Generative AI tools, such as ChatGPT, Google Gemini, and Copilot, are increasingly embedded in everyday life for quick answers for learning, research, and decision making. These AI responses tend to be presented in full confidence, which can lead users to accept information without skepticism. However, not all prompts have clear or definitive answers, and errors or hallucinations can occur. Absolute language can eliminate the instinct to fact check these answers, or question why the program chooses one option over another. When AI models use absolute language to deliver non-binary or objectively false answers, they bypass the user’s natural skepticism. This project investigates linguistic certainty indicators in generative AI responses through a multi-faceted approach: academic research, comparative AI response analysis, and the development of a diagnostic software tool. The project will be composed of research overview, analysis of generative AI responses, and the ethical implications of absolute language. By proposing a framework of uncertainty indicators, this project aims to shift user behavior from passive consumption to active verification, ultimately advocating for greater transparency in AI-human communication. This project was developed as part of the University Library AI Fellows Program, supported by the Library AI Studio and the Provost’s AI Acceleration Program.
Aniela Haines (Fri,) studied this question.