Machine Behaviorism: Exploring the Behavioral Dynamics of Large Language Models in Decision Making

Date

2025-01-07

Contributor

Advisor

Department

Instructor

Depositor

Speaker

Researcher

Consultant

Interviewer

Narrator

Transcriber

Annotator

Journal Title

Journal ISSN

Volume Title

Publisher

Volume

Number/Issue

Starting Page

4196

Ending Page

Alternative Title

Abstract

Large Language Models (LLMs) have empowered AI agents to autonomously make decisions on behalf of humans, exhibiting humanoid behavior. However, the intricate dynamics driving such behavior remain largely unknown, making it crucial to understand and guide the behavior of these LLM agents, especially in high-stakes decision environments. This paper proposes a new concept of Machine Behaviorism, unlocking the potential of LLM-based agents by providing a systematic framework to reshape the agent behavior. By tracking these agents’ actions under different decision environments, we identify AI behavioral biases when delegated with high-stakes decisions, investigate the origins of such behaviors, and propose corrective interventions accordingly. Specially, using a typical representational behavioral bias commonly found in human investors, we provide a detailed demonstration of machine behaviorism in the financial trading environment. Finally, we advocate for the establishment of machine behaviorism as an emerging discipline, as AI agents increasingly undertake autonomous decision-making and become more integrated into our daily lives.

Description

Keywords

Economic and Societal Impacts of Technology, Data, and Algorithms, ai agents, human-like bias, large language models, machine behaviorism

Citation

Extent

10

Format

Geographic Location

Time Period

Related To

Proceedings of the 58th Hawaii International Conference on System Sciences

Related To (URI)

Table of Contents

Rights

Attribution-NonCommercial-NoDerivatives 4.0 International

Rights Holder

Local Contexts

Email libraryada-l@lists.hawaii.edu if you need this content in ADA-compliant format.