In the first year we just told people, which was obviously very stupid in hindsight.

In the second year, we had a survey where devs could say what their preference was: (a) team/specific devs (b) topic (c) none

That worked pretty well.