Answers are grounded in this paper and cite their sources.
Explores trust in a chatbot, emphasizing explainability and user interactions to improve reliance on AI systems.
Intelligent systems require explanations of their underlying decision-making processes, which are vital for developing trust in AI. This is mostly achieved through cognitive modalities that improve trustworthiness using model-driven explanations. Such explanations are often not useful to users and may increase the perceived opacity of AI systems. Our research is focused on a model-driven Q\&A chatbot called MAX. The system was developed to provide access to a company’s internal knowledge during crisis situations. However, users experienced difficulties trusting the chatbot’s answers, particularly when dealing with crucial information. In this paper, we deconstruct the notion of trust in chatbots by conducting a workshop series with designers, developers, and users to identify design features that affect the trustworthiness of chatbots used for informational retrieval. The results show that achieving trust through explainability is not a one-step process. Different users require different levels of explanations and control mechanisms to trust the answers given by the chatbot. This study contributes to utilizing enhanced interactions to facilitate stepwise and appropriate reliance on AI for users who engage with systems pragmatically, provisionally, or skeptically. Interacting with explanations helps users understand the role of AI and supports calibrated trust in practice.
No takes yet. Share an insight, caveat, or question.
Mahmood et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: