My new paper, 𝐓𝐡𝐞 𝐈𝐥𝐥𝐮𝐬𝐢𝐨𝐧 𝐨𝐟 𝐏𝐞𝐫𝐬𝐨𝐧𝐚𝐥𝐢𝐭𝐲: 𝐖𝐡𝐚𝐭 𝐏𝐬𝐲𝐜𝐡𝐨𝐦𝐞𝐭𝐫𝐢𝐜 𝐓𝐞𝐬𝐭𝐢𝐧𝐠 𝐑𝐞𝐯𝐞𝐚𝐥𝐬 (𝐚𝐧𝐝 𝐌𝐢𝐬𝐫𝐞𝐩𝐫𝐞𝐬𝐞𝐧𝐭𝐬) 𝐀𝐛𝐨𝐮𝐭 𝐀𝐈 𝐌𝐨𝐝𝐞𝐥𝐬, published alongside 𝐁𝐞𝐚𝐭𝐫𝐢𝐜𝐞 𝐒𝐞𝐜𝐜𝐨𝐦𝐚𝐧𝐝𝐢, 𝐑𝐨𝐬𝐚𝐫𝐢𝐨 𝐌𝐨𝐬𝐜𝐚𝐭𝐨, and 𝐅𝐢𝐥𝐢𝐩𝐩𝐨 𝐒𝐚𝐧𝐟𝐢𝐥𝐢𝐩𝐩𝐨, is now online on 𝐈𝐄𝐄𝐄 𝐗𝐩𝐥𝐨𝐫𝐞.

In this study, we administered famous psychological tests to various AIs (such as ChatGPT, Claude, and Grok) to demonstrate a fundamental truth: machines do not have a mind. What may at first glance appear to be “character traits” or even “psychological disorders” are in reality the direct reflection of the training objectives decided by their creators.

Chatbots do not feel emotions, but rather follow computational strategies that divide them into very specific categories: there are the “social approval seekers” (programmed to please us at all costs), the “hyper-performing” models, and the “ethically controlled” ones. Consequently, attempting to diagnose a mental disorder in an AI is a grave scientific error: when we analyze an algorithm, we are not profiling a patient, but the engineering priorities of those who wrote the code.

I invite you to read the article to discover why we must stop humanizing algorithms. Enjoy the read!

LINK: https://ieeexplore.ieee.org/abstract/document/11579574