The scenario described in the video, which is the main image of the article, is just fiction for an ordinary person, but for a Finnish athlete it could be a reality in the next few years.
Canadian professor Richard McLaren and his research colleagues Martin Dubbey became famous for exposing the Russian doping scandal, but now they are working on a new project. McLaren and Dubbey are pushing technology based on sound analysis for international doping testing.
The topic surfaced at the Play The Game conference in Trondheim in early February.
There, McLaren and Dubbey demonstrated an automated telephone survey for athletes. The athlete must answer each question either yes or no in a phone call lasting about three minutes. After this, the artificial intelligence begins to evaluate the athlete’s answers to doping-related questions using voice analysis.
The topic raised a lot of eyebrows among the participants of the conference, and Dubbey promised to be interviewed by Urheilu the next day.
– It’s not about a lie detector, Dubbey underlines right at the beginning of the interview.
According to Dubbey, the task of artificial intelligence is only to assess whether the athlete who answered the questions belongs to a high, medium or low risk group in terms of doping. Even if the answers arouse suspicions in the artificial intelligence, the athlete cannot receive a doping verdict from them.
– If an athlete ends up in the high-risk category, the most likely explanation is that the athlete is lying. However, it does not rule out other possibilities. This produces a risk assessment that anti-doping authorities can use in their work, says Dubbey.
Why is this kind of artificial intelligence technology called Protect, which has its roots in the war in Afghanistan, even considered as a new tool for anti-doping work? The main reason is inefficiency.
Help with inefficient testing
According to Wada, the World Anti-Doping Agency, a total of 278,000 doping tests were conducted in 2019.
Every out-of-competition test has a subscriber. The test designer and the test implementer are often different people. The latter travels to the athlete to get a doping sample.
After this, either the tester or the courier service transports the sample to the laboratory, which must have a Wada license. In the world, 30 laboratories have such a license.
Accredited personnel in the laboratory process and analyze the sample and possibly transfer it to long-term storage. From there, it can be applied for retesting within the next ten years from the moment of sampling.
When the entire life cycle of testing and international differences are taken into account, the research group led by Dubbey has estimated that the average price of one doping test hovers around one thousand dollars. So 278,000 doping tests, one thousand dollars per case. Using this math, in 2019 alone, 278 million dollars, or 259 million euros, were spent on doping testing.
So what can you get with this brisk quarter of a billion? According to Wada’s statistics, only 1537 of the tests carried out in 2019 – i.e. 0.55 percent – resulted in a doping conviction.
– If someone told me about a business proposal asking for a quarter of a billion with a return of 0.55 percent, they would be laughed out of the door, Dubbey says and announces the goal of his program, which relies on artificial intelligence, is to reduce the costs of anti-doping operations by 70 percent.
– Based on our pilot projects, 55-60 percent of athletes do not pose a doping risk. If testing is to be targeted, it would be possible to free up the resources spent on the non-risk group for high-risk athletes.
Developed for the war in Afghanistan
Artificial intelligence and its development has been a topic of conversation in recent years, which has aroused a lot of opinions. Some consider it reasonable to utilize it, while others would put the development on hold completely.
The roots of technology based on question battery and artificial intelligence voice analysis go back to the beginning of the 21st century.
At the time, the US military was at war in Afghanistan against the terrorist organization Taliban and sought to recruit local informants. At the same time, the Taliban tried to infiltrate its fighters into the American ranks.
The task of the artificial intelligence utilized by the US military was to use voice analysis to distinguish whether the interviewee was an infiltrator or not.
– In the beginning it was about life and death, Dubbey states.
In the 2010s, technology began to be used in other fields as well. For example, several insurance companies in England harnessed it for their use. Dubbey believes that the 2020s can be the decade of sports.
Dirty boxing referees
Dubbey and McLaren took their first steps with artificial intelligence voice analysis just over two years ago. At that time, AIBA, the international amateur boxing association, was acting as a guinea pig, which was trying to clean its nest of corruption.
In the spring of 2021, AIBA commissioned an external investigation from a company bearing the McLaren name when several AIBA judges were suspected of numerous match manipulations. Dubbey was named the lead investigator on the case.
Through interviews and whistleblowing, Dubbey and his research team found out that AIBA’s leadership had approved match manipulation in events such as the 2016 Rio de Janeiro Summer Olympics. The research team estimated that the bribed ring and score judges had influenced the outcome of at least 11 matches.
However, the results of the matches did not change, because McLaren and Dubbey felt that they did not have sufficiently solid evidence, such as a positive doping test, in the legal room.
The Rio report, made using traditional research methods, was published on September 30, 2021. Just a few weeks later, McLaren and Dubbey began to approach the match-fixing suspicions using artificial intelligence sound analysis.
The work took place at the AIBA World Championships held in Belgrade, Serbia, where Dubbey and his team interviewed the judges and entered audio data for analysis by artificial intelligence.
According to him, artificial intelligence is able to extract revealing elements from a person’s voice when the person is exposed to cognitive pressure such as questions about match manipulation and bribes.
– Artificial intelligence can recognize 40 different risk assessment factors from a person’s voice, says Dubbey.
Although none of the 2016 Olympic referees suspected of corruption were on the job at the 2021 World Championships, AI voice analysis indicated that the culture of corruption had not disappeared from amateur boxing in five years.
As stated at the beginning, based on the voice analysis, artificial intelligence places the target person in either a high, medium or low risk category. At the 2021 World Cup, 30 percent of AIBA’s judges met the high-risk criteria. With the help of the information, AIBA continued its own investigations, which led to numerous dismissals.
– In our most recent review, there were only two percent of high-risk judges, Dubbey says.
Out of the body, but not out of the mind
Doping testing has changed significantly in the current millennium.
The development was very fast, especially in the last decade, when the ability of laboratories to find traces of prohibited substances in stored samples increased up to a thousandfold compared to the previous decade.
This led to hundreds of results being invalidated and medals being redistributed years after the competitions ended.
The same trend has not continued in doping verdicts. The 0.55 percent share from 2019 mentioned by Dubbey is the previous so-called normal year before the corona pandemic. Since then, the percentage of all doping samples that led to a conviction has dropped to less than 0.4.
– Doping testing requires creativity. This technology alone does not solve the challenges related to inefficiency, but it can help direct testing to risk groups, Dubbey says and repeats one phrase during the interview that summarizes the importance of cognitive pressure in the doping issue.
– The prohibited substance may have already left the body, but not the mind.
Dubbey emphasizes that the retesting of samples and the redistribution of medals must not become the new normal. The main purpose of the risk analysis is to shift the focus of anti-doping activities to a more proactive direction.
Dubbey and McLaren’s technology has been peer-reviewed in the United States, and its operation is overseen by Carnegie Mellon University’s New York-based unit. A new peer review round is underway for the development version of the project.
Now the research duo is looking for national anti-doping offices to join their pilot project.
– The Paris Olympics will be embraced too quickly, but the practice can play a significant role in, for example, the next Winter Olympics, says Dubbey and refers to the 2026 Games in Cortina and Milan.
Will the Finnish athlete get a call soon?
Suek’s general secretary also presented at the Play The Game conference in Trondheim Teemu Japisson says that Suek is considering taking part in the AI pilot.
– In Finland, society’s resources are getting tighter all the time. Testing is very expensive because it has to be done in accordance with all processes and athlete-oriented.
– All testing is already very planned, but any means by which testing can be carried out even more precisely are welcome, says Japisson, who lectured on Finnish anti-doping activities.