Why Developers are Creating an Angry AI

Customer service isn’t a bad gig most of the time. It’s when that one wrathful person walks into your store that it becomes a nightmare. During my college years working at a coffee shop and a video game store, the easy parts were dealing with stacking shelves and making sure there was enough brew to keep the morning rush happy. But the bile that would be slung at me over the price of tea or why I was unqualified to work in a store because I was a female (true story) would take me aback. But over time, you learn to respond with a calm head and the phrase, “If you have a complaint, please feel free to take it up with our district manager’s office.”

The training videos in conflict management did little to prepare me for a lot of the nuances involved in these confrontations. Much of my learning came from experience in the trenches and advice from managers. However, a company out in New Zealand wants to change that, and make those blissful training videos a thing of the past. Instead, they’re designing a computer program that will mimic the hatred and bile of angry customers to prepare reps to deal with the wrath of riled-up consumers in a multitude of scenarios.

Danny Lewis from Smithsonian writes that the project is called Radiant. The name is a reference to the supercomputer from Isaac Asimov’s Foundation series that had the ability to predict the future.

The designers are currently working on feeding Radiant data over the next six months, which is comprised of collection calls showing human beings at their worst. Radiant will be digesting the data, and sifting through the calls to determine what variables could set people off at any given point on a customer service call. The computer will then turn what it has learned on mankind in order to help train bank telemarketers and customer service representatives in conflict management. The developers are hoping to complete Radiant by the end of the year.

From the description of the project, it sounds like Radiant has been tasked with identifying patterns of what sets off different emotions. According to Stefan Weitz, a senior director of search at Microsoft who knows quite a bit about machine learning, the future of machine learning will be in teaching robots to identify patterns — critically analyzing queries rather than pinging the web to find results.