What active inference still can't do: the (frame) problem that just won’t go away

The frame problem, or problem of relevance, concerns the capacity of cognitive agents to zero in on relevant information during action and perception, whilst intelligently ignoring everything else. Although this is an ability that such agents realise even in the most seemingly novel of situations, it is generally accepted that no comprehensive explanatory account of it has been provided by cognitive-scientific researchers. However, a new account deriving from the popular active inference framework purports to solve the problem of relevance, an achievement which, if realised, would give strong evidence for the framework’s claim to be an all-embracing theory of cognition. Unfortunately, this account, alongside previous active inference attempts to respond to the problem of relevance, is explanatorily inadequate. This means that, as is the case for all other frameworks in cognitive science, active inference has heretofore failed to resolve the problem of relevance.