A presentation at PyConline Au by Laura Summers
How do we connect high-level fairness principles with day-to-day product decision making? How do we move past the AI Ethics hype and start trying, testing and implementing practical approaches? In this talk the presenter will share stories, discoveries, and pointers from their time as an ‘ethics ops’ consultant embedded with a small fraud detection team in a big telco. From improving the science bit of data science, to developing the collective sensitivity of the team, to designing recourse for false positives, tune in for pragmatic pointers and actionable take-aways.
In a perfect world we want to build technology that’s perfectly fair, responsible and ethical. In reality it’s hard, for reasons as varied as the difficulty of recovering trust once it’s lost, to pinning down what ‘fair’ is exactly, anyway, to working out how to design solutions at scale that simultaneously cater to individual preferences.
Adding to these baseline challenges, we consider the implications of customers acting in bad faith - everything from minor infractions to fully-fledged fraud that costs thousands of dollars a day. What does this mean for interpreting our in-coming data? Hint: it adds both uncertainty and complexity.
We’ll discuss how to acknowledge these uncertainties without becoming paralysed, the challenges of designing alerts and notifications for high-stress situations, and the difference between an exploratory data science practice versus data science outputs that are ‘load-bearing’.