Algorithmic bias

The technical term is creeper glasses, not pervert glasses.

But Flock's mission, according to the company, is to, quote, "eliminate crime with a principled framework that protects privacy and promotes objectivity." Safiya, given what you know about algorithmic bias, which we'll get in more into in a moment, and the way these systems are built, what do you make of this statement?

Yeah, I mean, I think that the underlying logics of so many of these systems is about using historical data to determine where hotspots are of concern.

And of course, we know that communities of color, Black communities in particular, have been overpoliced.

Today, we welcome Craig Haubin.

He's a healthcare executive, and today's KevinMD article is AI Bias in Healthcare Reads the Writer, Not the Symptom.

Craig, welcome to the show.

Thanks, Kevin.

And I think a lot of people don't know, and I think they'll find interesting and a little scary and a little exciting that I wanted to share and bring people along their AI journey.

All right.

So you tell us a story about one particular example in your KevinMD article, AI bias in healthcare reads a writer not to symptom.

Why did you decide to share this story and then tell us about it for those that you get a chance to read it?

So shall we start by talking

where AI bias actually comes from? Mm-hmm.

So there's three main things, basically, is who builds the tools, who use the tools, and the data that goes in.

What I shared in the presentation is that in this case, when we talk about women and men or genders, Every single one of them is just men's.

But obviously, it must be used, uh, in, in the proper way, uh, with the proper safeguards in place.

Let's understand how we stop AI bias from influencing the outcomes.

In preparation, as they're going through the documents, as they're comparing and contrasting, as they're gathering information, is there...

I mean, already, I think we are all encountering AI bias that happens to some extent in the works that we're doing as we're starting to pivot to these large language models as well, and then I'm learning as well what I can and cannot rely on it for, for, for my industry, for the idea of news gathering, for the idea of, oh, it's missed out this whole segment.

It doesn't even consider this.

It doesn't even factor that in.

How do we make sure AI bias is not gonna be a part of natural or neutral mediations rather?

Hmm.