Moving the Goal Posts: Exposure to AI-made Products Changes Evaluation Criteria for Human-made Products

How is human creativity valued in an era of intelligent machines? The question has gathered particular significance as AI becomes increasingly adept at a range of cognitively complex tasks, most notably creative endeavors seen as uniquely human. Prior work on the "AI effect" shows that evaluators "move the goalposts" for AI, changing the criteria for 'intelligence' and 'humanness' to those that remain out of reach for AI. In this study, we ask if the influx of AI-made products changes the way human-made products are evaluated. Using an observational study and a pre-registered experiment, we test how exposure to AI-made art changes the evaluation of human-made art. We find that after the release of OpenAI's ChatGPT, evaluators on online discussions in Reddit art forums and critical reviews on Artnet see criteria not applicable to AI (versus criteria applicable to AI) as more important and valuable. In the online experiment, participants who first evaluate an AI-made sculpture place greater importance on criteria that are less applicable to AI when they subsequently rate a human-made sculpture, relative to participants who only evaluate human-made sculptures.