When Chatbots Can Start to Seem Like They Have Minds

"Mind perception is the tendency to attribute mental capacities such as thinking, intentions, or feelings to another entity. Humans naturally do this with other people and animals, but we can also extend it to robots, computers, and even abstract things like corporations or nature."

"Across four experiments involving large language models (LLMs) such as ChatGPT, LLaMA, and Claude, we found that even brief exposure can increase people's perception that these systems possess mind-like qualities."

"Psychologists often describe mind perception along two main dimensions. The first is agency, which refers to the capacity to think, plan, and act. The second is experience, which refers to the capacity to feel emotions or sensations such as pleasure, pain, or fear."

Mind perception is the psychological tendency to attribute mental capacities such as thinking, intentions, or feelings to entities. While humans naturally extend this to people and animals, they can also apply it to robots, computers, and abstract concepts. Research across four experiments with large language models including ChatGPT, LLaMA, and Claude demonstrates that even brief interaction increases mind attribution to these systems. Psychologists describe mind perception along two dimensions: agency (capacity to think, plan, and act) and experience (capacity to feel emotions). Machines typically receive higher agency attribution than experience attribution, but large language models blur this boundary through fluent, conversational communication.

Read at Psychology Today