Perceptual sensitivity to an artificial co-actor in competitive 2D Pong

Gaurav Patil, Lillian Rigoli, Christopher Wahlin, Patrick Nalepka, Rachel W. Kallen, Michael Richardson

    Research output: Chapter in Book/Report/Conference proceedingConference abstract


    Deep reinforcement learning (Deep RL) methods can train artificial agents (AAs) to reach or exceed human-level per-formance. However, in multiagent contexts requiring competitive behavior or where the aim is to use AAs for humantraining, the qualitative behaviors AAs adopt may be just as important as their performance to ensure representativetraining. This paper compares human behaviors and performance when competing against either a human expert oran AA opponent trained using Deep RL on a 2-dimensional version of Pong. Results show that participants were notsensitive to the movement differences between the human expert and AA. Further, the participants did not alter theirbehaviors, except to compensate for differences in the environmental states caused by the opponents. The paper con-cludes with discussion on the potential impacts of AA training on human behavior with regard to representative designin the areas of skill development and team training.
    Original languageEnglish
    Title of host publicationCogSci 2021: program for the 43rd Annual Meeting of the Cognitive Science Society
    Subtitle of host publicationcomparative cognition animal minds
    Place of PublicationSeattle, WA
    PublisherCognitive Science Society
    Number of pages1
    Publication statusPublished - 2021
    EventAnnual Meeting of the Cognitive Science Society (43rd : 2021) - Vienna, Austria
    Duration: 26 Jul 202129 Jul 2021


    ConferenceAnnual Meeting of the Cognitive Science Society (43rd : 2021)
    Abbreviated titleCogSci 2021

    Cite this