When Can Predictive Brains be Truly Bayesian?

tive coding framework does not yet make stringent commitments as to the nature of the causal models that the brain can represent. Hence, contrary to suggestions by Clark (in press), the framework does not yet have the virtue that it effectively implements tractable Bayesian inference. At this point in time three mutually exclusive options remain open: either predictive coding does not implement Bayesian inference, or predictive coding is not tractable, or the theory of hierarchical predictive coding is enriched by specific assumptions about the structure of the brain’s causal models. Assuming that one is committed to the Bayesian Brain Hypothesis, the first two options are out and the third is the only one remaining. Formal analyses expanding on this option are beyond the scope of this commentary (see e.g., Blokpoel et al., 2010; van Rooij et al., 2011), but Table 1 qualitatively sketches the space of causal models that could (or could not) yield tractable Bayesian cause estimation. We will discuss the viability of the options in more detail below. To start, causal models could be assumed to be quite simple, e.g., having high degrees of statistical independencies of variables. In this case, it may be that heuristic methods, such as those based on gradient ascent (Friston, 2002, p. 13) or a Kalman filter (Rao It is thus a major virtue of the hierarchical predictive coding account that it effectively implements a computationally tractable version of the so-called Bayesian Brain Hypothesis. (Clark, in press)