What are the advantages of having a stateless class injected via Dagger/Hilt which houses methods that operate on input data according to business logic, vs just using a file with top-level functions for this purpose? I have an app which recommends albums to listen to from your library based on your current mood. To do this recommendation, is there any downside to just passing the data set to pick from along with the mood parameters into top-level functions? Why use a class/object if it holds no state?

the big one is dependency inversion, so your "recommend" function isn't relied on directly by, say android specific code if you wanted to do something like KMM.

If using KMM (which I do plan to in the future), what is the issue with android-specific code relying on the top-level "recommend" function, rather than an interface/class? If the function's input and output are objects in my domain (platform-agnostic), then can't the android-specific code and iOS-specific code call it just as easily as if it were behind an interface?

yea you could, you're just depending on concretes at that point ( which isn't necessarily bad to do, if you want to, in general I try to not depend on concretes), with the interface it allows you to have that seam if you need it. really