Research on Time Discounting and Magnitude Effect in Intertemporal Decision-Making from the Perspective of Dual-System Theory
Download as PDF
DOI: 10.25236/ieesasm.2024.053
Corresponding Author
Kun Zhang
Abstract
Intertemporal decision-making, which involves choosing between immediate and delayed outcomes, is shaped by time discounting and the magnitude effect. Traditional models, such as exponential and hyperbolic discounting, often fail to capture the dynamic inconsistencies and reward size effects seen in human behavior. To address these gaps, this paper introduces the Dual-System Reinforcement Learning (DSRL) algorithm, which integrates dual-system theory with reinforcement learning. The DSRL model dynamically adjusts the influence of impulsive and rational decision-making systems based on reward magnitude and time delay. By incorporating a hybrid discounting mechanism, the DSRL algorithm better models how individuals weigh short-term versus long-term rewards. Experimental results demonstrate that DSRL surpasses baseline models in handling time discounting and adapting to different reward magnitudes, achieving higher cumulative rewards over time. Additionally, the DSRL model learns more efficiently, converging faster to optimal decision strategies. These findings suggest that DSRL provides a more accurate and adaptable framework for understanding human decision-making in complex, delayed-reward environments.
Keywords
intertemporal decision-making, time discounting, dual-system theory, reinforcement learning