Paper Plain: Making Medical Research Papers Approachable to Healthcare Consumers with Natural Language Processing

Tal August, Lucy Lu Wang, Jonathan Bragg, Marti A. Hearst, Andrew Head, Kyle Lo

Bridging the Gap in Health Literacy

When standard medical pamphlets don't provide enough detail, many healthcare consumers turn to primary research literature. However, scientific papers are often impenetrable for non-experts due to technical jargon, dense formatting, and a lack of "structural literacy"—the knowledge of how to navigate a paper to find specific answers.

To address these barriers, researchers developed Paper Plain, an interactive reading interface that leverages natural language processing (NLP) to make medical research more accessible. By providing on-demand "plain language gists" and a question-driven navigation index, the system aims to reduce the cognitive burden of reading dense scientific text without sacrificing objective comprehension.

The Paper Plain Interface

The system integrates four primary NLP-powered features into a web-based PDF reader designed to act as a scaffold for the reader:

  1. Term Definitions: Using medical entity recognition (scispaCy), the system identifies complex terms and provides tooltips with definitions from sources like the Unified Medical Language System (UMLS).
  2. Section Gists: In-situ, plain-language summaries of subsections help readers understand the "big picture" of a section before diving into the details. These are generated by GPT-3 prompted for a 5th-grade reading level.
  3. Key Question Index: A sidebar provides curated questions (e.g., "What did the paper find?") based on the PICO (Population, Intervention, Comparison, Outcome) framework.
  4. Answer Gists: Clicking a question jumps the reader to the relevant passage and displays a simplified summary of the answer.
The Paper Plain interface showing the Key Question Index (left) and an Answer Gist (right) summarizing a highlighted passage.

Key Insights: Scaffolding for Confidence

The researchers evaluated Paper Plain through a usability study with 24 participants, comparing it against a standard PDF reader. The results highlighted several critical insights into how non-experts interact with scientific literature:

  • Guidance Overcomes Linearity: Non-experts often attempt to read papers linearly from start to finish, frequently getting stuck in dense "Methods" sections. The Key Question Index encouraged a more efficient, non-linear navigation strategy.
  • Gists as Scaffolding: Users did not replace the original text with the summaries; instead, they used the "gists" as a mental map to help them parse the more nuanced technical details in the source text.
  • Confidence vs. Comprehension: One of the most striking findings was that while objective comprehension scores remained stable across both groups, participants using Paper Plain reported significantly lower reading difficulty and higher confidence in their ability to find and understand relevant information.

Results and Future Directions

The study confirmed that Paper Plain significantly improves the reading experience for healthcare consumers:

  • Reduced Difficulty: Reading difficulty ratings were significantly lower ($p < 0.0001$).
  • Increased Confidence: Users were more certain they had identified all relevant information ($p = 0.0030$).
  • Feature Preference: The Key Question Index and Answer Gists were the most valued features, helping users bypass the often "demoralizing" experience of navigating complex papers alone.

While promising, the authors note that future versions must address the risk of AI hallucinations in summaries and explore how to tailor information to a reader's specific personal medical history. Ultimately, Paper Plain demonstrates that AI-driven "scaffolding" can empower patients to engage directly with the primary sources of medical knowledge.

Summary updated

Abstract

When seeking information not covered in patient-friendly documents, like medical pamphlets, healthcare consumers may turn to the research literature. Reading medical papers, however, can be a challenging experience. To improve access to medical papers, we introduce a novel interactive interface-Paper Plain-with four features powered by natural language processing: definitions of unfamiliar terms, in-situ plain language section summaries, a collection of key questions that guide readers to answering passages, and plain language summaries of the answering passages. We evaluate Paper Plain, finding that participants who use Paper Plain have an easier time reading and understanding research papers without a loss in paper comprehension compared to those who use a typical PDF reader. Altogether, the study results suggest that guiding readers to relevant passages and providing plain language summaries, or "gists," alongside the original paper content can make reading medical papers easier and give readers more confidence to approach these papers.
BibTeX
@article{august2023paper,
  author = {Tal August and Lucy Lu Wang and Jonathan Bragg and Marti A. Hearst and Andrew Head and Kyle Lo},
  title = {{Paper Plain: Making Medical Research Papers Approachable to Healthcare Consumers with Natural Language Processing}},
  journal = {ACM Trans. Comput. Hum. Interact.},
  year = {2023},
  doi = {10.1145/3589955},
}
Figure