Replying to @jw_lockhart and @msalganik
Protecting recommendations from being gamed is definitely useful, but the context here is thinking about how and whether to regulate them. The threat model is more around destroying democracy, radicalisation, attentional addiction, etc.