What hypnosis can teach us about Artificial Intelligence

A new article by Prof. Giuseppe Riva, Director of the Humane Technology Lab (HTLab) at Università Cattolica, published in the international journal Cyberpsychology, Behavior, and Social Networking (Mary Ann Liebert), proposes an original comparison between hypnosis and Artificial Intelligence.

The study, conducted together with Brenda K. Wiederhold (Virtual Reality Medical Center, San Diego) and Fabrizia Mantovani (University of Milano-Bicocca), shows how the human mind in a hypnotic state and Large Language Models such as ChatGPT share important functional characteristics.

In both cases, behavior emerges from automated pattern-completion processes, without strong executive control. Under hypnosis, the brain reduces critical monitoring and becomes highly sensitive to context; similarly, language models generate fluent responses but may produce errors or false information with high confidence, as they lack genuine internal verification mechanisms.

The most relevant parallel concerns the so-called "meaning gap": both the hypnotized mind and the AI produce linguistically coherent content without authentic understanding. Meaning emerges only through interpretation by the human observer.

The article highlights important implications for the development of safer and more reliable AI. Research on hypnosis offers a useful model for understanding phenomena such as suggestibility, confabulation, and implicit goal formation, which are now central topics in the debate on Artificial Intelligence safety.

As the authors emphasize, the future of AI depends not only on larger and more fluent models, but on the design of hybrid architectures capable of integrating automatic generation with supervision and control systems inspired by the functioning of the human mind.

Read the full article on Cyberpsychology, Behavior, and Social Networking: Automatic Minds: Cognitive Parallels Between Hypnotic States and Large Language Model Processing