Machine Learning in Psychology (final part)

Hi everyone! After covering the second article last week which was on machine learning being used to identify young offenders, I’m excited to write about the third and final article about the uses of machine learning in cognitive psychology. In today’s blog, I will be writing about how machine learning was used in measuring decision-making based on pupil response in a study done in 2018.

Essentially, the premise of Van Slooten and her team’s research was that your pupils dilate both before and after you make decisions. Your pupils are important indicators about your thinking. Based on that idea, Van slooten and her group investigated how decision making caused pupil size fluctuations and the cognitive processes that were demonstrated through those fluctuations.

The research involved 34 participants and involved a learning task. During the learning task, participants selected between pairs of options on a screen, each with varying probabilities of reward. These probabilities were undisclosed, requiring participants to adapt based on feedback. The task included randomness, so even “better” choices didn’t always yield rewards, mimicking real-life decision-making. During all this, pupil size was continuously tracked with eye-tracking to observe how reinforcement learning and decision-making influenced physiological responses; focusing on pupil dilation and constriction during decision-making and feedback.

The approach that was used during this experiment was one known as the Bayesian Approach. Utilizing the Bayesian approach, researchers were able to model the decision-making process of the participants and statistically predict the results. Quite simply, the Bayesian provides a way to increase the hypothesis accuracy based on newly found data; similar to making an educated guess that becomes more and more right as you continue to learn. You have the initial established probability and as more and more data is acquired, the posterior probability (updated hypothesis) updates. Through their work, the researchers observed that pupil size exhibited a pattern of dilation about one second before a decision was made, suggesting increased cognitive processing or anticipation of the choice. Following the decision, pupil size tended to constrict, particularly around a second after the decision, reflecting the processing of the outcome or feedback.

As I said before, these studies on the applications of machine learning in cognitive psychology will only continue to improve. As indicated by the Bayesian in this experiment, it’s extremely important to acquire as much data as possible because the more data that is acquired, the more accurate the results will be which will continue to improve the researchers in their mission in predicting how the brain functions.