In-Context Learning for Human-AI collaboration: Methods and Measures
Visiting speaker
Sarah M. Preum
Past Talk
Hybrid
Thursday
Oct 17, 2024
Watch video
3:30 pm
EST
Virtual
177 Huntington Ave.
11th floor
11th floor
Devon House
58 St Katharine's Way
London E1W 1LP, UK
58 St Katharine's Way
London E1W 1LP, UK
In-Context Learning (ICL) in Natural Language Processing (NLP) enables large language models, such as GPT-3 and GPT-4, to perform tasks based on examples provided in the input prompt, without the need for explicit fine-tuning on task-specific data. ICL has proven to be a promising approach for various domain-specific tasks in NLP. However, it still faces several challenges, such as sample selection, task variability, generalization, and understanding implicit context. In this talk, I will explore how ICL can enhance human-AI collaboration in online environments, e.g., question answering, and knowledge discovery. I will present our recent advancements in developing transformer-based statistical text representation methods to improve sample selection for ICL. By examining a series of challenging NLP tasks, I will highlight cases where ICL is effective and where it encounters limitations. Furthermore, I will introduce novel methods and metrics for evaluating ICL's effectiveness and discuss strategies to enhance its performance in tasks where it currently struggles. The talk will also feature practical use cases of human-AI collaboration in health communication, covering both traditional settings (e.g., online patient portals) and non-traditional settings (e.g., online health communities).
About the speaker
About the speaker
Sarah M. Preum is an Assistant Professor in the Computer Science Department at Dartmouth College and serves as the Technical Associate Director at the Dartmouth Center for Precision Health and Artificial Intelligence. Her research combines natural language processing (NLP) and human-AI interaction to tackle challenges in computational healthcare, contributing to advancements in both fields. She leverages data from Electronic Health Records (EHRs), patient portals, online health communities, social media, and sensors to improve health communication at individual and community levels. To date, Sarah has authored 42 peer-reviewed full papers in top computer science venues, including AAAI, EMNLP, ACL, ICWSM, CHI, IMWUT, JMIR, ICDE, PerCom, and ACM Computing Surveys. Before joining Dartmouth, Sarah was a postdoctoral research scholar at Carnegie Mellon University's School of Computer Science and earned her Ph.D. in Computer Science from the University of Virginia in 2020. She was recognized as one of the "Rising Stars in EECS" in 2020, an international group of emerging female academics demonstrating excellence and a commitment to advancing equity and inclusion in the field. Additionally, she has been awarded the UVA Graduate Commonwealth Fellowship, the Adobe Research Scholarship, the NSF Smart and Connected Health Student Award, and the UVA Big Data Fellowship. Sarah also contributes as an associate editor for ACM Health. Her research has received support from the NSF, NIH, and various pilot grants.
Sarah M. Preum is an Assistant Professor in the Computer Science Department at Dartmouth College and serves as the Technical Associate Director at the Dartmouth Center for Precision Health and Artificial Intelligence. Her research combines natural language processing (NLP) and human-AI interaction to tackle challenges in computational healthcare, contributing to advancements in both fields. She leverages data from Electronic Health Records (EHRs), patient portals, online health communities, social media, and sensors to improve health communication at individual and community levels. To date, Sarah has authored 42 peer-reviewed full papers in top computer science venues, including AAAI, EMNLP, ACL, ICWSM, CHI, IMWUT, JMIR, ICDE, PerCom, and ACM Computing Surveys. Before joining Dartmouth, Sarah was a postdoctoral research scholar at Carnegie Mellon University's School of Computer Science and earned her Ph.D. in Computer Science from the University of Virginia in 2020. She was recognized as one of the "Rising Stars in EECS" in 2020, an international group of emerging female academics demonstrating excellence and a commitment to advancing equity and inclusion in the field. Additionally, she has been awarded the UVA Graduate Commonwealth Fellowship, the Adobe Research Scholarship, the NSF Smart and Connected Health Student Award, and the UVA Big Data Fellowship. Sarah also contributes as an associate editor for ACM Health. Her research has received support from the NSF, NIH, and various pilot grants.
Want to be notified about upcoming NetSI events? Sign up for our email list below!
Thank you! You have been added to our email list.
Oops! Something went wrong while submitting the form
