Cassie Kozyrkov is more than a data scientist – she's Google’s chief decision scientist, a title that is as much mission statement as job description.
For Kozyrkov, artificial intelligence and big data analytics are far removed from the dystopian visions of science fiction movies. She sees them as just a few new tools used by humans. To make these tools work better, she brings together data and behavioural science with human decision-making.
Born in South Africa, Kozyrkov’s fascination with data began when she was a child, entering information about her gemstone collection into Excel while her friends played outside. After moving to the US as a teen, she got a degree in economics at the University of Chicago – where she “realised what incredible possibilities lie in data”.
She would later get degrees in statistics, neuroscience and psychology, a combination that raised some eyebrows. “When I was being trained, it was really unusual for someone to do both [data and decision science],” she says. “I remember college advisers telling me I was crazy with my list of courses, and why on earth would I want to [study] these completely different things that had nothing to do with one another.”
Kozyrkov first thought that she would become a university professor. But a chance meeting with a data scientist at Google led to a summer internship at the company, followed by jobs as a statistician and machine-learning expert. After a stint as Google Cloud’s chief data scientist, she took on the role of chief decision scientist for the whole company in 2018.
What she does is to bridge departments that usually keep to themselves, all the way from research to the teams that apply algorithms to business functions – training more than 17,000 Google staff in the process.
“Typical data science training might teach you how to analyse survey data but not how to design the survey in the first place. If the survey is poorly designed, no amount of math can help,” she says. Kozrykov wants to use applied data science, AI and analytics to create better tools and products – a discipline that she calls decision intelligence.
In real life, it means letting “well thought-out projects flourish, as well as identifying ill-advised projects so these can be shut down before they begin”. And, by bringing psychology into data science, she hopes to reduce bias in algorithms.
This article was originally published by WIRED UK
