[1]

Some challenging questions surface here as to whether prediction errors are used to update priors, which corresponds to standard Bayesian inference, or whether they are used to update the underlying generative/predictive model, which corresponds to learning.

[2]
This underlines the close links between cybernetics and behaviourism. Perhaps this explains why cybernetics was so reluctant to bring phenomenology into its remit, an exclusion which, looking back, seems like a missed opportunity.
[3]
Allostasis: the process of achieving homeostasis.
[4]
It is interesting to consider possible dysfunctions in this process. For example, if high-level predictions about the persistence of low blood sugar become abnormally strong (i.e., low blood sugar becomes chronically expected), allostatic food-seeking behaviours may not occur. This process, akin to the transition from hallucination to delusion in perceptual inference (Fletcher & Frith 2009), may help understand eating disorders in terms of dysfunctional signalling of satiety.
[5]
Interestingly the expectation of perceptual correlations seems to be sufficient for inducing the rubber hand illusion (Ferri et al. 2013).
[6]
These are long-range projection neurons found selectively in hominid primates and certain other species.
[7]
Evolutionary robotics involves the use of population-based search procedures (genetic algorithms) to automatically specify control architectures (and/or morphologies) of mobile robots. For an excellent introduction see (Bongard 2013).
[8]
Videos showing the evolution of both gait and self-model are available from http://creativemachines.cornell.edu/emergent_self_models
[9]
At a recent symposium of the AISB society that focused on SMC theory, it was stated that “the main question is how to get the brain into view from an enactive/sensorimotor perspective. […] Addressing this question is urgently needed, for there seem to be no accepted alternatives to representational interpretations of the inner processes” (O'Regan & Dagenaar 2014).
[10]
See Beaton (2013) for a distinct approach to incorporating counterfactual ideas in SMC theory. Beaton’s approach remains squarely within the strongly enactivist tradition.
[11]
There are also some challenges lying in wait here. For instance, it is not immediately clear how important assumptions like the Laplace approximation can generalize to the multimodal probability distributions entailed by counterfactual PP (Otworowska et al. 2014).
[12]
There is a more dramatic conflict with “radical” versions of enactivism, in which mental processes, and in some cases even their material substrates, are allowed to extend beyond the confines of the skull (Hutto & Myin 2013).
[13]
Presence may also depend on the hierarchical depth of predictive models inasmuch as this reflects object-related invariances in perception. For further discussion see commentaries and response to (Seth 2014b).
[14]
Thanks to my reviewers for raising this provocative point.
[15]
Chris Thornton came up with this term (personal communication).
[16]
The term “dark room problem” comes from the idea that a free-energy-minimizing (or surprise-avoiding) agent could minimize prediction error just by finding an environment that lacks sensory stimulation (a “dark room”) and staying there.