Reproduce “Thinking Assistants: LLM-Based Conversational Assistants that Help Users Think By Asking rather than Answering” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Soya Park, Hari Subramonyam, Chinmay Kulkarni

Researchers at Emory and Stanford Universities developed "Thinking Assistants," LLM-based conversational agents that scaffold productive self-reflection by asking expert-informed questions rather than primarily providing answers. Evaluations demonstrated these assistants elicited significantly more reasoning and self-disclosure in users compared to traditional advice or question-only chatbots.

Key insights

  • Combining expert-informed questions with guidance ("informed inquiry") is significantly more effective in promoting deep reflection, reasoning, and self-disclosure than providing advice alone or asking questions without domain expertise.
  • Users often initially approach conversational AI with an information-seeking mental model, highlighting a need for design strategies that guide them toward an inquiry-driven interaction style.
  • Thinking Assistants offer a scalable and accessible alternative to human mentorship, proving valuable for low-risk brainstorming and preparation for high-stakes decisions.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.