WADA looks to artificial intelligence to catch dopers

May 27, 2020
In  this February 5, 2010, file photo, a laboratory technician prepares samples of urine for doping tests during a media open day, at the King’s College Drug Control Centre in London.
In this February 5, 2010, file photo, a laboratory technician prepares samples of urine for doping tests during a media open day, at the King’s College Drug Control Centre in London.

DUSSELDORF, Germany (AP):

With sports around the world shut down by the coronavirus pandemic, the World Anti-Doping Agency (WADA) is looking to artificial intelligence (AI) as a new way to detect athletes who cheat.

WADA is funding four projects in Canada and Germany, looking at whether AI could spot signs of drug use which might elude even experienced human investigators. It's also grappling with the ethical issues around the technology.

Athletes won't be suspended solely on the word of a machine. Instead, AI is a tool to flag up suspect athletes and make sure they get tested.

Previous results

"When you are working for an anti-doping organisation and you want to target some athletes, you look at their competition calendar and you look at their whereabouts, you look at the previous results and so forth," WADA senior executive director Olivier Rabin told The Associated Press in a recent interview. "But there is (only) so much a brain can process in terms of information."

The pandemic has shut down anti-doping testing in many countries, but it's pushed AI work to the fore, since much research can be done remotely.

Analysing an athlete's blood or urine sample is about more than just finding a performance-enhancing substance. Tests also track numerous biomarkers like an athlete's red blood cell count or testosterone levels.

That kind of information is already used by anti-doping bodies in the "biological passport" programme to detect the effects of using something like the blood-booster EPO, the substance used by Lance Armstrong.

WADA hopes AI can help improve that system by tracking patterns between those markers and cross-referencing them with other information. One of WADA's projects aims to make EPO detection more precise and another hopes to do the same for steroids.

Machine learning systems can be taught by showing them confirmed "dirty" and "clean" profiles to detect similarities which may not be visible on the surface.

There's also what Rabin calls a "global" project in Montreal which could predict the risk of doping by evaluating data from a wider range of sources, possibly including the information athletes are required to file about their whereabouts. Athletes' personal data and even the names of the cities where they live and train will be anonymised because of privacy concerns.

Other Sports Stories