@darius simple version: if Alice trusts Bob, Bob tells Alice that he trusts Charlie, then Alice can trust Charlie. If Bob tells Alice he distrusts Eve, then Alice knows not to trust Eve either.
In a more complex setup you could let Alice say she has 40% trust in Charlie and 80% trust in Dave, so if Charlie trusts Eve but Dave says Eve is dangerous, then Alice would distrust Eve because she trusts Dave's judgment more than Charlie's.
Many Mastodon instances already share what instances they limit or suspend...if that was accessible through an API then creation of limit/suspend lists could be automated for instance admins by choosing a few starting instances that they trust; and new instances could get "verified" by only needing one or a few known instances agree to trust them, so long as they did not act poorly enough to have that trust overruled by others.
@admin ah, yes, I was a (distant) advisor on a master's thesis that was similar to this
https://cblgh.org/trustnet/