The sporting world often only knows months or years later, through complex verification procedures, whether top performances really deserve their fame – or whether doping was involved. Scientists from Saarland believe that this could soon change thanks to artificial intelligence.
2028 Summer Olympics in Los Angeles, the final of the 100 meter run. The fastest men in the world compete against each other. A showdown, a spectacle for the audience. And when the sprinters reach the finish line, inspectors with rapid testing devices are already waiting for them and ask them to take an immediate doping test. As soon as the medals are awarded, it is almost certain whether the athletes will receive a rousing reception at home – or a doping ban.
This is the vision of Wolfgang Maaß, professor at Saarland University. “Currently, blood and urine samples must first be sent from the sports facility to the doping control laboratory. Only there can they be analyzed and only then will it be clear who really won the medals. With rapid doping tests we could bring justice to clean athletes more quickly“, says Maaß in an interview with the ARD doping editorial team.
AI should bring speed and efficiency
Maaß is neither a sports scientist nor a biochemist. And, as he himself says, he actually has no idea what happens in the body during doping. Maaß is a business IT specialist and head of the Smart Service Engineering research group at the German Research Center for Artificial Intelligence DFKI. His expertise lies in optimizing processes in industrial companies and in the healthcare sector – with the help of artificial intelligence.
A few years ago he came up with the idea that the methods he uses for industry should also be applicable in sport – especially in the fight against doping, when countless data is collected. This should make the hunt for fraudsters faster and more efficient. “There will be no other way to do this than with artificial intelligence” says Maaß.
“Just numbers, nothing more“
With a small staff of scientific staff, he feeds AI models with values from positive and negative doping samples. The anonymized data comes from clinical studies at the University of Copenhagen and the Cologne Institute of Biochemistry.
“We just get numbers, nothing more“, explains Maaß. However, it doesn’t work without basic knowledge of biochemistry. The computer scientists use the values of certain biomarkers, i.e. variable parameters in blood or urine that are routinely recorded in every sample. The AI models compare these values , examine all possible connections, look for connections and ultimately the smallest deviations that indicate doping. With the help of machine learning methods – deep learning – the researchers teach the systems to create patterns and thus detect doping more and more accurately.
The end result is not a positive or negative doping test, but rather a finding as to how certain a doping offense took place. For example, the AI would detect much more quickly and with a high degree of accuracy if urine samples were manipulated, as was the case at the 2014 Olympics in Sochi.
Thevis: AI has “big potential“
The prerequisite for such a scenario: sufficient amounts of comparative data. And in practice, a digital athlete passport would have to be created for every athlete who is officially tested, in which the values of all tests are recorded, similar to the existing biological athlete passport.
Mario Thevis, the head of the Cologne doping control laboratory, sees artificial intelligence as a useful addition to the anti-doping fight: “Sifting through the incredible amounts of data we produce here is very time-consuming, very complicated, and perhaps not as detailed as computer algorithms could provide. And here we see great potential for artificial intelligence.“
New EPO proof in sight?
Thevis finds particularly helpful an AI model from the Saarbrücken researchers that can detect the ingestion of artificial EPO, a hormone that is often used by dopers in endurance sports and that the body also produces itself. Thevis: “Distinguishing the artificially produced hormone from the body’s own production is analytically difficult and challenging. If we had further starting points that were filtered out by AI, then we would have taken a big step forward.“
Wolfgang Maaß is also in contact with the World Anti-Doping Agency WADA. But at WADA, the use of artificial intelligence in the fight against doping is apparently not yet at the top of the priority list. “We have not yet been able to develop any major programs with WADA“, says Maaß: “In other areas where I work, for example in healthcare, we are moving at a much faster pace.“
WADA still cautious
However, Maaß does not believe that artificial intelligence could decide the race between fraudsters and investigators in the anti-doping fight in favor of clean sport in the long term. Because the other side, he says, doesn’t sleep either. “Of course, the experts on the other hand will look at how sensitive our methods are and then develop appropriate methods themselves to undermine this sensitivity“, says Maaß.
The scientist had actually hoped that his team’s AI models would celebrate their premiere at the Olympic Games in Paris this summer. But the sports associations are still having a hard time. The Saarbrücken researchers do not receive financial support from sports.
“I hope that in the near future WADA and the IOC will recognize that AI is an important tool for the anti-doping fight of the future“, says Maaß. He is counting on the 2028 Olympic Games in Los Angeles as the start of a new era in the anti-doping fight. And then immediately with a rapid test device in the finish area