The negative thesis of this paper states that the free energy principle’s explanatory power, unlike its unificatory power, can be regarded as weak, since it does not fulfil the 3M requirement posited by mechanists. This follows from the fact that the free energy principle is a functional theory, thus also employing functional concepts. Yet these do not explain how a given phenomenon in fact does work but only how it should work. However, Piccinini & Craver (2011) propose that functional analyses, ultimately, are nothing else but mechanism sketches, i.e., incomplete mechanistic explanations.
In this paper I have tried to make a positive contribution to the discussion by arguing for two claims: first, since the free energy principle incorporates empirical results from psychology and neuroscience it provides a stronger case of unification (SUT) than the unification provided by Bayesian theory alone. By not solely relying on its mathematical foundation, the free energy principle can try to ground its findings empirically in the brain. As a result, both the free energy principle and theories from psychology and neuroscience can constrain each other, thus being beneficiary to one another. Second, I argue that the free energy principle can act as a guide to finding multilevel mechanistic explanations. By linking mechanistic concepts with functional concepts from the free energy principle, the 3M requirement posited by mechanists can be fulfilled, consequently leading to actual explanations. This relates to the accused preposterousness of the free energy principle: with increasing explanatory power it becomes more and more difficult to deny that the free energy principle itself is, in fact, true.