Read on Twitter
In reaction to the sheer volume of bad papers I have had to review lately, I am starting to draft a new paper called "How not to write about the free-energy principle"
Some protips: 1. Don't discuss another formalism. Telling me about Bayes's rule doesn't qualify as a substitute for talking about nonequilibrium steady states, probability density dynamics, and so on. Saying that it's very difficult, mysterious, and technical doesn't excuse this
2. Write with people who are technically savvy. So much energy would be saved if non-technically savvy folks worked with people who are fluent in the maths. Showing your paper to friends in non-technical fields like philosophy isn't enough. Collaborate between sets of expertise!
3. FEP exists in the context of an overarching experimental framework. The origin of the free-energy construct is the evaluation of evidence for models of the causal structure of processes generating data. To ignore this is to ignore the methodology that underwrites the approach
4. Work with the constructs a bit before trying to get critical. If you've not been involved in modeling work, derivations of the equations for FEP models, or experimental work using FE-based metrics like SPM or DCM, there's good reason to think you may be missing methods/context
5. Falsificationism has no bite. The variational method operationalizes the quantification of evidence for alternative hypotheses. It is a mathematical machine for hypothesis testing. Most discussions of falsification miss this fact and so miss the entire point of the approach
6. If the maths look daunting it's because they are. The formalism in Friston's monograph reviews formulations from quantum, classical, and statistical mechanics before *building on* them. If you wouldn't be comfortable discussing the former you may not be equipped for the latter
7. Engage the formalism directly or don't write the paper. The papers I've reviewed almost all start by noting the maths are mysterious to the author; and then proceed to make a point that is actually defused by recent extensions of the formalism. This benefits nobody.
8. Ask yourself who your audience is. If you're writing for a non-technical audience, they will not benefit from a low quality reconstruction of complex maths using first-order logic. If you're writing for a technical audience, make sure they haven't already defused your point
