Tweetorial Hooks: Generative AI Tools to Motivate Science on Social Media (2305.12265v2)

Abstract: Communicating science and technology is essential for the public to understand and engage in a rapidly changing world. Tweetorials are an emerging phenomenon where experts explain STEM topics on social media in creative and engaging ways. However, STEM experts struggle to write an engaging "hook" in the first tweet that captures the reader's attention. We propose methods to use LLMs to help users scaffold their process of writing a relatable hook for complex scientific topics. We demonstrate that LLMs can help writers find everyday experiences that are relatable and interesting to the public, avoid jargon, and spark curiosity. Our evaluation shows that the system reduces cognitive load and helps people write better hooks. Lastly, we discuss the importance of interactivity with LLMs to preserve the correctness, effectiveness, and authenticity of the writing.

• The paper introduces an interactive LLM-based tool that generates engaging tweetorial hooks to make STEM topics accessible on social media.
• It demonstrates that incorporating examples in prompts significantly enhances the relatability and curiosity of the generated hooks.
• User studies reveal that the system reduces cognitive load and improves hook quality, helping scientists communicate effectively.

An Expert Overview of "Tweetorial Hooks: Generative AI Tools to Motivate Science on Social Media"

The paper "Tweetorial Hooks: Generative AI Tools to Motivate Science on Social Media," presented at the ICCC '23, addresses the challenge of effectively communicating complex scientific content to the public through social media, specifically focusing on the emerging format of Tweetorials. Tweetorials are a sequence of tweets that explain STEM concepts in engaging, informal narratives to broaden public understanding and engagement.

Key Contributions

The research explores how LLMs can aid STEM experts in crafting the initial "hooks" of Tweetorials—critical first tweets meant to capture readers' attention. The paper presents an interactive system designed to leverage LLMs to support the process, ensuring hooks are relatable, devoid of jargon, and designed to spark curiosity.

1. *Prompting Strategies:*
   • Three prompting strategies were explored to determine how effectively LLMs could generate hooks without human intervention. The strategies progressed from basic prompts to more sophisticated chains incorporating examples and user interactions. The findings indicated that providing examples within prompts enhanced the quality of the hooks generated.
2. *Interactive System:*
   • A web-based tool was developed to guide users through writing engaging hooks. The tool utilizes a step-by-step process where LLMs propose ideas that the user can refine or replace, encompassing everyday examples, common experiences, and personal anecdotes to ensure the output is both accurate and engaging.
3. *User Study:*
   • A paper involving ten participants confirmed that using the system reduced cognitive load and improved the quality of Tweetorial hooks. The structured support allowed users to focus more on content rather than form.

Analytical and Numerical Insights

The analyses revealed that examples in the prompts significantly improved the relatability and curiosity-evoking nature of the generated hooks, as scored by expert annotators. The system incorporating prompting chains saw improved participation and reduced effort, as demonstrated by subjective user ratings in the NASA TLX questionnaires (indicating lower cognitive load and higher satisfaction).

Practical and Theoretical Implications

Practically, the paper underscores the potential of LLMs as cognitive scaffolds in science communication, aiding in the translation of complex scientific concepts into public-friendly narratives. Theoretically, it reinforces the importance of co-creative systems that preserve user agency and style, presenting a hybrid approach where machine suggestions do not overwrite human creativity but support the ideation process.

Future Directions

Future research could expand the system’s applications beyond scientific topics, leveraging the findings to communicate other technical topics to non-expert audiences. Additionally, integration with real-time audience feedback mechanisms could further customize the LLM outputs to match diverse audience expectations across different cultural and social settings.

Ultimately, "Tweetorial Hooks" paves the way for innovative applications of AI in science communication, demonstrating how tools powered by LLMs can facilitate public engagement by making complex topics accessible yet engaging. This bridging of the communication gap empowers both scientists and the public to engage in more meaningful dialogues about science and technology.