Information-Seeking, Curiosity,
& Random Exploration
PhD Dissertation
Experiment 1 - Non-Instrumental Information (NII)
Research Question: How do people sample non-instrumental information (NII)? Definition of NII: Counterfactual information—e.g., the amount behind the unselected door. Inspiration: Kobayashi et al. (2019); FitzGibbon et al. (2019); Gershman (2019); Daw et al., (2006) Summary: Participants selected one of two doors to contribute to a cumulative point total, which determined their monetary prize. After each selection, they were offered the option to view the outcome behind the unchosen door. Key Design Feature: Participants were explicitly told that door outcomes were random and unrelated to future trials. Thus, the counterfactual information had no instrumental value for updating decision strategies. Key Findings: Participants frequently sampled the non-instrumental information early in the task, but this behavior declined sharply across blocks. Interim Interpretation: Participants may have used NII to test the environment’s reward structure—initially treating counterfactual feedback as a probe for randomness. Once convinced of the system’s stochastic nature, their sampling behavior diminished.
Experiment 2 - Procedural and Epistemic NII
Research Question: How do people sample epistemic and procedural non-instrumental information (NII)? Definition of NII: Procedural: Counterfactual outcome (same as Experiment 1) Epistemic: Answer to a trivia question previously rated by the participant during a pre-task phase Inspiration: Litman et al. (2005); Barnes et al. (2015); Berlyne (1960); Kang et al. (2009); Feng et al. (2016) Summary: Participants selected one of two doors to contribute to a cumulative point total, which determined their monetary prize. After each selection, they were offered non-instrumental information: either the outcome behind the unchosen door (procedural condition) or the answer to a previously rated trivia question (epistemic condition). Key Design Feature: In the second half of the experiment, half of the participants experienced a reward distribution shift—door outcomes were no longer equally likely to yield higher rewards. This allowed for testing whether NII sampling was sensitive to environmental change. Key Findings: The procedural condition replicated findings from Experiment 1: sampling declined over time. In the epistemic condition, NII selection was better predicted by participants’ confidence in knowing the trivia answer. Interim Interpretation: Information value emerges from multiple sources of uncertainty—not just those tied to instrumental goals. Participants may pursue information to resolve curiosity or test environmental structure, even when doing so does not directly improve task performance. Concurrent motivations (e.g., epistemic curiosity) can override point-maximizing strategies.
Experiment 3 - NII Reward Processing (EEG)
Research Question: Are there neural signatures of reward processing for epistemic and procedural non-instrumental information (NII)? Inspiration: Cockburn et al. (2022); Metcalfe et al. (2021); Ruterbories et al. (2024); Mordirshanechi et al. (2023) Summary: This ongoing within-subjects study interleaves blocks of epistemic and procedural NII conditions while participants undergo EEG monitoring. The design aims to identify whether distinct neural correlates emerge for different types of non-instrumental information. Pre-registration: Available upon request—please email me.