Director, Pharmacometrics & Systems Pharmacology Pfizer, Inc. tempe, Arizona, United States
Disclosure(s):
Luke Fostvedt, PhD: No financial relationships to disclose
Maximizing the benefits of AI goes in parallel with minimizing the risks, particularly in drug development where the risks of misdiagnoses can be life-threatening. AI and ML methodologies can inherit and even amplify biases present in the training data, resulting in unfair or discriminatory outcomes. Addressing ethical concerns in AI requires a multidisciplinary approach across scientists, policy makers, and professional societies leading the discussion together. This presentation will provide an overview of ongoing efforts from diverse professional and governmental organizations, including the National Institute of Standards and Technology (NIST), FDA, European Union (EU), Institute of Electrical and Electronics Engineers (IEEE), and American Statistical Association (ASA), to define standards and principles to ensure safe and responsible implementation of AI. Specifically, we will learn about the guiding principles of ‘Accountability, Transparency, and Fairness’ and address sources of bias critical to modeling activities. Through awareness of some of the most common sources of bias we can avoid unintended consequences and improve clinical decision making, moving towards a common goal of ensuring ‘trustworthy’ AI
