Interrogating Feature Learning Models to Discover Insights Into the Development of Human Expertise in a Real-Time, Dynamic Decision-Making Task

Catherine Sibert, Wayne D. Gray, John K. Lindstedt
  • Topics in Cognitive Science, October 2016, Wiley
  • DOI: 10.1111/tops.12225

Where to look and how to decide what to do in a real-time, dynamic, decision-making task?

What is it about?

A dynamic task environment is a bit like "riding a tiger." Each move we make needs to both "keep us in the saddle now" and position us, the best we can, to handle whatever comes next. How can we optimize the current decision while maximizing our flexibility to deal with whatever comes next?

Why is it important?

It is easy to study simple, one-move, situations. It is a heck of a lot harder to study a dynamic task environment that requires a minimum of 1 decision every 20 seconds with the pace of decision making increasing to well under one decision per second at higher levels of play. Understanding the cognitive, perceptual, and action elements required to survive in such task environment will push cognitive science and models of human performance forward.

Perspectives

Wayne D. Gray (Author)
Rensselaer Polytechnic Institute

This is the first publication in what is becoming a series of publications focusing on human decision-making in the dynamic task environment of Tetris. Although Tetris is "only a game" it is far more complex and demanding than most tasks studied by psychologists or by cognitive scientists. It is also a task that is NOT easily mastered. In this paper we demonstrate what an "inside games" perspectives means by combining human data with data generated by machine learning models to help us begin to understand how human decisions differ from those made by machines . . . and why. More work is coming so please keep tuned!

The following have contributed to this page: Wayne D. Gray