“What is the significance of feedback and reinforcement learning mechanisms in aiding dynamic decision making?” – proposing a revised hybrid reinforcement learning initiated instance-based model of dynamic decision making

Abstract

Aim: The current study aimed to propose a new hybrid-model called Reinforcement-Learning initiated Instance-based Model which acknowledges the significance of feedback in promoting learning and memory-mechanisms over a period of time to aid Dynamic Decision-Making (DDM), a crucial aspect disregarded by a previous DDM-model called Instance-based Learning Model proposed by Gonzalez et.al.,(2003) as also test this hybrid-model in a novel Financial Expenditure DDM-environment with varying income-periods like income/employment and retirement-phase. Method: In a mixed-measures experimental-design the current study subjected participants (n=48) to a financial DDM-task wherein they undertook discretionary-expenditures during varying income-periods (income-phase/retirement-phase), across two conditions—feedback and no-feedback in two separate decision-making sessions (DMS) 1&2 in each condition. Results: Results indicated that participants in feedback-condition (DMS-1&2) outperformed in the financial DDM-task as compared to those in no-feedback (DMS-1&2) condition. Furthermore, participants in feedback-condition in DMS-1 performed significantly better in DMS-2 without explicit-feedback, while those in no-feedback condition in DMS-1 deteriorated in DMS-2 in DDM-task without feedback, suggesting that feedback facilitated development of memory-instances during the learning-process in DMS-1 which aided DDM in similar DDM-environment (DMS-2). Conclusion: The current study highlighted the significance of feedback in aiding learning and memory-mechanisms in DDM-settings.