Gastroenterology and Artificial Intelligence: 4th Annual Global Artificial Intelligence Summit (On-Demand | August 2022-Clinicians Trust in AI, Fairness and Bias

Pdf Summary

In this document, Dr. Sravanthi Parasa discusses the issue of bias and fairness in artificial intelligence (AI) in the context of healthcare, specifically in endoscopy. The document highlights the importance of clinicians trusting AI systems and provides insights into how bias can emerge in AI algorithms.<br /><br />The document begins by illustrating the potential for bias by presenting a simple image of a watermelon and showing how different individuals may perceive and describe it differently. This serves as a metaphor for how bias can be introduced into AI systems through incomplete or biased data.<br /><br />Dr. Parasa outlines various factors that contribute to bias in AI, including asking the wrong question, unrepresentative training data, bias within the training data, and curation and optimization choices that lead to differential performance. These factors can result in problematic outcomes and disparate impact.<br /><br />The document emphasizes the need for clinicians to understand the algorithm behind AI systems and highlights the importance of clear design specifications, training data, performance evaluation, and robustness of the model. Trustworthy AI requires a basic understanding of algorithms, traceability of data, explainability, and an understanding of bias.<br /><br />The document also discusses the role of model governance in ensuring trustworthy AI. It suggests transparency in disclosing the data used to train the algorithm, including information about the demographics represented in the training data.<br /><br />Regulatory and legal aspects related to AI in healthcare are also mentioned, including compliance with medical device regulations, data protection regulations, and federal efforts to address bias and fairness in AI.<br /><br />Overall, the document emphasizes the importance of clinicians trusting AI systems in healthcare and provides insights into how bias can be addressed and mitigated in AI algorithms used in endoscopy.

Asset Subtitle

Sravanthi Parasa, MD

Keywords

bias

fairness

artificial intelligence

healthcare

endoscopy

clinicians

algorithm

training data

transparency

regulations