Automated systems take into account thousands of variables to make decisions that affect our lives. People are calling for more transparency in AI, but not everyone agrees on what constitutes a fai...
The logical implication of enabling counterfactual assessment is that algorithms have to run on attributes, not identities. Thought experiment: I applied for auto insurance and accidentally typed in the wrong zip code, a neighborhood 2 miles away in Brooklyn. The quote was 40% higher than when I fixed it. Is this fair? At least it was possible to run the quote with both zip codes (in other words by attribute).
A thought-provoking article. I don’t know if algorithms can truly be fair because fairness can mean different things to different people. I do believe transparency of how decisions are made is an important step to get closer to fairness.
AI reflects our histories with the data it is fed and our history is made up of racism and bias. The human hand is’t so invisible with AI and its ability to analyze sentiment. Man making AI in his own image is clunky and problematic.AI making AI in his own image...Essentially, I think they’ll be writing programs run by programs to improve this problem.
Doesn’t matter, we will never know. Fairness is an impossible ideal because bias is inevitable. Just like most ideals, fairness doesn’t have a common definition across all humans. The best we can do is to strive to build tools that reflect the way we want the world to be — not the way the world is.
The new existential reality. There may be no easy or clear answers, but answers must we have.