As part of the honors program, I was able to partake in a course teaching proper oration called "Rhetorics and Ethics". While being equiped with a theory-based toolset for both performing and analyzing speech, we were asked to prepare a TED-style talk to be performed at a concluding event — in front of an audience. Having just taken a class on ethics in data mining as part of my bachelor's, I knew just what topic I wanted to explore: how does bias slip into data-driven decisions?
Every day, decisions are being made that affect each and every one of us. Often, these decisions are based on data, i.e. "data-driven". How come we hear about so much discrimination in those systems (e.g. sexism in application handling, racial bias in sentencing, etc.)?
The primary aim was to sensitize an audience made of non-experts to some basic ways bias and error can slip into datasets and, in turn, into the results of data-driven decisions. Due to this, I had to leave a lot of detail and nuance out while still keeping the core point.
You can find the talk under the following link: YouTube