認知、心理、與文化因子的突現性複雜現象---整合代理人基計算經濟學、實驗經濟學、與腦神經經濟學之研究架構-總計畫及子計畫一

Abstract

本計畫基於研究目標與年度工作分配，於本年度(98-99年)建立受試者資料庫並召募受試者，其中受試者資料庫運作所需之各項系統皆已完成並架設網頁對外開放，目前資料庫儲存約1300名受試者資料，其中529位受試者已取得其認知能力中工作記憶測驗的結果。同時，研究團隊亦著手進行包含真人與軟體代理人的雙方喊價市場(double auction markets)實驗，以了解認知能力對於真人決策的影響以及真人與軟體代理人的差異。整體結果而言，真人受試者在市場中的表現比軟體代理人佳，而不同工作記憶測驗成績的受試者在市場上的表現亦有顯著的差異，即工作記憶成績越好的受試者多能在市場中取得更高的利潤，而後我們將更進一步探討認知能力在各項經濟實驗與心理實驗中的重要性，實驗包含：multi-armed bandit problem、預測實驗(prediction experiment or Feldman’s experiment)、蛛網模型 (cobweb models)、資本市場 (asset markets)、寡勝賽局(minority games) 與選美結果預測賽局(beauty contests)等等。

To support the goal of the overall 3 year project, the Experimental Subject Database (ESD) is constructed in the first (98-99) year. The complete ESD are performed on an open website, which is furthered extended with a number of on-going user-supporting systems. About 1300 human subjects are involved in ESD. Among them, 529 subjects have finished their working memory capacity (WMC) test, a test for cognitive capacity. Some of these subjects attend the double auction market experiments with software agents. For a comparative study purpose, we also conduct agent-based market simulation with software agents only. The results, based on limited experiments, have shown that human subjects generally perform better than software agent. Furthermore, on the average, human subjects with high WMC perform superior to those with low WMC. In the next stage, we will expand our current study into other economic and psychological experiments to see how extensive that cognitive capacity can matter, such as multi-armed bandit problem, prediction experiment (Feldman’s experiment), cobweb models, asset markets, minority games and beauty contests.