An AI-Resilient Text Rendering Technique for Reading and Skimming Documents

TL;DR

GP-TSM introduces a novel text rendering paradigm that preserves grammar while progressively de-emphasizing non-core details via multi-level opacity, guided by recursive sentence compression and an AI-assisted extraction backend. The approach aims to reduce reading effort and improve skimming without sacrificing content integrity, addressing the shortcomings of traditional summarization that can omit or distort information. Across formative and fully automated user studies, GP-TSM demonstrated faster task completion and higher comprehension, with strong user preference over unigram-frequency opacity baselines. The work argues for AI-resilient reading interfaces that surface the AI’s decisions and preserve access to original text, enabling users to notice and recover from potential AI errors. It also identifies practical considerations, including backend transparency, speed, accessibility, and generalizability to diverse text genres, charting a path toward broader adoption and extension as a Chrome extension or multi-domain tool.

Abstract

Readers find text difficult to consume for many reasons. Summarization can address some of these difficulties, but introduce others, such as omitting, misrepresenting, or hallucinating information, which can be hard for a reader to notice. One approach to addressing this problem is to instead modify how the original text is rendered to make important information more salient. We introduce Grammar-Preserving Text Saliency Modulation (GP-TSM), a text rendering method with a novel means of identifying what to de-emphasize. Specifically, GP-TSM uses a recursive sentence compression method to identify successive levels of detail beyond the core meaning of a passage, which are de-emphasized by rendering words in successively lighter but still legible gray text. In a lab study (n=18), participants preferred GP-TSM over pre-existing word-level text rendering methods and were able to answer GRE reading comprehension questions more efficiently.

Figures (10)

• Figure 1: An illustration of how a paragraph shortens with each round of extraction. Each level stays grammatical after shortening. The increasingly faded text at each level before the final most concise extractive summary show what will be removed at each level; the most faded text at the top level was removed first. What is rendered at the top level in this figure is the only rendering of this process that readers see.
• Figure 2: Screenshot of the HITL-GP-TSM-Interactive interface in the preliminary user study. The (static) HITL-GP-TSM interface is exactly the same, but without the sliders or the responsiveness to mouse scrolling to hide segments of text below a certain level of opacity.
• Figure 3: In the preliminary user study, HITL-GP-TSM-Interactive resulted in significantly better performance on the reading comprehension task than Control---on the order of nearly an entire reading comprehension question out of a total of 4, though participants in the HITL-GP-TSM condition were not far behind. In the HITL-GP-TSM condition, participants completed their reading comprehension tasks significantly faster than when using the Control. The error bars represent standard error.
• Figure 4: Screenshots of the GP-TSM (left) and WF-TSM (right) interfaces in the main user study.
• Figure 5: Participants performed significantly better and significantly faster in the reading comprehension task when using GP-TSM compared to the control conditions in the user study. The error bars represent standard error.