What question did this study set out to answer?

This research investigates how different memory types affect decision-making quality in AI agents within a simulated economic environment.

May 19, 2026Open Access

Memory Presence Matters, Mechanism Does Not: Evidence from a 21-Agent Organizational Simulation on a Historical Economic Benchmark

Key Points

This research investigates how different memory types affect decision-making quality in AI agents within a simulated economic environment.
Used a multi-agent simulation framework with 21 AI agents for decision-making.
Compared three memory conditions: bio-inspired memory, flat retrieval memory, and no memory.
Conducted experiments using historical economic data spanning from 1925 to 2024 with a sample size of n=78 per arm.
Bio-inspired memory (d=5.30, p<1e-58) and flat retrieval memory (d=4.94, p<1e-58) substantially outperform no memory.
Both memory types are statistically indistinguishable from each other (d=0.14, p=0.390).
In a larger paired analysis, a strong performance is confirmed (d=6.28, p=2.10e-195) under crisis conditions.

Abstract

We introduce YMERA, a multi-agent simulation framework in which 21 AI executive agents deliberate over strategic, operational, financial, and risk decisions using a historical economic data surface spanning 1925-2024. In the current benchmarked experiments, we evaluate the 1925-1934 decade and compare three memory conditions: bio-inspired memory, flat retrieval memory, and no memory. In the canonical three-condition run (n=78 per arm), bio-memory and flat retrieval each substantially outperform no memory (d=5.30 and d=4.94, pflat signal for CEO+CHRO agents in crisis years after agent-year normalization (d=1.03, Welch p=0.022). We conclude that memory presence strongly improves organizational AI decision quality, while bio-inspired mechanism complexity yields no broad advantage over flat retrieval at this model scale.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper