← Back to arXiv
arXivLogicarXiv:2609.00633

Infinite Communication Complexity and KW Games

The paper takes a classic idea from computer science and extends it into the realm of infinite mathematics. In the 1990s, Karchmer and Wigderson showed that the complexity of computing a function with a circuit is intimately connected to how hard it is for two parties to communicate and solve a related puzzle together. The authors ask what happens when you stretch this game to be infinitely long, played over infinite sequences of zeros and ones rather than finite strings. They define a precise notion of "infinite communication complexity" and show that a set of infinite sequences is Borel (a well-behaved, constructively definable collection) if and only if the corresponding infinite communication game can be "solved" by the two players working together.

Borel sets are a central concept in descriptive set theory, a branch of mathematics that studies how complicated sets of real numbers or similar spaces can be. Classically, proving theorems about these sets requires sophisticated abstract machinery. By translating the questions into game-theoretic and combinatorial language, the authors are able to give new, more elementary proofs of several important classical results. These include the analytic separation theorem, which says that two disjoint "nicely definable" sets can always be separated by a Borel set, and a theorem connecting two different ways of defining Borel sets using monotone and positive conditions.

As a further application, the authors describe when one set can be separated from another using a Borel set, framing the answer in terms of who wins a simple "cut-and-choose" game, where one player proposes a division and the other picks a side. Using this framework, they give clean combinatorial arguments for why certain natural sets are not Borel, including the collection of infinite trees that have no infinite branch and any infinite generalization of the parity function from circuit complexity. The overall contribution bridges ideas from complexity theory, logic, and set theory in a way that makes results from each field more accessible through the language of the others.

Read original →