Original title: Designing Proactive Thought Partners for Writing
Chao Zhang, Abe Davis, Chih-Wei Chen, Chin-Chia Hsu · September 2026
Writing takes many kinds of thinking, like coming up with ideas or editing drafts, and people need different help at different times. Most automated tools only offer basic text help, like autocomplete. To address this, researchers explored proactive thought partnersAI that offers help without being asked, tailored to your writing tasks.. These are AI helpers that step in on their own with higher-level mental support. The team built a test tool that lets writers set up partners by choosing their roles and how proactive they should be. While someone writes, these partners choose the right moments to suggest ideas. The researchers found that writers set up helpers through future planning, used the tips to spark ideas and check their own work, and preferred gentle wording with simple visual designs so interruptions stayed out of the way.
You're in the kitchen, halfway through a recipe, and you suddenly realize you're out of a spice. Usually, you'd have to stop and hunt through cabinets, breaking your rhythm. Here, the AI partners are set up to notice when you might need something and offer a suggestion just in time, so you can keep writing without interruption.
Writers often struggle with getting the right help at the right time. This work designs AI that proactively offers support tailored to individual needs, potentially making the writing process smoother and more effective.
The study involved only 16 participants over one week, which may not generalize to all writers. The probe is a prototype, so real-world implementation might differ. The findings are based on self-reported experiences and may not cover all cognitive aspects of writing.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Designing Proactive Thought Partners for Writing — Chao Zhang, Abe Davis, Chih-Wei Chen et al.
Abstract licence: CC0
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.