Interface
[Overview] The interface created for the experiment in this paper is used for studying how AI assistance affects participants’ interpretive performance and the pleasure they derive from close reading. The main task interface features four components: a task instruction panel, a poem panel, an AI panel (if applicable for the experimental condition), and a text editor for written responses. The interface captures keystroke and cursor events across all panels, enabling analysis of how participants engage with the poem and the AI interpretation, and how they compose their written responses. These raw logs yield behavioral metrics — see the data dictionary.
For reuse:
The interface can be used for human-centered experiments studying AI assistance in literary interpretation and other reading and writing tasks. Researchers can replace the reading materials and AI interpretations, and revise or skip the post-task questions.
The interface has three base configurations (no AI assistance, single AI interpretation, and multiple AI interpretations), which can be used individually, randomly assigned per participant from any subset of the three, or extended by adjusting the number of AI interpretations displayed and their default visibility.
Data and Analysis
© 2026 by Jiayin Zhi.