New advances are constantly made in biomedical and health natural language processing, but very few of these advances translate to measurable impact in medicine. When new AI and NLP methodologies are used in practice, they often magnify social biases or exhibit other undesirable behaviors. These failures of translation and AI-related injustices stem from a common source: the need to move beyond benchmarked technology design alone and account for the contexts in which AI and NLP systems are designed, evaluated, and deployed.
In this talk, Newman-Griffis highlights recent work articulating translational research challenges and opportunities in the biomedical NLP space, using disability and rehabilitation as a case study area for technological innovation. Newman-Griffis illustrates the vital role of critical methodologies in aligning NLP system design with clinical goals and reducing AI bias, and calls out key directions for new research on translational challenges in biomedical NLP and laying concrete pathways to equitable impact.
Watch the Recording
Presenter
Denis Newman-Griffis
Denis Newman-Griffis (they/them) is a Lecturer/Assistant Professor in Data Science at the University of Sheffield and a Member of the UK Young Academy. Their work investigates the processes and practices that inform data science and AI development, and how sociotechnical design thinking can help reduce and manage bias in AI/data systems. They bring their work into practice at the intersection of disability and data science, and have developed some of the first deep learning-based NLP systems for analyzing information on disability experience. Denis was recognized with the 2021 AMIA Doctoral Dissertation Award, and their work has been funded by the NIH Clinical Center, the US Social Security Administration, and the National Library of Medicine. Denis is an advocate for LGBTQIA+ support and inclusion in STEM and a member of the AMIA DEI Committee.