well students performed on a given test. This visualization shows the given average health feature values of students and shows how they relate to how well each student did on a given test. Try selecting different exams and health variables using the buttons on the left and right and see if you see any patterns/trends!

Also, hover over the points on the graph for specific info on the students!

Interesting Takeaways: Surprisingly, in all exams, the more stressed students tended to perform better on average. The same is true for students with higher average heart rate and temperature as well. Meanwhile, blood volume pulse and electrodermal activity seem to have little correlation with how well students perform.

Finally, in the visualization below, users can adjust health feature sliders to set desired physiological values and see the student whose exam data closely matches those values along with their grade. Try to notice any patterns! Which sliders seem to have the most and least impact on how many predictions appear?

Interesting Takeaways: Some sliders seem to have much more impact on the prediction. The visualization and the chosen grade is frequently changing while dragging the Average EDA and STRESS sliders, while the HR and TEMP slider seem to cause fewer changes, and the BVP slider rarely changes the chosen grade. This might be an indicator that some features are more correlated with a student’s grade. The combinations of which sliders are being chosen and the values they are set at also seems to have an impact on the frequency of changes in the prediction.

So what did we learn?

Wow, our research on this dataset has led us to some interesting conclusions! The most important one is the correlation between overall stress levels and grades. Contrary to the general assumption, the students in this study performed better under more stress. The reason behind this phenomenon requires further study. We have some hypotheses as to why this may be the case. Higher tress levels may enhance students’ focus and make them better prepared for the exams. Students may also be more stressed if they prepared a lot for an exam and are worried that their efforts will be wasted, while students who didn't prepare for an exam as hard are less stressed about their studying being wasted. Maybe these students were just outliers and the study needed a larger sample size.

Another thing we found are interesting trends in certain health features as an exam goes on. Skin surface temperature always increased significantly at some point during tests and remained high until the end. In contrast, heart rate steadily decreased for the entire duration of the exams. Some features also didn't have clear trends as time passed, such as BVP, while others had different trends for specific exams, such as EDA, or a bit of both, like STRESS. Of note is that, for EDA and BVP, there were massive spikes in the data for specific students at seemingly-random intervals.

Finally, we noticed that some features have a big impact on the student’s grade and might be good indicators of stress, like EDA and heart rate. In contrast, other features like BVP and temperature are less helpful for indicating stress and should potentially be omitted in future studies on stress analysis.

Hopefully, the next time you feel stressed out and under pressure for an exam, you will remember the results of this project and feel a little bit better about your chances at getting the grade that you want!