I just wanna blame somebody, not something! Reactions to a computer agent giving negative feedback based on the instructions of a person. Issue 154 (October 2021)

• Record Type: Journal Article
• Authors: Horstmann, Aike C.; Gratch, Jonathan; Krämer, Nicole C.
• Is Part Of: International journal of human-computer studies. Issue 154(2021)
• Publication Date: 2021-10
• Subjects: Human-agent interaction -- Automated agents -- Agent representatives -- Agency -- Blame -- Attribution theory; Human-machine systems -- Periodicals; Systems engineering -- Periodicals; Human engineering -- Periodicals; Human engineering; Human-machine systems; Systems engineering; Periodicals; Electronic journals; 004.019
• DOI: 10.1016/j.ijhcs.2021.102683
• Languages: English
• ISSNs: 1071-5819

Abstract

Highlights: Attribution of agency and blame is affected by the interaction partner's type of agency. Most agency and blame is attributed to an avatar controlled by a person in real-time. This is less the case for a person's agent representative, which was instructed by a person beforehand, and the least for a virtual agent, which is controlled by a computer program. Abstract: Previous research focused on differences between interacting with a person-controlled avatar and a computer-controlled virtual agent. This study however examines an aspiring form of technology called agent representative which constitutes a mix of the former two interaction partner types since it is a computer agent which was previously instructed by a person to take over a task on the person's behalf. In an experimental lab study with a 2 × 3 between-subjects-design (N = 195), people believed to study either together with an agent representative, avatar, or virtual agent. The interaction partner was described to either possess high or low expertise, while always giving negative feedback regarding the participant's performance. Results show small but interesting differences regarding the type of agency. People attributed the most agency and blame to the person(s) behind the software and reported the most negative affect when interacting with an avatar, which was less the case for a person's agent representative and the least for a virtual agent. Level of expertise had no significant effect and other evaluation measures were not affected.