Schelling Point
A Schelling point is a solution that several people arrive at independently because it seems obvious to all of them – without being able to coordinate with one another. The term originates from game theory and explains why groups sometimes coordinate perfectly well without any communication at all.
Imagine you’re supposed to meet a stranger in Berlin. Neither of you may call or write, and no one has told you a location. Yet remarkably many people, in such an experiment, turn up at the Brandenburg Gate. The reason: everyone is not looking for the best place, but for the place they believe the other person also considers obvious. Such a meeting point in people’s minds is called a Schelling point, named after the economist Thomas Schelling. It describes a solution that suggests itself, because it is conspicuous, unique, or familiar to everyone.
Coordination without communication
Very many situations in everyday life and in the economy are coordination problems. There are several solutions that would be fine for everyone, but everyone must choose the same one. Which side of the road to drive on is one such case. Right or left doesn’t matter, as long as everyone does it the same way. As soon as people cannot calmly agree with one another, the question of which option seems most conspicuous becomes decisive.
Schelling showed in the 1960s that classical game theory could not explain this phenomenon. Purely mathematically, all equally valid solutions are interchangeable. But people do not treat them as interchangeable. They draw on shared knowledge: round numbers, well-known landmarks, traditions. It is precisely this shared prior knowledge that turns an option into a Schelling point.
This becomes practically important wherever coordination is expensive, slow, or impossible. Negotiations often end at round numbers, such as a fifty-fifty split, even though 47 to 53 would be just as conceivable. The round number acts as a natural stopping point at which both sides stop haggling.
What turns a point into a focal point
What matters is not what you consider correct. What matters is what you believe the others consider obvious. And they think exactly the same thing about you. This loop of mutual expectations is called shared knowledge: everyone knows something, and everyone knows that everyone knows it. An option wins if it is stable within this loop.
Typical features are uniqueness and conspicuousness. If only one number among ten is printed in red, most people will pick the red one. If a place is the only landmark in a city, it becomes the meeting point. Habit is enough too: whatever has always been done that way remains the reference point, even if it is objectively worse.
A common misconception is to regard a Schelling point as the best solution. It isn’t. It is merely the most expectable one. A group can thus jointly settle on something mediocre, simply because everyone thinks of it immediately. And Schelling points differ culturally: in New York, most people would name Grand Central Station, in Paris the Eiffel Tower. Someone from a different culture can simply miss the point.
From blockchains to AI safety
In tech news, the term appears especially often in the context of cryptocurrencies. A network like Bitcoin only works if thousands of computers, without a central authority, agree on the same version of the chain of data. If the network splits, the community must decide which branch is the real one. Such decisions frequently run through a Schelling point, such as the longest chain or the software maintained by the best-known developer team.
The principle is also embedded in standards. Why did charging cables suddenly end up almost all using the same connector? Because one format became the obvious reference point, and manufacturers didn’t want to be the only ones left out. Similarly, experts in AI safety discuss whether companies, absent a binding agreement, would adhere to shared limits. A clearly formulated, publicly visible threshold can act as a Schelling point that everyone orients themselves by.
The term has also come up in discussions about coordinated behavior among multiple AI systems. If, in the future, many programs negotiate or set prices autonomously, they could settle on shared reference points without communicating with one another. For competition authorities, this is an open question: whether such tacit coordination is still permissible depends on whether it is judged to constitute collusion.