German coach asks AI to avoid the other hand, from God or from Cucurella

On June 22, 1986, Maradona leapt into the air, taking the ball, England’s World Cup hopes and the authority of the referee with the “hand of God”.

38 years later, the controversy with other actors is still ongoing. German coach Julian Nagelsmann began his press conference after the Euro 2024 quarter-final match against Spain by calling for the use of artificial intelligence to referee situations like that of Spanish player Marc Cucurella. He intercepted the ball with his hand in the penalty area in the 106th minute of the match, but the action was not sanctioned.

Spain beat hosts Germany thanks to a goal from Mikel Merino in the 118th minute, shortly before the end of the second half of extra time.

Spain’s epic win was thrown into doubt after a defender allegedly handballed the ball red inside the zone, which would have meant a penalty in Germany’s favor.

Not only do we not evaluate the sporting aspects, but we also wonder whether Nagelsmann’s statement makes sense. That is, if artificial intelligence can help in this regard.

Technology in football

Although FIFA has been reluctant to introduce technological advances into elite football for many years, today we have several examples of successful use, such as video refereeing, semi-automated offside or goal-scoring technologies. The latter consists, according to FIFA, of an automatic goal-line system that instantly determines whether the ball has completely crossed the goal line.

Another technology application that has been introduced recently is the so-called tied ball (the ball is tied) Using the sensor, the ball records up to 500 data per second.

Its processing allows each individual hit that the ball receives to be identified. Thanks to this, the existence of contact with the ball is now beyond doubt. The rest is done by the images, determining whether contact was with the hand or another part of the body. These two aspects, contact and location of contact, seem to be resolved with modern technology.

Trajectory and Intent: Nagelsmann’s Doubts

After the match, the media repeated two statements by the German coach. Infobae quotes the following: “There must be artificial intelligence that calculates where this ball will go.”

And the newspaper Marca highlights the following expression: “I would like your intention with the ball when you hit it with your hand to be valued. It is not known what the intention is (…) there must be an artificial intelligence that would determine this.”

In the first case, it is a trajectory calculation problem. That is, a very complex physical and mathematical problem. It is necessary to take into account variables characteristic of the ball, such as its exit trajectory, linear speed, slope and rotation speed. In addition, it is necessary to take into account external factors such as wind, air humidity or even atmospheric pressure.

Thanks to technology tied balleigenvariables are data that can be obtained realistically. As for atmospheric factors, the only one that is difficult to predict is the exact wind at a given time.

That is, it is possible to build at least an approximate model of the trajectory, which determines with an error whether the trajectory is directed to all three suits. Not counting, of course, the football debates that this forecast can provoke.

Determine intentionality

On the other hand, the sports newspaper emphasizes the statement about intentionality. It seems obvious, given that we are trying to determine the intentionality of a gesture, that from the point of view of artificial intelligence this is a classification problem, in which the possible outcomes are: “intentional” and “unintentional”.

The AI-based classifier system works by learning from previously identified examples. It does not depend on the type of data used for classification. They can be numerical, text, audio or graphic. In this case, we must focus the analysis on image data, given that the few cameras that record elite matches offer almost all angles in very high resolution.

Now that we have the images, we need to train the system on previously classified examples. That is, cases where there is consensus among the judges. Although there are many matches of this image quality collected today, there are a few cases that we could classify as similar to the one we are going to analyze. For example, in the case of Cucurella’s hand, we will not find many examples to train the system on.

On the other hand, the problem of explainability needs to be addressed. It is quite possible that an artificial intelligence can make a decision. You might even make a wise decision. However, let us remember that this is a black box system, which does not offer any explanation of the reasons for the decision. It is clear that this would not be accepted by the relevant hobbies. Therefore, the validity of these systems will be questioned.

Jurisprudence and video arbitration

Now we could approach the problem differently by looking for similar examples. Given the system’s inability to make decisions, we can ask it to look for similar examples.

This already exists. Internet search engines allow searching for images similar to a given one. Thus, the video arbitration system can offer the judge a catalog of previous examples in which consensus was reached on the application of the criterion. The judge’s decision will be based on these examples.

Judges already undergo training in which they are exposed to examples of this type, so we will talk about a decision support system rather than a decision-making system.

Now we are in for a passionate sport. Of course, we would like to move the debate forward. Instead of witnessing an argument about whether Cucurella’s actions were cruel or not, we would start hearing things like: “John Doe’s (fictitious name) shot should have been invalidated in 1996 (fictitious year)” or “if Maradona was a god, his hand would be too.”