In the absence of that, at least aim to "first, do no harm" — which means don't propose an intervention at all unless you're confident you know the first, second, and third order effects over a long time scale across all impacted groups.

→ View original post on X — @jeremyphoward