Overview

• The peer-reviewed paper, published August 20 in Communications Psychology, reports eight studies involving hundreds of participants.
• Researchers analyzed lab tasks, real Reddit threads, and responses from ChatGPT to map how advice is framed across contexts.
• Across settings, suggestions to add activities far outnumbered recommendations to remove harmful behaviors.
• Participants rated additions as easier and more beneficial, and they viewed cutting harmful habits as easier to advise for close friends than for themselves.
• The team found ChatGPT mirrored the same additive pattern and recommended design changes to elicit subtractive options in AI-delivered support.