21 Degrees Digital has launchd a public AI experiment to explore not only whether AI can predict football, but also whether we trust its answer.
The Leeds agency set up the AI World Cup Predictor at the beginning of the tournament, creating a live public forecast that attempts to call every match, knockout stages and eventual winner while updating daily as the tournament evolves.
The platform combines rankings, squad data, injuries, current form and market indicators to generate predictions and explain its reasoning.
At the moment it believes (spoiler alert):
- Spain will win the tournament
- England will reach the final
- Harry Kane will get the Golden Boot
- Morocco will be as dark horse contenders
So far its track record is just over 60% (so we’ll retract that spoiler alert).
But according the 21 Degrees Digital, the interesting part is what happens when AI encounters uncertainty.
READ MORE – AI-first Leeds agency hires search specialist as growth plans accelerate
It’s found that the project has become less about predictions and more about how AI behaves in unfamiliar environments.
With AI drawing from historical information, this tournament in particular has added issues, with the expanded format, changing qualification rules and more.
“One of the biggest lessons for us has been that AI still needs context,” said Rory Mason, co-founder of 21 Degrees.
“It can only reason from information and patterns it has seen before. When the environment changes, or sport behaves unpredictably, you start to see the limits as well as the opportunities.
“That’s actually where the learning starts.”
The platform has been set up through a visible process using public information and then updated as reality changes.
“We deliberately built something people could disagree with. That sounds strange, but disagreement creates engagement,” continued Mason.
“If someone thinks the prediction is wrong, they can challenge the inputs and understand the reasoning.”
Mason added: “We’ve enjoyed seeing people challenge it. People have debated the predictions, argued with them and questioned the logic.
“That tells us something bigger. The future probably isn’t people blindly trusting AI.
“It’s people using it as a tool, understanding its reasoning and combining that with human judgement.”
