Science mag recently published a report which revealed that a health risk prediction algorithm used by a major healthcare company was significantly racially biased.
Racial bias in AI is everywhere
You’ve probably heard from us or other outlets about how AI algorithms are biased against people of color. Like how Google’s facial recognition tech identified black people as gorillas and how self-driving cars are more likely to crash into black people.
Because these AI algorithms are shaped and developed by engineers who tend to be predominantly white and male, the parameters used for the machine learning process and the datasets being fed into the system do not reflect the whole spectrum of the human race and we end up with having to deal with systematic racism in the most literal sense.
Less $ gets spent on caring for black patients than white patients
The report found out black patients who were considerably sicker than white patients were rated at the same health risk score by the algorithm – leading to narrower access to health care for black people when compared to white counterparts. Fortunately, the researchers were able to tweak the algorithm and increase the % of black patients receiving additional care from 17.7% to 46.5%.
We need real diversity
The bias problem remains, tho – we can’t be fixing those biases after the fact and instead need to bring in real diversity into AI development remove biases fundamentally.
Google recently got caught using faces of homeless black people to inject some “diversity” into its facial recognition tech.
Fucked up, right? It’s time to cut the BS and actually groom and hire talents from diverse backgrounds.