New studies from OpenAI and MIT explore how using ChatGPT might affect emotional and social well-being. OpenAI reviewed 40 million real interactions and surveyed over 4,000 users. Meanwhile, MIT ran a 28-day controlled study with 1,000 participants, each using ChatGPT for at least five minutes a day.
Led by PhD student Cathy Fang, the MIT study focused on areas like loneliness, emotional dependence, and social interaction. Results showed that both the chatbot’s behavior and the user's approach significantly influenced emotional outcomes. Surprisingly, using emotionally engaging voice modes did not result in more harm than neutral voice or text.
One complex finding showed that personal conversations were linked to more loneliness but less dependence, while non-personal conversations in heavy use led to higher emotional dependence. Researchers didn’t explain why, but noted the need for further study.
The research stresses that AI developers should focus not only on performance and human-like design, but also on how chatbots affect people’s mental health. The teams hope their findings will inspire more institutions to explore these emotional impacts.
Vocabulary:
• spontaneous (adj): happening naturally without planning.
• correlated (verb): connected or related to something.
• dependence (noun): a state of needing something or someone too much.
• outcomes (noun): results or consequences.
• mutual (adj): shared by two or more people or things.
• engaging (adj): attracting interest or attention.
• contradiction (noun): a situation where two things are opposite or inconsistent.
• observing (verb): watching or monitoring something carefully.
• problematic (adj): causing difficulty or trouble.
• controlled environment (noun phrase): a setting where conditions are managed for study purposes.
• collaborated (verb): worked together on a project.
• platforms (noun): digital services or tools.
• institution (noun): an organization or structure, often related to education or research.
Source: CBS News