What question did this study set out to answer?

To evaluate whether autonomous agents powered by large language models can realistically simulate human cognitive processes, decision-making, and social dynamics within recommendation platforms.

December 20, 2024ACM Transactions on Information Systems

User Behavior Simulation with Large Language Model-based Agents

View Full Paper

Ask AI

Bookmark

Authors

LWLei WangRenmin University of China JZJingsen ZhangRenmin University of China HYHao YangUniversity of California, San Francisco

Discussion

Member takes

Implication

Simulation study demonstrates believable user interactions in recommendation systems, indicating potential to analyze social phenomena like information cocoons.

Key Points

To evaluate whether autonomous agents powered by large language models can realistically simulate human cognitive processes, decision-making, and social dynamics within recommendation platforms.
Constructed autonomous software agents equipped with customized profile, memory, and action modules driven by large language models.
Designed an interactive sandbox environment enabling individual agent interaction with recommendation systems alongside one-to-one and broadcast agent communications.
Conducted subjective and objective believability evaluations and executed targeted simulation experiments modeling information cocoons and user conformity.
Agent-generated behavioral patterns demonstrated high believability across both subjective qualitative reviews and objective performance metrics.
Restricting the personalization degree of recommendation algorithms and elevating the diversity of social relations significantly mitigated information cocooning.
User conformity behavior emerged systematically and was strongly mediated by the overall volume of an agent's social network connections.

Cite This Study

Wang et al. (2024) studied this question.

synapsesocial.com/papers/6a7e1191bcaf28b5c937ec2e https://doi.org/10.1145/3708985

View Full Paper

Ask AI

Bookmark

Implication

Simulation study demonstrates believable user interactions in recommendation systems, indicating potential to analyze social phenomena like information cocoons.

User Behavior Simulation with Large Language Model-based Agents

Discussion

Member takes

Implication

Key Points

Cite This Study

User Behavior Simulation with Large Language Model-based Agents

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider