Embedded Files
EXPLORATION-EXPLOITATION TRADE-OFF
Exploration vs. exploitation in a novel complex card sorting task
The Complex Card Matching Task (CCMT) is a novel paradigm I developed to study the exploration-exploitation tradeoff in relation to the locus coeruleus-norepinephrine (LC-NE) system. Participants learn classification rules through trial-and-error feedback, switching between exploring new rules and exploiting the one that currently applies, with two versions of the task manipulating difficulty through the number of relevant card features. Combined with tonic and phasic pupil recordings, the CCMT provides a window onto how LC-NE activity supports adaptive behavior under uncertainty.
Across both difficulty levels, exploration was associated with larger pupil size than exploitation, consistent with heightened LC-NE activity during uncertain decision-making. Task difficulty selectively modulated task-evoked, but not pretrial, pupil responses, suggesting that the pupil dynamics underlying the exploration-exploitation tradeoff are more multifaceted than any single theoretical account currently captures.
The CCMT is available for research use on GitHub (both a behavioral and pupillometry versions): https://github.com/giovannacdelsordo/Complex-Card-Matching-Task.git
Check out the manuscript here (published in Cognitive, Affective, & Behavioral Neuroscience).
How does task demand influence the Exploration-Exploitation tradeoff?
Psychonomics_2025_poster.pdfIn November 2025, I presented a replication and extension of my first Exploration–Exploitation study at the Psychonomic Society’s 66th Annual Meeting in Denver, CO.
This work replicates my earlier findings on how the locus coeruleus–norepinephrine (LC–NE) system supports shifts between exploration and exploitation and extends the investigation to pupil responses during feedback.
Results show that exploration elicits stronger LC–NE activation across all task phases, reflecting heightened arousal and information-seeking under uncertainty, whereas exploitation engages a more stable tonic state that supports efficient, rule-based performance. Although task difficulty modestly increases arousal during exploration, LC–NE dynamics appear primarily state-driven rather than load-driven
Page updated
Google Sites
Report abuse