Search Papers | Poster Sessions | All Posters
Poster C69 in Poster Session C - Friday, August 9, 2024, 11:15 am – 1:15 pm, Johnson Ice Rink
Perceived AI Consciousness: A Quantitative Exploration of Human Responses
Bongsu Kang1 (), Jundong Kim1, Tae-Rim Yun1, Hyojin Bae2, Chang-Eop Kim1; 1Gachon University, Republic of Korea, 2Seoul National University, Republic of Korea
This study investigates the characteristics that contribute to the perception of artificial intelligence (AI) consciousness in human-AI interactions. Drawing from a pilot survey of 29 participants and their assessments of 39 human-AI exchanges, we quantitatively analyzed the influence of nine key features on the Perceived Artificial Consciousness Index (PACI). Utilizing multiple linear regression models and hierarchical clustering, we identified significant features that lead humans to ascribe consciousness to AI, such as 'Metacognitive Self-reflection' and 'Emotionality', while also revealing individual differences in sensitivity to these features. Our study provides preliminary insights into the factors that might shape perceptions of AI consciousness and highlights the variability in human responses to AI. These insights are important for improving AI design and deepening our understanding of consciousness.
Keywords: Large Language Model Consciousness Human-AI Interactions