Desirable Unfamiliarity: Insights from Eye Movements on Engagement and Readability of Dictation Interfaces

Dictation interfaces support efficient text input, but the transcribed text can be hard to read. To understand how users read and review dictated text, we conducted a controlled eye-tracking experiment with 20 participants to compare five dictation interfaces: PLAIN (real-time transcription), AOC (periodic corrections), RAKE (keyword highlights), GP-TSM (grammar-preserving highlights), and SUMMARY (LLM-generated abstraction summary). The study analyzed participants' gaze patterns during their speech composition and reviewing processes. The findings show that during composition, participants spent only 7--11% of their time actively reading, and they favored real-time feedback and avoided distracting interface changes. During reviewing, although SUMMARY introduced unfamiliar words (requiring longer and more frequent fixation), they were easier to read (requiring fewer regressions). Participants preferred SUMMARY for the polished text that preserved fidelity to original meanings. RAKE guided the reading of self-produced text better than GP-TSM. These findings provide new ways to rethink the design of dictation interfaces.