Individual Differences that Predict Interactions in Mixed-Initiative Teams: A Big Five Approach
Advisor Information
Doug Derrick
Location
Dr. C.C. and Mabel L. Criss Library
Presentation Type
Poster
Start Date
6-3-2015 9:00 AM
End Date
6-3-2015 10:30 AM
Abstract
Humans and machines are collaborating in new ways and organizations are increasingly leveraging mixed-initiative teams to enable better, faster, and more effective decisions. We examine the effect that an individual’s personality has on his or her willingness to: (1) seek assistance from and/or (2) accept the recommendations of an automated teammate. We use a game of pure strategy with a perfectly accurate decision-assisting automated agent to examine how personality predicts these interactions. Forty-nine participants played 3 rounds of a decision game called “Pirate Island”. Each participant made 27 total decisions (9 decisions per round over 3 rounds) and had the option to solicit assistance from an automated agent for each decision. Participants were not told that the agent was 100% accurate only that it could help them. Using multi-level modeling, we found that people low on extroversion and high on agreeableness were more likely to solicit recommendations from an agent. However, only those high on agreeableness actually accepted the recommendations. We also found that over time, the willingness of users to engage with the agent increased over time. This study bifurcates the behavioral outcomes of soliciting versus accepting instruction from a machine in a mixed-initiative team. This has implications for the type of outcome needed in a mixed initiative team; in some cases, merely asking for more options might improve performance, regardless of whether a user takes a specific option suggested by an automated teammate. This research will lead toward a tighter integration of human and machine intelligence.
Individual Differences that Predict Interactions in Mixed-Initiative Teams: A Big Five Approach
Dr. C.C. and Mabel L. Criss Library
Humans and machines are collaborating in new ways and organizations are increasingly leveraging mixed-initiative teams to enable better, faster, and more effective decisions. We examine the effect that an individual’s personality has on his or her willingness to: (1) seek assistance from and/or (2) accept the recommendations of an automated teammate. We use a game of pure strategy with a perfectly accurate decision-assisting automated agent to examine how personality predicts these interactions. Forty-nine participants played 3 rounds of a decision game called “Pirate Island”. Each participant made 27 total decisions (9 decisions per round over 3 rounds) and had the option to solicit assistance from an automated agent for each decision. Participants were not told that the agent was 100% accurate only that it could help them. Using multi-level modeling, we found that people low on extroversion and high on agreeableness were more likely to solicit recommendations from an agent. However, only those high on agreeableness actually accepted the recommendations. We also found that over time, the willingness of users to engage with the agent increased over time. This study bifurcates the behavioral outcomes of soliciting versus accepting instruction from a machine in a mixed-initiative team. This has implications for the type of outcome needed in a mixed initiative team; in some cases, merely asking for more options might improve performance, regardless of whether a user takes a specific option suggested by an automated teammate. This research will lead toward a tighter integration of human and machine intelligence.