| 研究課題/領域番号 |
24K15688
|
|---|
| 研究種目 |
基盤研究(C)
|
| 配分区分 | 基金 |
|---|
| 応募区分 | 一般 |
|---|
| 審査区分 |
小区分90030:認知科学関連
|
|---|
| 研究機関 | 沖縄科学技術大学院大学 |
|---|
研究代表者 |
SANGATI Ekaterina 沖縄科学技術大学院大学, 神経計算ユニット, ポストドクトラルスカラー (70898246)
|
|---|
| 研究分担者 |
チャン ユーチャン 立教大学, 現代心理学部, 助教 (50831484)
|
|---|
| 研究期間 (年度) |
2024-04-01 – 2025-03-31
|
| 研究課題ステータス |
中途終了 (2024年度)
|
| 配分額 *注記 |
4,550千円 (直接経費: 3,500千円、間接経費: 1,050千円)
2026年度: 1,430千円 (直接経費: 1,100千円、間接経費: 330千円)
2025年度: 1,690千円 (直接経費: 1,300千円、間接経費: 390千円)
2024年度: 1,430千円 (直接経費: 1,100千円、間接経費: 330千円)
|
|---|
| キーワード | collective behavior / collective intelligence / reinforcement learning / Bayesian inference / multi-armed bandit / agent-based simulation / emergence / collective cognition / cognitive cost / social coordination |
|---|
| 研究開始時の研究の概要 |
This project will contribute to understanding the phenomenon of emergent collective cognition: in which sense group behavior can be "emergent" and "cognitive". The first property will be assessed by using novel information-theoretic tools to data recorded in joint perception tasks. The cognitive status will be tested as the ability of collectives to adapt to changing task conditions. This research will inform further theoretical developments in cognitive and behavioral sciences, lead to more refined experiments, and have practical implications for supporting collective cognitive performance.
|
| 研究実績の概要 |
Before grant termination, we established the foundations for investigating collective cognition through agent-based simulations and human studies. We developed a multi-agent multi-armed bandit task, where agents or participants coordinate to maximize collective rewards. Our simulations show that groups of simple learning agents with social mechanisms can perform nearly as well as a centralized, cognitively complex agent, exemplifying distributed cognition. A publication on these results is in preparation. We also implemented an online experiment mirroring the model and conducted internal testing. In the coming months, we plan to collect preliminary human data and fit models to uncover the cognitive mechanisms driving collective behavior.
|
