Know Your Audience: The benefits and pitfalls of generating plain language summaries beyond the "general" audience

TL;DR

The paper examines when and how to adapt plain-language summaries for diverse general audiences using three within-subject studies that vary complexity and information content. It contrasts expert-written and machine-generated summaries, showing that low-complexity text benefits readers with little topic knowledge, while more complex, content-rich summaries engage readers with higher familiarity but may reduce information retention. When information content is preserved, the benefits of plain language extend to audiences with limited background, though longer texts risk fatigue. The authors provide practical guidance for adaptive plain-language design and emphasize expert oversight to guard against factual inaccuracies in generated text, aiming to make scientific communication more inclusive without compromising accuracy.

Abstract

Language models (LMs) show promise as tools for communicating science to the general public by simplifying and summarizing complex language. Because models can be prompted to generate text for a specific audience (e.g., college-educated adults), LMs might be used to create multiple versions of plain language summaries for people with different familiarities of scientific topics. However, it is not clear what the benefits and pitfalls of adaptive plain language are. When is simplifying necessary, what are the costs in doing so, and do these costs differ for readers with different background knowledge? Through three within-subjects studies in which we surface summaries for different envisioned audiences to participants of different backgrounds, we found that while simpler text led to the best reading experience for readers with little to no familiarity in a topic, high familiarity readers tended to ignore certain details in overly plain summaries (e.g., study limitations). Our work provides methods and guidance on ways of adapting plain language summaries beyond the single "general" audience.

Figures (6)

• Figure 1: Flowchart of the study method, with shared features of all studies listed once. Ordering of summaries were randomized.
• Figure 2: The study interface for reading the article summaries. The accordions started closed.
• Figure 3: Distribution of ratings for each subjective reading experience measure across complexity levels. The ratings were based on the following questions: Reading ease: "How easy was it for you to read the article?", Understanding: "How confident do you feel in your understanding of the article?", Interest: "How interesting did you find the article?", Value: "How much would you agree that this article contained valuable information?" Notice the greater number of high ratings (purple) and fewer low ratings (orange) as participants are presented with less complex summaries.
• Figure 4: Distribution of ratings for each reading experience measure across complexity and participant topic familiarity for study 1 (expert written summaries).
• Figure 5: Distribution of ratings for each reading experience measure across complexity and participant topic familiarity for study 2 (machine-generated summaries and no information restriction).