Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

"I never would have thought to say this": Example-Based Exploration to Balance Scientists' Writing Preferences with Public Science Communication Strategies

Grace Li, Yuanyang Teng, Juna Kawai-Yue, Unaisah Ahmed, Anatta S. Tantiwongse, Jessica Y. Liang, Dorothy Zhang, Kynnedy Simone Smith, Tao Long, Mina Lee, Lydia B Chilton

TL;DR

This paper investigates how readers respond to three public-science-writing strategies— relatable examples (E), step-by-step walkthroughs (W), and personal language (P)—and introduces a contrastive-example system to help scientists balance these strategies with traditional scientific norms. Through a reader study using AI-generated parallel narratives and a writer study with 10 CS researchers, the work shows that reader preferences are nuanced and context-dependent, with examples and personal language generally favored, while walkthroughs yield topic-dependent effects. The findings inform a two-step writing-support framework: first decide structure (one, none, or many examples) and then calibrate style (with or without personal language), allowing scientists to adapt writing to topic and audience. The research demonstrates that presenting contrastive options helps scientists make deliberate, topic-aware decisions about public science communication, with practical implications for AI-assisted writing tools and outreach practices.

Abstract

Public-facing science communication is important in garnering interest, engagement, and trust in science. Social media platforms provide scientists with opportunities to reach broader audiences, yet many resist adopting social media writing strategies because the strategies conflict with traditional science writing norms and personal preferences. To address this gap, we first evaluate readers' preferences for strategies such as examples, walkthroughs, and personal language. While many readers enjoyed science narratives that used these strategies, their effectiveness was nuanced and context-dependent, varying by topic and individual preference. Building on these findings, we design a system that uses contrastive examples to help scientists adopt and integrate these social media science writing strategies. In a user study with scientists, we found that presenting contrastive examples helped writers critically evaluate different narrative options, balance competing goals, and gain confidence in adapting social media writing strategies to fit both their topic and audience.

"I never would have thought to say this": Example-Based Exploration to Balance Scientists' Writing Preferences with Public Science Communication Strategies

TL;DR

This paper investigates how readers respond to three public-science-writing strategies— relatable examples (E), step-by-step walkthroughs (W), and personal language (P)—and introduces a contrastive-example system to help scientists balance these strategies with traditional scientific norms. Through a reader study using AI-generated parallel narratives and a writer study with 10 CS researchers, the work shows that reader preferences are nuanced and context-dependent, with examples and personal language generally favored, while walkthroughs yield topic-dependent effects. The findings inform a two-step writing-support framework: first decide structure (one, none, or many examples) and then calibrate style (with or without personal language), allowing scientists to adapt writing to topic and audience. The research demonstrates that presenting contrastive options helps scientists make deliberate, topic-aware decisions about public science communication, with practical implications for AI-assisted writing tools and outreach practices.

Abstract

Public-facing science communication is important in garnering interest, engagement, and trust in science. Social media platforms provide scientists with opportunities to reach broader audiences, yet many resist adopting social media writing strategies because the strategies conflict with traditional science writing norms and personal preferences. To address this gap, we first evaluate readers' preferences for strategies such as examples, walkthroughs, and personal language. While many readers enjoyed science narratives that used these strategies, their effectiveness was nuanced and context-dependent, varying by topic and individual preference. Building on these findings, we design a system that uses contrastive examples to help scientists adopt and integrate these social media science writing strategies. In a user study with scientists, we found that presenting contrastive examples helped writers critically evaluate different narrative options, balance competing goals, and gain confidence in adapting social media writing strategies to fit both their topic and audience.

Paper Structure

This paper contains 44 sections, 7 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Background on Tweetorials
  3. Key Strategies of the Tweetorial Body
  4. Example (E)
  5. Walkthrough (W)
  6. Personal Language (P)
  7. Related Work
  8. HCI for Science Communication
  9. LLM-based Writing Support
  10. Reader Study: Methodology
  11. Research Questions
  12. Automatic Narrative Generation Strategy
  13. Generating Tweetorials with and without Examples (H1)
  14. Generating Tweetorials with and without Personal Language (H3)
  15. Generating Tweetorials with and without Walkthroughs (H2)
  16. ...and 29 more sections

Figures (7)

  • Figure 1: Annotated Tweetorial on the topic of Walker's Action Decrement Theory in Psychology with color highlights corresponding to Example, Walkthrough, and Personal Language.
  • Figure 2: H1: Example Sample explanation generations on the topic of "Walker's Action Decrement Theory" in Psychology comparing the experimental condition (EWP) and baseline condition (WP).
  • Figure 3: H3:Personal Language Sample explanation generations on the topic of Walker's Action Decrement Theory in Psychology comparing the experimental condition (EWP) and baseline condition (EW).
  • Figure 4: H2: Walkthrough Sample explanation generations on the topic of Back Propagation in Computer Science contrasting experimental condition (EWP-NoFewShot) and baseline condition (EP-NoFewShot).
  • Figure 5: Study interface: Viewing different structure options (1st of 3 steps)
  • ...and 2 more figures