A web-based feedback study on optimization-based training and analysis of human decision making

Abstract

The question “How can humans learn efficiently to make decisions in a complex, dynamic, and uncertain envi- ronment” is still a very open question. We investigate what effects arise when feedback is given in a computer- simulated microworld that is controlled by participants. This has a direct impact on training simulators that are already in standard use in many professions, e.g., flight simulators for pilots, and a potential impact on a better understanding of human decision making in general. Our study is based on a benchmark microworld with an economic framing, the IWR Tailorshop . N=94 partic- ipants played four rounds of the microworld, each 10 months, via a web interface. We propose a new approach to quantify performance and learning, which is based on a mathematical model of the microworld and optimiza- tion. Six participant groups receive different kinds of feedback in a training phase, then results in a perfor- mance phase without feedback are analyzed. As a main result, feedback of optimal solutions in training rounds im- proved model knowledge, early learning, and performance, especially when this information is encoded in a graphical representation (arrows).