Desirable Behaviors for Companion Bots in First-Person Shooters

Title

Desirable Behaviors for Companion Bots in First-Person Shooters

Creator

Friedman, Adina

Schrum, Jacob

Date

2020-11-20

Date Available

2020-11-20

Date Issued

2019

Identifier

A. Friedman and J. Schrum, "Desirable Behaviors for Companion Bots in First-Person Shooters," 2019 IEEE Conference on Games (CoG), London, United Kingdom, 2019, pp. 1-8, doi: 10.1109/CIG.2019.8848036.

uri

https://collections.southwestern.edu/s/suscholar/item/214

Abstract

First-Person Shooter games are a popular genre that often includes a team deathmatch mode of play: teams of agents score by killing members of the other team. When played without other humans, this mode features both opponent bots and companion bots. This paper uses a human subject study with 30 participants to analyze player preference for cooperative teammates vs. skilled, but less cooperative, teammates in the game Unreal Tournament 2004. Specifically, participants play with both a skilled bot based on neuroevolution and a less skilled bot hand-coded to be more cooperative. Survey results indicate that users perceived significant differences between the bots in several categories, e.g. following behavior and skill at scoring, but did not have a significant preference for one bot over another. However, participants did significantly prefer whichever bot they personally felt was more "helpful" and also preferred whichever bot they happened to see more of. These data indicate how user perception can strongly depend on coincidental interactions, such as being seen more. They also identify some qualities that humans desire in teammates, and indicate that simply scoring more will not necessarily result in a higher user preference.

Language

English

Publisher

Proceedings of the Conference on Games (CoG 2019)

Subject

Computer games

Neuroevolution

First-person shooter games ,

Type

Article