Oliver Laas: The Eliza Effect and Critical Thinking

Oliver Laas, in a commentary on Vikerraadio, examines the connection between artificial intelligence and critical thinking. He recalls Alan Turing's test, where a machine is considered intelligent if a human cannot distinguish it from another human during conversation. While this test avoided philosophical debates about the nature of intelligence, it opened the door for Joseph Weizenbaum's chatbot Eliza, created in 1966.

Eliza simulated a psychotherapist by turning users' statements into questions. Despite its simplicity, users perceived it as intelligent and understanding. This phenomenon became known as the Eliza effect – humans' tendency to view coherent speech as meaningful regardless of its source. Weizenbaum unintentionally exploited these psychological traits; his followers did so deliberately.

In cybersecurity, such psychological manipulation is called a social engineering attack, with fraudulent phone calls as a local example. Laas argues that Turing's test enabled "human hacking" in AI development: if a machine's intelligence depends on user perception, creating a convincing impression seems sufficient.

Modern chatbots are far more powerful than Eliza, and their impact on susceptible users may have intensified. Stories increasingly emerge of users convinced that ChatGPT has consciousness, helped them achieve a scientific breakthrough, or even has divine origins. This condition is informally called "AI psychosis."

While OpenAI and other AI companies could address this issue, like social media companies, they may lack incentive – an addicted user is a good customer. Laas emphasizes that critical thinking, fostered by openness and moderate skepticism, is hindered by cognitive biases such as groupthink. He believes the Eliza effect also impedes critical thinking about AI. Although it cannot be entirely eliminated, critical thinking teaches awareness and limitation of its influence, enabling a more rational use of artificial intelligence.