‹
Back
11:00 - 12:30
Parallel sessions 2
›
11:00 - 12:30
Room: HSZ - N2
Chair/s:
Nadia Said
The introduction of new technologies has always shaped societies. Artificial intelligence (AI) applications, especially AI chatbots, are already part of everyday human life. Robots – for example in healthcare but also in other service areas – are also becoming more and more common. Generally, perceptions of these new technologies are mixed. Whereas some of them are widely accepted (e.g., use of generative AI tools like ChatGPT or DeepL), others are highly controversial (e.g., use of AI in classrooms or robots as companions in elderly homes). This raises the question of which factors influence human perceptions of and ultimately human interaction with AI and robots? The aim of this symposium is to present novel insights into human-AI and human-robot interaction by taking three different perspectives: (i) how far do social perceptions also extend to robots and how does this influence the interaction with robots?, (ii) what factors shape humans’ interaction with generative AI tools and how does such an interaction impact them?, (iii) do people differ in their perception of AI and robots? To provide answers to these questions, the first talk investigates the perception of robots as social actors. More specifically, the talk focuses on how similar robots are perceived to be, for example, to human partners. The second talk then tackles the question whether established social heuristics (such as the bystander effect) govern human behavior toward robots. Moving from embodied artificial actors to generative AI tools, the third talk focuses on the influence of external factors (explainability, content, culture) on the perception of an AI chatbot. The fourth talk investigates factors influencing the choice to use generative AI as a cognitive offloading tool and its consequences for human memory and performance. In the final talk, the question of whether Artificial Intelligence and robots are perceived differently is discussed. Jointly, these talks provide a broad overview of human-AI and human-robot interaction by examining the topic from different perspectives.
Submission 196
The Influence of Content and Explainability on the Perception of the AI Chatbot „ExplainitAI“ in a Cross-Cultural Comparison Between Germany and South Korea
SymposiumTalk-03
Presented by: Nadia Said
Nadia Said 1, Sora Kang 2, Andreea-Elena Potinteu 1, 3
1 Department of Psychology, University of Tübingen, Germany, Germany
2 Seoul National University, South Korea, Korea, Republic of (South)
3 Leibniz-Institut für Wissensmedien, Germany, Germany
In our study, we investigated whether the perception of an Artificial Intelligence (AI)-driven ChatBot (based on the large language model GPT-4) differs between Germany and South Korea. Furthermore, we investigated possible topic dependencies of AI perception and explored the impact of varying one dimension of explainability, namely "contextual information." To this end, we collected two quota-based samples from Germany and South Korea with a total of N = 297 participants. Results showed significant differences between German and Korean participants in their perceptions of AI. More specifically, Korean participants displayed higher trust, expressed more positive user experiences, and perceived AI as more of an opportunity compared to German participants. Furthermore, we showed that the content provided by the AI influenced participants’ trust in AI. That is, participants reported lower trust in AI for societally debated topics (e.g., migration) compared to health or entertainment topics. Additionally, interactions between countries, content provided by the AI, and explainability conditions further shaped participants’ perceptions, highlighting the interplay of cultural, contextual, and content-specific factors in AI evaluation. These result emphasize the importance of integrating cultural and contextual nuances into the design of AI systems, offering actionable insights for the development of culturally adaptive and explainable AI tailored to diverse user needs and expectations across domains.