This summer, thanks to the Undergraduate Experiential Learning Opportunities Support Fund, I worked on 鈥淚nferring cognitive variables from behavior鈥� under Dr. Paul Masset at 91丝瓜视频, analyzing data from da Silva Castanheira et al. (2021) on how subjective confidence forms in uncertain, value-based decisions.
The study found higher confidence for certain choices and that risk attenuated links between confidence, response time, and subjective value difference. Further analysis by a lab member revealed that participants often expressed high confidence when choosing high-probability risky options (prisky = 0.9), even when those options were less valuable, suggesting overconfidence and overvaluation. We tested whether this pattern could be explained by side bias (left/right) or certainty bias (preference for certain over risky options). We found that risky options with prisky = 0.9 modulated preferences toward certainty: participants more often chose the certain option when the alternative was a high-probability risky option. The original study used Altered Prospect Theory with a probability-weighting parameter to capture risk preferences. We hypothesized that the discrepancy came from how subjective values were computed. Comparing raw value differences (value risky 鈥� value certain) to expected values (which include probabilities) improved model fit, as shown by higher total log-likelihood. Adding a utility parameter (伪) further improved the fit, successfully replicating Altered Prospect Theory. Next, we fit subjective probabilities for risky options, yielding our best-fitting model. This indicated that participants relied on personal subjective probabilities across low- and high-probability gambles rather than objective probabilities, shaping choices in ways the original model did not predict. To further explore the mechanism, we plotted the probability of gambling as a function of prior decision and subjective valuation. Participants were more likely to choose an objectively less valuable gamble if they had previously gambled, with this effect strongest when the prior gamble had also been objectively less valuable. These findings suggest gambling behavior may be history-dependent. Cross-correlation analyses revealed a strong negative relationship between 伪 and subjective probability, indicating a potential trade-off between the parameters. We are now beginning to test alternative models, such as Dynamic Prospect Theory, which may better capture how humans make value-based decisions under risk, including variability across high- and low-probability gambles.
I was drawn to the ARIA project because of the structure and support it provides for students. As I prepare to graduate from my bachelor鈥檚 degree, I am considering a master鈥檚 in research. Before applying this fall, I wanted to be sure of my path, and ARIA offered the ideal opportunity to explore this decision without the added burden of working multiple jobs. I began my summer research with the goal of stepping out of my comfort zone and further developing my research skills. I did so by drafting an article summarizing my work, collaborating with lab members who had contributed to the project before me, and practicing communicating my research for the upcoming Annual Faculty of Arts Undergraduate Research Event. All of this was possible thanks to ARIA.
This summer was filled with highlights: finally getting my code to work, seeing the pieces of the puzzle come together, and enjoying lab lunches outside Stewart Bio. I had a rewarding experience that deepened my knowledge in computational modeling of decision-making. There were challenges, especially as a Psychology major navigating models, machine learning, and coding. Yet with the support of my supervisor, Dr. Masset, and my lab members, I overcame them by asking questions, doing extra research, and engaging with papers that sometimes pushed me beyond what I fully understood.
ARIA has genuinely helped shape my future educational path. It confirmed my desire to pursue research at the master鈥檚 level and gave me the confidence to work outside my comfort zone. Most importantly, none of this would have been possible without the generous funding from my donor. Because of this support, I covered my living costs and fully immersed myself in research, rather than dividing my time between multiple jobs. This made my experience far richer and more impactful, and I accomplished more as a result. To my donor, I extend my most sincere thanks. Your support allowed me not only to learn and grow this summer but also to feel more confident in my own 鈥渞isky decision鈥� to pursue graduate studies.
