We are pleased to announce our latest peer-reviewed publication, “World-making for a future with sentient AI” in the British Journal of Social Psychology. In this article, Janet Pauketat, Ali Ladak, and Jacy Reese Anthis examined the psychological antecedents to world-making beliefs, attitudes, and expectations for a possible future with widespread sentient AI with data from Waves 1 and 2 of the Artificial Intelligence, Morality, and Sentience (AIMS) survey.
Abstract
The ways people imagine possible futures with artificial intelligence (AI) affects future world-making—how the future is produced through cultural propagation, design, engineering, policy, and social interaction—yet there has been little empirical study of everyday people's expectations for AI futures. We addressed this by analysing two waves (2021 and 2023) of USA nationally representative data from the Artificial Intelligence, Morality, and Sentience (AIMS) survey on the public's forecasts about an imagined future world with widespread AI sentience (total N = 2401). Average responses to six forecasts (exploiting AI labour, treating AI cruelly, using AI research subjects, AI welfare, AI rights advocacy, AI unhappiness reduction) showed mixed expectations for humanity's future with AI. Regressions of these forecasts on demographics such as age, the year the data was collected, individual psychological differences (the tendency to anthropomorphise, mind perception, techno-animist beliefs), and attitudes towards current AI (perceived threat and policy support) found significant effects on all forecasts from mind perception, anthropomorphism, and political orientation, and on five forecasts from techno-animism. The realized future that comes to pass will depend on these dynamic social psychological factors, consequent changes in expectations, and how those expectations shape acts of world-making.
Author’s preprint version (non-paywalled): https://sentienceinstitute.org/downloads/World-Making-for-a-Future-with-Sentient-AI.pdf
British Journal of Social Psychology version of record (paywalled): https://doi.org/10.1111/bjso.12844