Bad Choices!! Using Big, Long, and Multivariate Data to Explore Blunders Made by Teams and Individuals, Experts and Novices in Dynamic Skilled Performances
Abstract
Bad choices!! We all make them. Indeed, when made by someone who should know better, say someone who is an expert, we call them blun"ders. How do the blunders made by experts differ from the bad choices made by novices? Does the phenomena of expert blunders just apply to individuals or does it extend to bad choices made by expert teams? Do the causes and patterns of extreme expertise generalize across domains?Our proposal builds on past work that sampled individual expertise across a wide range of skill on dozens of measu"res (i.e., multivariate data) on two complex dynamic decision tasks. It extends that work by collecting longitudinal data (i.e., Lon"g Data) on tasks performed by individual and harvesting Big Data to supply Long Data on a competitive team task. All three are tasks" in which even hesitating requires a decision to hesitate. For all three, our research focus will be on the bad choices made by novi""ces and blunders made by experts.The Big Data analyses will use a dataset of 15 million team games, harvested from the web, from a"" realtime, team versus team, combat game. For each expert team we will have records for, at least, 150 team competitions. As team me""mbership often changes across games, developing and validating measures of team familiarity are prerequisite to predicting team perf""ormance. l Multivariate data on individuals will include time and location of each screen object, user keystrokes and eye fixations,"" and dozens of features and measures all collected and timestamped to the nearest ms. Some studies will sample expertise"" by examin"ing individual games collected at tournaments or an hour of in-lab play from our archive > 300 individual players. The Long Data studies will collect and examine 40 hours of in-lab game play of two very different individual performance games.Hypotheses based on empirical data will be tested by mathematical and computational models of human performance. Current models have revealed that the" closer we get to the data, the more complex the challenges appear. The novice, whether individual or team, is challenged to simply"" see"" a good opportunity and move"" their markers to those locations. As expertise emerges the challenge becomes one of winning poi""nts or conquering a rival human team. As extreme expertise is acquired, schemes are hatched for disaster avoidance"" to anticipate a"" bad sequence of pieces (in an single-player game) or an unexpected move (in a team game) but when those schemes fail, tried and tru""e disaster recovery"" plans can be invoked so that the individual or team can recover from all but the most dire straits. However, t""he Icarus Hypothesis (discussed in the paper) predicts that inevitably, experts will blunder. Developing cognitive theories and mode"ling formalisms that model the acquisition of expertise and predict performance outcomes will point the way to developing training programs that ensure expertise and reduce expert blunders during fleet operations.This work seeks to break new ground for cross domain generality in the development of extreme expertise. Where past work has noted generalities in the perceptual advantage of experts" in domains as diverse as Medical Diagnosis and Chess, we seek evidence for the increasingly complex role of perceptual learning as"" the demands of domain expertise transitions from static (medical diagnoses based on static imagining technologies), to slow moving"" (such as chess positions), to dynamic (such as for videogames).
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 29, 2017
- Source ID
- N000141712943
Entities
People
- Wayne D. Gray
Organizations
- Office of Naval Research
- Rensselaer Polytechnic Institute
- United States Navy