Пошук академічних текстів

Інформація × Реєстраційний номер 2124U004694, Матеріали видань та локальних репозитаріїв Категорія Препринт Назва роботи Entropy-Based Approach to Modeling Reward-Dependent Choice Behavior in a Feature-Specific Reinforcement Learning Task Автор Nahorniuk Marta Дата публікації 01-01-2024 Постачальник інформації Український католицький університет Першоджерело https://hdl.handle.net/20.500.14570/4845 Видання Опис A decision is a central concept of behavioral economics, a detailed understand- ing of which has been a focus of research in various fields, and yet - it is still not fully discovered. The evolution of choice behavior computational models, particu- larly within the reinforcement learning framework, significantly improved the field. However, traditional models often struggle to accurately capture the complexities in decision-making and frequently demonstrate inconsistencies between expected and actual behavior - a phenomenon known as undermatching. This study lever- ages entropy-based method, a model-free approach based on information theory, to address the problem in feature-specific reinforcement learning tasks in which the agent is unaware of the task-reward connections. Unlike previous methodolo- gies that primarily assess general decision-making tendencies, this study applies an entropy-based approach to explore the dynamics of choice behaviors driven by spe- cific features in controlled experimental settings with monkeys. Our findings sug- gest that this entropy-based method not only explains feature-specific choice behav- ior significantly better than previously existing methods but also improves the pre- dictive accuracy of computational models of choice. This approach represents a sig- nificant advance in integrating behavioral insights into computational frameworks across various scientific and technological domains. It also deepens our understand- ing of feature-specific choice behaviors and their impact on decision dynamics. Fi- nally, the study contributes to the broader field of behavioral economics by provid- ing a more precise way to explain the decisions and bridge gaps between theoretical economics and practical, observable behaviors in controlled experiments. Додано в НРАТ 2025-05-09 Закрити