AIHEALTHCARE
Exploring the Impacts and Implications of Artificial Intelligence & Medical Technology
Ryan Nguyen, Pooja Thorali, Melanie Kuo, Kassidy Gardner
Solutions
Preventing algorithmic bias in healthcare
Health professionals are becoming more aware of algorithmic bias in healthcare, and companies are promoting diversity, equity, and inclusion in their teams to combat this issue. Here are some potential solutions for combatting algorithm bias, as ideated by scholars at Harvard University and the Duke-Margolis Center for Health Policy:
Calibrating Incentives
For professionals in the field of medical AI, power lies within their ability to incentivize private companies to analyze and change biases before their occurrences through the use of legislation in class action lawsuits.
Implementing formal legislation
Designing formal legislation to account for and remove variables that contribute towards unfair judgement according to various subgroup identities (race, gender, socioeconomic status, and disabilities) is another crucial step in combatting medical AI bias. Formal legislation that accounts for these factors must be implemented through a system of checks and balances.
Call for diverse expertise in the field of AI
Diversity is a key factor to reducing biases in artificial intelligence as a whole, as those who come from diverse backgrounds will ultimately have a deeper understanding of problem areas that from experiences within their own backgrounds. By celebrating these differences and using these as a tool to unite those across different subgroups within medical fields, more innovative solutions can be created to solve the issues that are systemically rooted within the medical use of AI.
Hold investors accountable for testing and transparency
Investor stakeholders in healthcare-based AI fields have the power to create change through their funding. Ultimately, their purchasing power should be dedicated towards thorough testing for affected subpopulations and transparency in their work.
Emphasize data equity and fair use
Those who record and handle data for AI research should ensure that their data is recorded ethically and in a just manner, and emphasize its usability for replication and thorough algorithm testing.
FDA and federal agency monitoring of medical AI
Ensure that medically-based AI products are clearly labelled, monitored, and are observed in their performance across different subgroup populations. These governmental agencies should also seek to build systems to regularly observe and monitor medical products on a performance-based level.