In this post I’ll describe something that’s kind of common knowledge among applied category theorists, that when you describe behaviour via a (pseudo)functor to the category of relations, emergent effects are described by the failure of that functor to be strong, ie. to be an actual functor. This is more or less in Seven Sketches (Fong and Spivak say “generative effects”, which as far as I can tell is an exact synonym for emergent effects), but I’ll write it in one place and in my own words.

Suppose we have a domain-specific category . It’s hopefully monoidal and most likely a hypergraph category, but the basic idea of what I’m saying applies just to the category structure. The morphisms of are open systems of some sort, the objects describe the open boundaries that a system can have, and composition describes coupling two systems together along a common boundary. The standard sledgehammer for building categories like this is decorated cospans, with structured cospans as a closely related new alternative.