Billions of people are about to watch the World Cup. And millions of them will spend part of it furious, not at the players, not at the coaches, or even the referee, but at a piece of technology called the Video Assistant Referee or VAR. 

A poll of over 7,000 fans found that 91% think football is better off without this system of video analysis and off-pitch officials designed to review and correct on-field referee decisions. 

VAR’s failure is a story about what happens when you deploy technology without asking whether it actually serves the people it's meant to help.  And we are making the same mistake in how we design, build, and use A.I.

FIFA has been promoting the claim that its new AI-enabled VAR will make decisions "faster."  To understand why fans hate this “improvement,” you have to go back to where it all began.

FIFA has been promoting the claim that its new AI-enabled VAR will make decisions "faster."  To understand why fans hate this “improvement,” you have to go back to where it all began: a far simpler, and far more successful, experiment called Goal Line Technology.

In a crucial game between England and Germany in the 2010 FIFA World Cup, English midfielder Frank Lampard's long-range shot struck the goalpost, bounced inside the goal, and into the hands of German goalkeeper Manuel Neuer. It happened so quickly that neither the referee nor his assistants were certain if the ball had crossed the goal line, and they refused to award the goal.

England lost the game, and its “golden generation” of footballers, who, on paper, were among the favorites, were knocked out of the tournament. Television replays showed that the referees had gotten it wrong, and the backlash from fans around the world was loud and clear: referees need the support of technology to do their jobs better. 

In response to this and other “ghost goal” incidents like it, football's global governing body, FIFA, introduced technology into refereeing for the first time through "goal line technology" - a set of 7 high-speed cameras dedicated to automatically checking if the ball had crossed the goal line or not.

An elegant solution with 99.99% accuracy and an instant success. It was this successful experiment that led FIFA to design VAR to help with other types of decisions referees make.  

The reason one worked and the other faces massive backlash from fans, offers lessons that go well beyond football, into how we introduce A.I. into our schools, governments, and homes.

The first is to identify a problem before creating the cool solution. Using high-speed videos for goal line technology was a great solution only because the underlying question was unambiguous: “Did the ball cross this line or not?”, but analyzing 5 different video angles cannot answer a question like “did this player intentionally touch the ball or was it accidental?”

If the only tool you have is a hammer, you tend to see every problem as a nail. 

It's a classic Maslow's hammer: if the only tool you have is a hammer, you tend to see every problem as a nail. And right now, a lot of people are walking around with an awful lot of AI.

The second is to have a clear way to evaluate the solution (and be transparent about it). If the solution isn't working, change the solution. Resist the urge to simply change the parameters by which it is measured.

When VAR repeatedly got handball decisions wrong, instead of changing VAR, FIFA changed the decades-old rules of football to say that "intent doesn't matter. If the ball touches the player's hand even by accident, it will still be ruled handball."

That small change to accommodate the technology created new problems for referees, changed how players play the sport, and shifted the goalposts for measuring VAR's effectiveness. 

Define clear metrics for success, and if the tool fails to meet them, blame the tool, not the user. 

Define clear metrics for success, and if the tool fails to meet them, blame the tool, not the user. The goal is to improve the system, not to redesign reality to accommodate the tool.

The third is to design with people, not for them. FIFA introduced and continues to develop VAR without any meaningful involvement of fans or players - the two key stakeholders in the sport. 

In our roles as designers of tools, especially in the public interest, we’re stewards of a toolkit of solutions, including A.I, which we can help deploy to improve people’s lives. Their experiences should be at the heart of the solutions we create.

Top-down deployment of cool tools developed without engaging those impacted by the problem is at odds with that mission. 

Top-down deployment of cool tools developed without engaging those impacted by the problem is at odds with that mission. 

Finally and perhaps most importantly, culture and emotion matter deeply. In many parts of the world, football is akin to a religion. Players spend their whole lives preparing for a 90-minute game.

For fans, watching their teams through ups and downs is among the most important aspects of their lives. Children learn what respect for referees, rules, and tradition means through sport. 

These are not measurable outcomes, but they’re key to understanding why football matters to people as much as it does. If adding technology detracts from the joy and disappointment of sport (and from respect for the human referee), it is doing a disservice to the people it is meant to serve. 

VAR failed because the people who built it never seriously asked whether it was improving football.

When graduating classes across the US boo mentions of A.I., and crowds protest data centers around the world, they are telling us something football fans told FIFA years ago about VAR: you built this for yourselves, not for us. VAR failed because the people who built it never seriously asked whether it was improving football.

That is exactly the question we are failing to ask about A.I.