Designing for Perceived Intelligence in Human-Agent Interaction: A Systematic Review

Abstract

AbstractThe aim of the current study was to identify design elements that influence the perceived intelligence of an agent to inform the design of human-agent interfaces. An agent’s level of perceived intelligence by the user is one of the essential dimensions of user experience in human-agent interaction. Studies have shown that perceived intelligence affects outcomes such as intentions to purchase and use. However, there is a lack of comprehensive interface design guidelines for improving perceived intelligence. We conducted a systematic literature review on empirical studies on perceived intelligence in human-agent interaction in academic literature databases and grey literature databases. A total of 2133 articles were screened, and subsequently, 55 relevant articles were identified, reviewed, and analyzed. Perceived intelligence was studied in various domains, such as robots, computer games, ambient intelligence, and interface agents. Design elements that can influence perceived intelligence mainly include anthropomorphism, animation, behavioral diversity, information pushing, decision recommendation, automated action implementation, interaction modality, transparency, and user control. These design elements were classified into three dimensions according to their conceptual similarity: appearance design, behavioral proactiveness, and communication effectiveness. The results can provide guidance to future research and practice for human-agent interaction. More research is needed to explore a broader range of design elements, such as types of human-agent relationships and changes in an agent’s behavior over time, that may also influence perceived intelligence.KeywordsPerceived intelligenceHuman-agent interactionSystematic literature review