I like the idea of building a reputation system that accepts bots as first class citizens instead of imagining them as an inherent problem. Anyone who meaningfully contributes is good even if they are a bot, and anyone who doesn’t gets left behind. (Easier said than done but the approach to the problem is what’s fundamentally different.)

Reputation systems, if they try too hard, quickly reach a point where ambitious fakes are more successful at convincing the system than casual real users.

Case in point: Google visibility, if you are not Wikipedia no amount of quality content will get you above the link farms if you don't also follow the customary rites of pleasing the algorithm.

reddit (or some reddit users, it's hard to know) does this. There's a bot crawling around and whenever someone replies to a bot saying "good bot" or "bad bot" it updates a ranking.

(there's always a relevant xkcd: https://xkcd.com/810/)

Bots can definitely contribute meaningfully in many ways. Mostly through organization (suggesting labels & other metadata, album creation, etc), spam filtering & rule enforcement (like reddit's automoderator), and even automatic editing (like some of the stuff Google Photos offers, like rotation suggestions and various automatic filters).

Custom moderator automation would be a very valuable tool, bots are terrible at beating bespoke solutions. Is there some open source Amazon lambda clone that community platforms could easily make available to their power users? Funneling automaton onto a platform runtime could be extremely helpful in distinguishing between good and bad automation.

Think of the bots on Reddit