AI systems are built by people, trained on data collected from the world, and then used to make decisions about hiring, lending, policing, and more. That means they can inherit and even amplify existing unfairness, often while looking neutral and objective. Understanding where bias comes from is a core part of building AI responsibly.
The lessons start with an overview of the main types of algorithmic bias, then look at a real case, bias in facial analysis, and finish with a talk on why algorithms need scrutiny and how an algorithmic audit can help.
Watch the lessons in order. If a video runs long, feel free to treat it as a reference you dip back into later rather than something to finish in one sitting.
Five common types of algorithmic bias, including data that reflects existing bias, unbalanced training data, feedback loops, and malicious data, plus ideas for spotting and reducing them.
Video 11minJoy Buolamwini shares how facial analysis software failed to detect her face, what she calls the "coded gaze," and why inclusive data and accountability in coding matter.
Video 8minData scientist Cathy O'Neil explains how secret, important algorithms can cause harm, why they are opinions embedded in code, and what an algorithmic audit looks like.
Video 13min