arxiv_ai 80% Match Research Paper AI Researchers,Cognitive Scientists,Neuroscientists,HCI Researchers 19 hours ago

Modulation of temporal decision-making in a deep reinforcement learning agent under the dual-task paradigm

reinforcement-learning › multi-agent

📄 Abstract

Abstract: This study explores the interference in temporal processing within a dual-task paradigm from an artificial intelligence (AI) perspective. In this context, the dual-task setup is implemented as a simplified version of the Overcooked environment with two variations, single task (T) and dual task (T+N). Both variations involve an embedded time production task, but the dual task (T+N) additionally involves a concurrent number comparison task. Two deep reinforcement learning (DRL) agents were separately trained for each of these tasks. These agents exhibited emergent behavior consistent with human timing research. Specifically, the dual task (T+N) agent exhibited significant overproduction of time relative to its single task (T) counterpart. This result was consistent across four target durations. Preliminary analysis of neural dynamics in the agents' LSTM layers did not reveal any clear evidence of a dedicated or intrinsic timer. Hence, further investigation is needed to better understand the underlying time-keeping mechanisms of the agents and to provide insights into the observed behavioral patterns. This study is a small step towards exploring parallels between emergent DRL behavior and behavior observed in biological systems in order to facilitate a better understanding of both.

Key Contributions

Investigates temporal decision-making interference in a dual-task paradigm using DRL agents trained in a simplified Overcooked environment. The study demonstrates that a dual-task agent exhibits significant time overproduction compared to a single-task agent, mirroring human timing research, and explores the underlying neural dynamics.

Business Value

Provides insights into how AI systems might handle multitasking and temporal judgments, which is crucial for developing more human-like AI assistants and understanding potential failure modes in complex operational environments.

Paper Metadata

Innovation Type

Experimental Study/Simulation

Deployment Feasibility

Moderate, as the findings are based on simulations and provide theoretical insights rather than direct deployable solutions.

Limitations Addressed

Lack of understanding of how concurrent tasks affect temporal processing in AI agents and how this relates to human cognitive behavior.

Technical Tags

Deep Reinforcement Learning (DRL)Dual-Task ParadigmTemporal Decision-MakingTime Production TaskNumber Comparison TaskOvercooked EnvironmentLSTMEmergent BehaviorAI TimingInterference

Research Topics

Temporal Processing in AICognitive Load and AIReinforcement Learning AgentsSimulating Human BehaviorAI Decision Making

Methods & Architectures

Deep Reinforcement LearningAgent TrainingComparative AnalysisLSTM Layer Analysis Deep Reinforcement Learning AgentsLSTM (Long Short-Term Memory)

Applications & Tasks

Cognitive Science Simulation AI Behavior Modeling Human-AI Interaction Modeling temporal decision-makingInvestigating interference in dual-task scenariosSimulating human timing behavior Temporal decision-makingTime productionNumber comparison

Related Fields

Cognitive ScienceNeuroscienceArtificial IntelligenceReinforcement LearningHuman-Computer Interaction

Keywords

deep reinforcement learningdual tasktemporal processingdecision makingOvercookedLSTMAI timingcognitive loadhuman behaviorsimulation

Academic Context

#Temporal Processing in AI#Cognitive Load and AI#Reinforcement Learning Agents#Simulating Human Behavior#AI Decision Making

Commercial Potential

Potential Products

AI agents with improved multitasking capabilitiesTools for simulating cognitive processes

Target Industries

GamingRoboticsHuman-AI Interaction DesignCognitive Modeling

Use Case Examples

Developing AI agents that can manage multiple concurrent tasks efficientlySimulating human multitasking performance for training or research

Competitive Edge

Focuses on a specific cognitive aspect (temporal decision-making under dual-task load) within RL, contributing to a deeper understanding of agent behavior.

Resource Requirements

Compute Needs

Requires compute for training DRL agents, potentially significant.

Data Requirements

Requires the Overcooked environment setup.

Deployment Constraints

The findings are primarily theoretical and simulation-based, not directly tied to a specific deployment scenario.

Scalability

The DRL training process can be computationally intensive and may scale with environment complexity and agent size.

Production Readiness

Maturity Level

Research/Experimental

Patent Potential

Low, focused on fundamental research.

View Full Paper Back to Papers