An artificial intelligence perspective: How knowledge and confidence shape risk and benefit perception

Artificial intelligence (AI) applications are increasingly used in everyday life. Whereas some are widely accepted (e.g., automatically compiled playlists), others are highly controversial (e.g., using AI in the classroom). The public discourse reveals a somewhat ambiguous perception of AI, with enthusiasm about the tremendous positive potential of AI on the one hand and growing concerns about the existential dangers of uncontrolled artificial intelligence on the other hand. Thus, understanding the underlying mechanisms of what shapes people's perceptions of AI is vital to help explain AI usage and acceptance trends. In our research, we take a cognitive approach by measuring the perceived risks and benefits of AI applications considering people's knowledge and confidence in their knowledge. To this end, we assessed in two studies (N = 394 and N = 437; representative) how knowledge about AI and confidence in AI knowledge is related to participants' risk-benefit perception of AI scenarios from three domains: media, medicine, and autonomous driving. Results showed that both AI knowledge and confidence in AI knowledge are predictors regarding people's risk-benefit perception beyond people's attitudes towards AI. More specifically, people with more knowledge about AI exhibited so-called risk blindness; that is, they underestimated the risks. On the other hand, higher confidence in one's AI knowledge impacted participants' benefit assessment. Knowledge and confidence thus open a new dimension of understanding people's perception of risks and benefits in AI.