Also see Interpreting the learning of deceit for another proposal/​research agenda to deal with this threat model.
On a quick skim, I think this makes additional assumptions that seem pretty uncertain.
On a quick skim, I think this makes additional assumptions that seem pretty uncertain.