How memory can affect collective and cooperative behaviors in an LLM-Based Social Particle Swarm

Taisei Hishiki; Takaya Arita; Reiji Suzuki

cs.AIApr 14, 2026

How memory can affect collective and cooperative behaviors in an LLM-Based Social Particle Swarm

Authors: Taisei Hishiki, Takaya Arita, Reiji Suzuki

Organizations: Graduate School of Informatics, Nagoya University, Nagoya, Japan

Abstract

This study examines how memory shapes the collective and cooperative dynamics of Large Language Model (LLM) agents in a multi-agent system. To this end, we extend the Social Particle Swarm (SPS) model, in which agents move in a two-dimensional space and play the Prisoner's Dilemma with neighboring agents, by replacing its rule-based agents with LLM agents endowed with Big Five personality scores and varying memory lengths. Using Gemini 2.0 Flash, we find that memory length is a critical parameter governing collective behavior: even a minimal memory drastically suppressed cooperation, transitioning the system from stable cooperative clusters through cyclical formation and collapse of clusters to a state of scattered defection as memory length increased. Big Five personality traits correlated with agent behaviors in partial agreement with findings from experiments with human participants, supporting the validity of the model. This effect of memory appeared whether or not personality was assigned. With heterogeneous personalities, individual behavior reflected the assigned traits and cooperation collapsed under long memory, whereas without personality Gemini's cooperative disposition dominated and cooperation was broadly maintained. Sentiment analysis of agents' reasoning texts showed that the model interprets memory increasingly negatively as its length grows, already in the early phase, providing a micro-level account of the suppression of cooperation. These results suggest that how an LLM interprets accumulated memory is a key driver of emergent social behavior in Generative Agent-Based Modeling.

Figures & tables

Figure 1: Three classes of the SPS model. (a) Class A: scattered and wandering defectors; (b) Class B: multiple and stable cooperative clusters; (c) Class C: repeated occurrences of emergence and collapse of cooperative clusters. Blue circles represent cooperating agents; red circles represent defecting agents.

Figure 2: Schematic overview of the LLM agent’s decision-making process. The left panel shows the overall structure of the prompt template; the individual components are described in detail in the text. The right panel shows a concrete example for agent $i$ at a specific time step: the top-right box presents the input data (current state, personality traits, interaction history, and neighborhood), the lower diagram visually represents the spatial configuration, and the output box shows the agent’s final decision (strategy, movement, and reasoning).

	Number of Neighbors		Cooperation Rate
$L_{m}$	Mean	Volatility	Mean	Volatility
0	17.6	6.41	0.899	0.0454
1	3.75	1.80	0.260	0.108
2	2.65	1.38	0.139	0.102
3	2.48	0.387	0.0776	0.0462

Table 1: Mean and volatility of the number of neighbors and cooperation rate for each

L_{m}

condition (Gemini 2.0 Flash, with personality). Volatility is the average of standard deviations calculated from the time series of each metric within a single trial, averaged over 10 trials.

Figure 3: Agent spatial configurations for $L_{m}=0,1,2$ , and $3$ (Gemini 2.0 Flash). Each row shows snapshots at $t=10,30,50,100$ , and $500$ for the corresponding $L_{m}$ condition. Blue circles represent cooperating agents; red circles represent defecting agents. (a) $L_{m}=0$ : cooperative clusters form and grow (Class B-like); (b) $L_{m}=1$ : cyclical formation and collapse of clusters (Class C-like); (c) $L_{m}=2$ : gradual transition toward defection; (d) $L_{m}=3$ : rapid convergence to scattered defection (Class A-like).

Figure 4: Pearson correlation coefficients between Big Five personality traits and agent behavioral characteristics ( $L_{m}=1$ condition, Gemini 2.0 Flash). Each cell shows the mean Pearson $r$ value averaged over 10 trials; cells marked “-” indicate non-significant correlations ( $p\geq 0.05$ ). Warmer colors indicate positive correlations; cooler colors indicate negative correlations.

	Number of Neighbors		Cooperation Rate
$L_{m}$	Mean	Volatility	Mean	Volatility
0	15.8	8.27	0.962	0.0292
1	7.12	1.24	0.834	0.0717
2	7.58	1.46	0.750	0.0821
3	7.12	0.996	0.509	0.0963

Table 2: Mean and volatility of the number of neighbors and cooperation rate for each

L_{m}

condition without personality assignment (Gemini 2.0 Flash, “Unspecified Personality” condition). Volatility is defined as in Table 1 , averaged over 3 independent trials.

$L_{m}$	Full ( $t=0$ – $500$ )	Early ( $t\leq 30$ )
0	$+0.60$	$+0.49$
1	$+0.06$	$+0.24$
2	$-0.02$	$+0.21$
3	$-0.24$	$-0.09$

Table 3: Average sentiment scores of memory-related sentences in Gemini 2.0 Flash agents’ reasoning texts. Mean sentiment score (

-1

: negative,

+1

: positive) of memory-related reasoning sentences, per memory length

L_{m}

, for the full experiment (

t=0

–

500

) and the early phase (

t\leq 30

). In both cases the score is positive at

L_{m}=0

and declines into negative territory at

L_{m}=3

. The consistency of this pattern in the early phase indicates that the increasingly negative memory interpretation is an intrinsic property of the model and not an artifact of the converged social state.

Large Language Model-Based Social Simulation

Large language model-based social simulation studies how language-model agents represent people and interact to produce social behavior.

See all 257 papers

Jun 29, 2026

Collective cooperation without individual fidelity in LLM agents

Large language models (LLMs) are increasingly used as agents in simulations of social systems, yet it remains unclear when their behavior can be interpreted as a faithful proxy for human decision-making. Here we test LLM agents against a direct empirical benchmark: a large-scale networked Prisoner's Dilemma experiment with human participants. Using the same interaction protocol, payoff structure, and network topologies, we compare nine open-weight LLMs with the human data. The selected model reproduces several macro-level features of cooperation dynamics, including the early decline and later stabilization of cooperation. This aggregate agreement, however, does not extend uniformly to finer levels of behavior. LLM populations underestimate individual-level heterogeneity and generate conditional cooperation patterns that differ from those observed in humans. Adding a fraction of random agents improves some aspects of micro-level agreement, but does not remove the mismatch in decision rules. These findings reveal a macro--micro dissociation in LLM-based social agents: collective outcomes can appear human-like even when the underlying behavioral distributions and mechanisms are not. They suggest that validating LLM agents as human surrogates requires comparisons across aggregate dynamics, individual heterogeneity, and context-dependent decision rules, rather than outcome-level agreement alone.

Henrique Ferraz de Arruda, Carlos Gracia Lázaro, Alberto Aleta +1

ARAID Foundation, Zaragoza, Spain. · Institute for Biocomputation and Physics of Complex Systems, University of Zaragoza, Zaragoza, Spain. · Universidad San Jorge (USJ), Zaragoza, Spain. +1

Multi-Agent LLM Systems Human Behavior Simulation LLM Agent Evaluation Large Language Model-Based Social Simulation

Ask PDF

May 8, 2026

The Memory Curse: How Expanded Recall Erodes Cooperative Intent in LLM Agents

Context window expansion is often treated as a straightforward capability upgrade for LLMs, but we find it systematically fails in multi-agent social dilemmas. Across 7 LLMs and 4 games over 500 rounds, expanding accessible history degrades cooperation in 18 of 28 model--game settings, a pattern we term the memory curse. We isolate the underlying mechanism through three analyses. First, lexical analysis of 378,000 reasoning traces associates this breakdown with eroding forward-looking intent rather than rising paranoia. We validate this using targeted fine-tuning as a cognitive probe: a LoRA adapter trained exclusively on forward-looking traces mitigates the decay and transfers zero-shot to distinct games. Second, memory sanitization holds prompt length fixed while replacing visible history with synthetic cooperative records, which restores cooperation substantially, proving the trigger is memory content, not length alone. Finally, ablating explicit Chain-of-Thought reasoning often reduces the collapse, showing that deliberation paradoxically amplifies the memory curse. Together, these results recast memory as an active determinant of multi-agent behavior: longer recall can either destabilize or support cooperation depending on the reasoning patterns it elicits.

Jiayuan Liu, Tianqin Li, Shiyi Du +7

Carnegie Mellon University · Foundations of Cooperative AI Lab (FOCAL) · University of Michigan +1

Social Dilemmas LLM Agent Memory Multi-Agent Collaboration Multi-Agent Reasoning

Ask PDF

Oct 23, 2026

The Butterfly Defects: Tracing Cooperation in LLM Agent Networks

As large language models (LLMs) increasingly operate as interconnected agents, a fundamental question emerges: does individual cooperation translate into collective cooperation? We investigate this question through the Iterated Prisoner's Dilemma (IPD), moving from dyadic interactions to homogeneous and heterogeneous networks of open-weight LLMs. Our findings reveal that cooperation is highly sensitive to model composition and seemingly minor linguistic changes. Models that cooperate with identical peers can shift toward defection in mixed populations. A simple uncertainty cue can increase or decrease cooperation, redistribute costs, and affect unexposed agents several hops away. Even reversing action-label order can dramatically alter behavior, with uncertainty framing sometimes amplifying this sensitivity. Together, these findings challenge the notion of cooperation as a stable model-level property, revealing how strategic behavior depends on interaction partners, network positions, and the language used to frame decisions.

Maria Sampani, Maria Lymperaiou, Angeliki Dimitriou +3

National Technical University of Athens

Multi-LLM Collaboration Multi-Agent Coordination Prompt Sensitivity

Ask PDF

Abstract

Large Language Model-Based Social Simulation

Explore similar work

Collective cooperation without individual fidelity in LLM agents

The Memory Curse: How Expanded Recall Erodes Cooperative Intent in LLM Agents

The Butterfly Defects: Tracing Cooperation in LLM Agent Networks