Conversational AI is increasingly used to help people make everyday decisions, but a new arXiv paper (2605.07789) argues that most systems approach these tasks with data-centric reasoning rather than the heuristics and interactional strategies people naturally use. The authors contend that this mismatch limits the quality of AI support in ordinary decision-making contexts.
To address that gap, the paper analyzes human heuristics and strategies in everyday decision-making conversations. The aim is to ground the design of conversational AI in how people actually talk through choices, rather than in purely statistical or data-driven models. The abstract suggests this could make AI assistants more natural and effective partners in decision-making.
Because the source is an abstract, the article does not detail specific methods or findings. The key contribution is the framing: conversational AI design should start from observed human conversational strategies, not just from large-scale data patterns.