Veröffentlichung von Paper in "Journal of Psychosocial Research on Cyberspace"
Montag, 3. Juli 2023
Marisa Tschopp (scip AG), Miriam Gieselmann (IWM Tübingen) und Prof. Dr. Kai Sassenberg (IWM Tübingen, Universität Tübingen) haben zusammen ein Paper mit dem Titel Servant by default? How humans perceive their relationship with conversational AI veröffentlicht. Der Beitrag ist in Cyberpsychology: Journal of Psychosocial Research on Cyberspace (Volume 17, No. 3, 2023) erschienen und setzt sich mit der Beziehung von Nutzern mit Conversdational AI, wie zum Beispiel Alexa, auseinander. Auf der Basis der Relational Models Theory wird ermittelt, inwiefern eine Anthropomorphisierung stattfindet.
Abstract:
Conversational AI, like Amazon’s Alexa, are often marketed as tools assisting owners, but humans anthropomorphize computers, suggesting that they bond with their devices beyond an owner-tool relationship. Little empirical research has studied human-AI relationships besides relational proxies such as trust. We explored the relationships people form with conversational AI based on the Relational Models Theory (RMT, Fiske, 1992). Results of the factor analyses among frequent users (Ntotal = 729) suggest that they perceive the relationship more as a master-assistant relationship (i.e., authority ranking) and an exchange relationship (i.e., market pricing) than as a companion-like relationship (i.e., peer bonding). The correlational analysis showed that authority ranking barely correlates with system perception or user characteristics, whereas market pricing and peer bonding do. The relationship perception proved to be independent of demographic factors and label of the digital device. Our research enriches the traditional dichotomous approach. The extent to which users see their conversational AI as exchange partners or peer-like has a stronger predictive value regarding human-like system perception of conversational
Links
- https://cyberpsychology.eu/
- https://cyberpsychology.eu/article/view/21003
- https://cyberpsychology.eu/issue/view/2453
Tags
Sie wollen eine KI evaluieren oder entwickeln?
Unsere Spezialisten kontaktieren Sie gern!