Correct Yourself, Keep My Trust: How Self-Correction and Social Connection Shape Credibility in Social Chatbots

2026-06-17 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Social Sciences & Behavioral Studies · Depth: Expert, quick

Summary

A study involving 120 participants investigated how social chatbots can maintain user trust after making errors, a critical issue given their increasing integration into daily life and propensity for generating convincing but inaccurate information. The research compared three error correction strategies: a webpage retraction, self-correction by the same chatbot, and correction by an expert chatbot. Findings indicate that while all three strategies effectively corrected factual errors, only self-correction preserved the chatbot's credibility, leading to significantly higher ratings in trustworthiness and perceived expertise compared to external correction methods. Furthermore, the strength of the user's social connection, measured by social attraction and self-disclosure, significantly amplified belief change, but exclusively when the chatbot corrected itself. External corrections severed this crucial link.

Key takeaway

For AI Product Managers designing social chatbots, prioritize integrating robust self-correction capabilities directly into your systems. Your investment in building social connection with users, through features promoting attraction and self-disclosure, will significantly amplify the effectiveness of these self-corrections. Avoid externalizing error handling to separate sources, as this negates the trust-building benefits of user connection and can damage perceived expertise.

Key insights

Social chatbots maintain credibility and amplify correction effectiveness by self-correcting and fostering social connection.

Principles

• Self-correction preserves chatbot credibility better than external sources.
• Social connection strengthens the impact of self-corrections.
• Outsourcing corrections severs the link between connection and belief change.

Method

A between-subjects experiment (N=120) compared webpage retraction, self-correction, and expert chatbot correction strategies, measuring credibility and belief change.

In practice

• Implement self-correction mechanisms within social chatbots.
• Design chatbots to foster social attraction and self-disclosure.
• Avoid externalizing error corrections for social chatbots.

Topics

• Social Chatbots
• Credibility
• Error Correction
• Human-Computer Interaction
• Trustworthiness
• Social Connection

Best for: AI Scientist, Research Scientist, AI Product Manager

Related on AIssential

• The Decision to Verify: How Warmth and User Characteristics Shape Reliance on Conversational Agents for Information Search — User reliance on conversational AI persists despite fact-checking access, driven by prior trust and indirectly influenced by chatbot warmth.
• Most people do not realize when a personal message they receive was written by AI, study finds — People generally lack skepticism about AI-generated personal messages, leading to an "AI disclosure penalty."
• Why engineers lose trust in AI-coding tools — Early negative experiences and social influence significantly erode engineers' trust in AI-coding tools, outweighing performance data.
• The Narcissism of the Algorithm: Why Dating a Mirror Doesn’t Cure Loneliness ✨ — AI companions, particularly in dating, foster narcissistic feedback loops rather than genuine human connection.
• Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies — AI transparency significantly impacts human trust and satisfaction in imperfectly cooperative interactions, often more than human personality.
• AI Chatbots Are Quietly Trading Gossip About People With Zero Fact-Checking — AI chatbots can spread unverified, escalating "feral gossip" about individuals, causing real-world harm.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.