When the source is a bot: How people adapt their evaluation strategies to assess AI-generated content

Area of Science:

• Information Science
• Human-Computer Interaction
• Artificial Intelligence Ethics

Background:

• Generative artificial intelligence (GenAI) is increasingly used for scientific content creation and distribution.
• This blurs the lines between expert and non-expert information sources.
• Understanding how users evaluate AI-generated scientific information is crucial.

Purpose of the Study:

• To examine how individuals adapt information evaluation strategies when using GenAI versus traditional search engines.
• To identify distinct evaluation strategies employed when interacting with AI-generated scientific content.
• To extend existing models of online information evaluation to AI-mediated contexts.

Main Methods:

• Performance tasks involving participants evaluating science-related socio-scientific dilemmas.
• Follow-up interviews with 30 adult participants from diverse educational backgrounds.
• Qualitative data analysis to identify emergent evaluation strategies.

Main Results:

• Users employed content and source evaluation strategies on both GenAI and search engines, but adapted them distinctly.
• Emergent strategies include 'representation evaluation' (assessing GenAI's source summarization) and 'meta source evaluation' (relying on source familiarity).
• Participants often exhibited a 'machine heuristic,' over-trusting GenAI's inherent correctness and not using dialogue for credibility assessment.

Conclusions:

• GenAI use necessitates new critical evaluation skills for scientific information.
• Existing information evaluation models need adaptation for AI-mediated environments.
• Addressing user over-trust in automated systems is vital for responsible AI integration.