Correct Yourself, Keep My Trust: How Self-Correction and Social Connection Shape Credibility in Social Chatbots
A new study published on arXiv explores how social chatbots can maintain user trust after making errors. Researchers found that chatbots which self-correct are perceived as more trustworthy and expert than those that rely on external sources for correction. The study also highlights that a strong social connection with the chatbot amplifies the effectiveness of self-correction, influencing belief changes more significantly. AI
IMPACT This research suggests that for AI systems designed for social interaction, self-correction is crucial for maintaining user trust and engagement.