Undergraduate Honors Theses

Thesis Defended

Spring 2018

Document Type

Thesis

Type of Thesis

Departmental Honors

Department

Psychology & Neuroscience

First Advisor

Dr. Yuko Munakata

Second Advisor

Dr. Eliana Colunga

Third Advisor

Dr. Iskra Fileva

Fourth Advisor

Jesse Niebaum

Abstract

Everyone has to make reward-based decisions. For example, a child may decide to behave well in a grocery store because they know they will get rewarded with ice cream. Strategies for learning from previous outcomes include model-based (uses a model of the environment, such as generalizing the good behavior to the library to get further reward) and model-free (reacts habitually to previous reward). Previous research used a two-step decision-making task to distinguish between these two strategies, which consisted of two decision stages. Adolescents and adults used both model-free and model-based strategies, whereas children only used model-free. However, there was a limitation to this task: the more effortful, model-based strategy did not receive greater reward. We hypothesized that when children were incentivized they would use model-based strategies and that age would be positively correlated with these strategies. A modified two-step task that incentivized model-based strategies was used. We found that children utilized a mixture of both model-based and model-free strategies with no significant correlation between model-based strategies and age. Additionally, in the original task, children demonstrated anticipatory preparation for a second-stage decision that could indicate proactive control tendencies, which is the process of maintaining goal-relevant information in anticipation of needing it. We hypothesized that proactive control would be positively correlated with model-based strategies. To assess this, we used the cued task-switching paradigm, which had participants sort toys based on shape or color in a proactive possible and proactive impossible condition. We found a marginally significant correlation between proactive control and model-based learning. We also included a measure of need for cognition and found a non-significant correlation with model-based learning. Our study was the first to find that children are capable of utilizing model-based strategies.

Share

COinS