Does ChatGPT have a typical or atypical theory of mind? DOI Creative Commons
Margherita Attanasio, Monica Mazza, Ilenia Le Donne

и другие.

Frontiers in Psychology, Год журнала: 2024, Номер 15

Опубликована: Окт. 29, 2024

In recent years, the capabilities of Large Language Models (LLMs), such as ChatGPT, to imitate human behavioral patterns have been attracting growing interest from experimental psychology. Although ChatGPT can successfully generate accurate theoretical and inferential information in several fields, its ability exhibit a Theory Mind (ToM) is topic debate literature. Impairments ToM are considered responsible for social difficulties many clinical conditions, Autism Spectrum Disorder (ASD). Some studies showed that pass classical tasks, however, response style used by LLMs solve advanced comparing their abilities with those typical development (TD) individuals populations, has not explored. this preliminary study, we administered Advanced Test Emotion Attribution Task 3.5 ChatGPT-4 compared responses an ASD TD group. Our results two had higher accuracy understanding mental states, although ChatGPT-3.5 failed more complex states. emotional performed significantly worse than TDs but did differ ASDs, showing difficulty negative emotions. achieved accuracy, recognizing sadness anger persisted. The adopted both appeared verbose, repetitive, tending violate Grice's maxims. This conversational seems similar high-functioning ASDs. Clinical implications potential applications discussed.

Язык: Английский

Does ChatGPT have a typical or atypical theory of mind? DOI Creative Commons
Margherita Attanasio, Monica Mazza, Ilenia Le Donne

и другие.

Frontiers in Psychology, Год журнала: 2024, Номер 15

Опубликована: Окт. 29, 2024

In recent years, the capabilities of Large Language Models (LLMs), such as ChatGPT, to imitate human behavioral patterns have been attracting growing interest from experimental psychology. Although ChatGPT can successfully generate accurate theoretical and inferential information in several fields, its ability exhibit a Theory Mind (ToM) is topic debate literature. Impairments ToM are considered responsible for social difficulties many clinical conditions, Autism Spectrum Disorder (ASD). Some studies showed that pass classical tasks, however, response style used by LLMs solve advanced comparing their abilities with those typical development (TD) individuals populations, has not explored. this preliminary study, we administered Advanced Test Emotion Attribution Task 3.5 ChatGPT-4 compared responses an ASD TD group. Our results two had higher accuracy understanding mental states, although ChatGPT-3.5 failed more complex states. emotional performed significantly worse than TDs but did differ ASDs, showing difficulty negative emotions. achieved accuracy, recognizing sadness anger persisted. The adopted both appeared verbose, repetitive, tending violate Grice's maxims. This conversational seems similar high-functioning ASDs. Clinical implications potential applications discussed.

Язык: Английский

Процитировано

0