The Risk of Racial Bias While Tracking Influenza-related Content on Social Media Using Machine Learning
This on-demand webinar does not offer CE credit.
Lwowski B, Rios A. The risk of racial bias while tracking influenza-related content on social media using machine learning [published online ahead of print, 2021 Jan 23]. J Am Med Inform Assoc. 2021;ocaa326. doi:10.1093/jamia/ocaa32.
Watch the Recording
Presenter
Anthony Rios, PhD
Manager
Hannah Burkhardt
Moderator
Jennifer Rosenbaum, MD, MEd.
Statement of Purpose
Natural language processing (NLP) has a wide array of uses in biomedical informatics. For instance, NLP is used for applications, including, but not limited to, the early detection of disease outbreaks, monitoring adverse drug reactions, behavioral risk surveillance (such as monitoring smoking/drug use), and mining individual's mental and physical health on social media. As NLP methods become more powerful and popular, researchers have raised concerns about bias and robustness. In biomedical applications, biased methods can harm the people they are meant to help.
This work measures potential racial bias in machine learning models trained to detect influenza-related content on social media. Specifically, we look to measure how well models perform on text with African American English-related characteristics. Our experiments show that a wide array of models, from neural networks to linear models, are biased. Moreover, the degree to which specific models are biased can vary, making it essential to test model bias before making significant decisions. Overall, this work is just the first step towards making more fair and equitable use of NLP for biomedical applications.
Target Audience
The target audience for this activity is professionals and students interested in health informatics.
Learning Objectives
The general learning objective for all of the JAMIA Journal Club webinars is that participants will
- Use a critical appraisal process to assess article validity and to gauge article findings' relevance to practice
After participating in this webinar, the listener should be able to:
- Measure potential bias in text classification models for biomedical applications
- Recognize the need for measuring bias across a wide array of biomedical applications
Commercial Support
No commercial support was received for this activity.
Disclosures for this Activity
The following presenters, planners, and staff who are in a position to control the content of this activity disclose that they and their life partners have no relevant financial relationships with commercial interests:
Presenter: Anthony Rios
JAMIA Journal Club planners: Hannah Burkhardt, Kirk E. Roberts, Jennifer Rosenbaum
AMIA Staff: Susanne Arnold, Pesha Rubinstein