Analysis of Team Tutoring Training Data

Thumbnail Image
Supplemental Files
Date
2017-01-01
Authors
Kohl, Adam
Gilbert, Stephen
Winer, Eliot
Dorneich, Michael
Bonner, Desmond
Slavina, Anna
Major Professor
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Person
Dorneich, Michael
Professor
Person
Gilbert, Stephen
Associate Professor
Research Projects
Organizational Units
Organizational Unit
Organizational Unit
Organizational Unit
Organizational Unit
Organizational Unit
Journal Issue
Is Version Of
Versions
Series
Department
Aerospace EngineeringMechanical EngineeringVirtual Reality Applications CenterPsychologyIndustrial and Manufacturing Systems EngineeringMechanical Engineering
Abstract

In 2015, the Army identified intelligent tutoring as a key tool for preparing soldiers in an ever-changing world. Intelligent tutoring uses targeted feedback to train soldiers in critical skills. Not only can tutoring train individuals, it can also train teams. As missions become more complex, success requires teamwork. Interactions between team members directly impacts outcomes regardless of individual performance. Currently, there exists some literature looking at the challenges of intelligent team tutoring, however, many of these scenarios have well defined roles that lend themselves well to constructing behavior rules. While these structured roles are easy to construct tutors for, they do not reflect real world scenarios. For the potential of intelligent team tutoring to be fully realized, more realistic tasks need to be studied. This work outlines the analysis strategies developed to decipher task performance from team tutoring data. A simulation-based military resonance task was developed that required communication and coordination between two trainees. The task was designed using VBS2 as the simulation engine and intelligent tutoring was implemented using the Generalized Intelligent Tutoring Framework (GIFT), adapted for teams. Quantitative data collected included entity positions, tutor feedback instances, and subtask performance. As is typical with team tutoring, the amount of data was very large (compared to individual tutoring) and noisy. This paper presents the multiple data parsing strategies and visualizations developed to fully understand the team interactions which took place. These strategies allowed targeted improvement of a team’s deficiencies.

Comments

This proceeding is published as MacAllister, Anastacia, Adam Kohl, Stephen Gilbert, Eliot Winer, Michael Dorneich, Desmond Bonner, and Anna Slavina. "Analysis of team tutoring training data." MODSIM World (2017). Paper no. 60. Posted with permission.

Description
Keywords
Citation
DOI
Source
Copyright
Sun Jan 01 00:00:00 UTC 2017