AI for Science Communication: Adapting to Different Stakeholders
Event details
| Date | 08.01.2025 |
| Hour | 11:00 › 12:00 |
| Speaker | Tal August |
| Location | Online |
| Category | Conferences - Seminars |
| Event Language | English |
Abstract: Communicating complex scientific ideas to the public is critical for an equitable, informed society, but doing so without misleading or overwhelming people is challenging. As large language models become more capable at summarizing and simplifying scientific text, we have a unique opportunity to use these models to make science more accessible. In this talk I will share my group’s research developing language tools and systems to help communicate science to more people. I will highlight two key communication strategies—based on our previous work—focused on different levels of language: explaining new findings from scientific papers and defining individual scientific terms. For both, I will discuss novel techniques we developed for adjusting generated language to fit the needs of different audiences and methods for modeling an individual reader’s background. I will close by discussing how these techniques generalize to other knowledge intensive communication tasks (e.g., legal and educational settings) and the opportunities of developing new techniques for these settings.
Bio: Tal August is an assistant professor in the Siebel School of Computing and Data Science at the University of Illinois at Urbana-Champaign. His group conducts empirical analyses and designs intelligent interactive systems to augment language in knowledge intensive domains like science, health and legal communication. Previously, he was a Young Investigator at the Allen Institute for AI. He received his PhD from the University of Washington, advised by Katharina Reinecke and Noah Smith, where he was supported by a Twitch Research Fellowship.
Practical information
- Informed public
- Free