Play It Safe or Take a Risk? Computational Modeling & Statistical Inference for the Effect of Emotional Valence on Risk-Taking
Huang, Jessica Li
MetadataShow full item record
CitationHuang, Jessica Li. 2020. Play It Safe or Take a Risk? Computational Modeling & Statistical Inference for the Effect of Emotional Valence on Risk-Taking. Bachelor's thesis, Harvard College.
AbstractDecision making is an important part of life. In the past 50 years, there has been a steady rise in the number and percentage of decision-making papers which also investigate the role of emotions. In this thesis, we are interested in whether the pleasantness of an emotion, or emotional valence, affects risk-taking decisions. This research is important for two main reasons: to better understand a component of the psychological phenomenon of decision making and to inform treatments and interventions for disorders which are symptomatic of altered emotion and decision making (e.g. depression, gambling addiction).
In the literature review, we present comprehensive scientific frameworks for explaining risk-taking and emotional valence during a risk-taking task. We also examine three hypotheses about how emotional valence affects gambling: the mood-maintenance hypothesis, the affect-infusion model, and the reward processing hypothesis. We also explore hurdles in incorporating these studies into mathematical models, which motivate our computational modeling and statistical analysis.
In our risk-taking task and data chapters, we explain why we chose the set of participants and the experimental setup, and we carefully profile participant behavior during the risk-taking task, to incorporate into our behavioral models.
In our theoretical developments and exploratory data analysis, we validate our data, we select covariates, we construct models of risk-taking which are scientifically informed and which demonstrate promise to learn three scientific hypotheses about how emotional valence affects risk-taking (mood-maintenance hypothesis, affect infusion model, reward processing hypothesis), and we test the stability of our models across regularization and resampling.
In our hypothesis test chapter, we build a scientifically informed statistical hypothesis test to begin to answer: does emotional valence affect risk-taking? This hypothesis test relies on conditional randomizations to generate empirical null distributions which boosts power relative to using estimated null distributions. This hypothesis test has advantages of comprehensively accounting for covariates which might affect risk-taking and integrating scientific hypotheses (the mood-maintenance hypothesis, the affect-infusion model, and the reward processing hypothesis) about how emotional valence affects risk-taking. One disadvantage of our hypothesis test is its large computational cost.
Finally, in our conclusion, we propose future computational, statistical, and psychological work to answer our research question, with respect to solving prevailing challenges about computational cost, statistical power, bidirectional relationships, going beyond psychological tasks and towards real-world utility, and modeling the effects of additional components of emotion beyond valence.
Citable link to this pagehttps://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37364742
- FAS Theses and Dissertations