Abstract
A new theory of the mind, the predictive processing model, is ascendant in recent work in cognitive science. According to this theory, all the mind ever fundamentally does is make hypotheses about the environment, generate prediction-errors by comparing its predictions with its sensory data, and use these prediction-errors to update its representation of the world. The theory of motivation and action to which the predictive processing model is committed has been the subject of lively debate in the literature. However, the upshots of the predictive processing model for the theory of practical reason have received relatively scant attention. Here I investigate how proponents of predictive processing should conceive of practical reason. The predictive processing model has it that the practical attitudes – desires and intentions etc. – reduce to certain theoretical attitudes – in particular, to beliefs and predictions. I argue that this psychological reduction precipitates a broader normative reduction, in which practical reasoning reduces to a kind of theoretical reasoning about what you will do, and in which practical normativity turns out to really be a species of epistemic normativity. In other words, that practical reason reduces to theoretical reason. If I am correct, the predictive processing model of the mind is committed to – what I am dubbing – a ‘radical cognitivism about practical reason’.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Where ‘the world’ is understood to include, not only the external environment, but also the subject’s internal states (including state of their body and mind).
Those familiar with the predictive coding literature should note that I am, following Hohwy (2013) and Clark (2015), assuming here a cognitivist and/or representationalist interpretation of the predictive coding model. This cognitivist/representationalist reading is questioned, and alternative non-cognitivist or non-representationalist interpretations of predictive coding are discussed, in – for example – Kirchhoff and Robertson (2018) and Downey (2018). The interested reader should consult these references for further discussion. I cannot defend the cognitivist/representationalist reading here. Rather, I shall simply be assuming it. Furthermore, it should be noted that the ‘Bayesian beliefs’ posited by the predictive processing model of the mind, even on the representationalist reading, are likely quite dissimilar to ordinary beliefs, since they are probability distributions of the expected causes of sensory states (Gładziejewski, 2016; Williams, 2017). According to the representationalist consensus, such beliefs are likely not propositional, admit degree, and may not be linguistically expressible.
Here I will follow previous commentators from the literature in speaking of the mind (or brain) as possessing, on the predictive processing model, a ‘drive’ or ‘goal’ or ‘aim’ or ‘imperative’ of minimizing prediction-error on its model of the world (Friston, 2005). To avoid any possible misinterpretations, I should be clear that I do not take this talk to be ontologically committing in any particular way. Indeed, in this paper, I will be neutral with respect to the nature of the ‘ontological correlates’, so to speak, of this talk. So, for example, I should not be construed as asserting that such a goal or aim of minimizing prediction-error has any psychological reality – as, say, Sun and Firestone (2020) seem to presuppose. Rather, it is consistent with everything I have to say here that talk of a ‘drive to minimize prediction-error’ is simply a useful facon de parler that serves as a placeholder for some more cumbersome assertion, such as the statement that ‘the mind implements a hierarchical prediction-error minimization regime on its total cognitive representation of the world’ or even the statement that ‘neural processes may be modelled as unfolding in ways that minimize prediction-error in the long run’.
Some evidence of the explanatory ambitions of proponents of the predictive processing model come from the following quotes: ‘…This mechanism is meant to explain perception and action and everything mental in between…’ (Hohwy, 2013); ‘…the first truly unifying account of perception, cognition, and action…’ (Clark, 2015).
By ‘predictive processor’ I mean advocates of the predictive processing model.
The mind implementing a hierarchical prediction-error minimization regime on its total cognitive representation of the world is hypothesized to be a way in which the mind can engage in approximate Bayesian inference concerning the causes of its sensory data. The interested reader should consult the prior literature – for example, Clark (2015) or Williams (2020) – for an explanation of the relationship between Bayesian inference and predictive processing.
Although the predictive processing model is typically framed as a theory of all mental processes, there have been few explicit attempts to show how the predictive processing model can account for cognition and conceptual thought. Indeed, Williams (2020) has argued that the predictive processing model cannot account for the generality or compositionality of cognition.
Precision is the measure of a variance of some set of signals. Roughly speaking, a set of signals is precise just when they have a low variance, or a low variance of values (Williams, 2020).
However, as I mentioned before, there is considerable skepticism over whether the predictive processing model can adequately account for cognition and conceptual thought – see, for example, William (2020).
The predictive processing model is held to provide an attractive model of various perceptual phenomenon (such as binocular rivalry, perceptual uncertainty, perceptual repetition etc.). Nevertheless, it looks to fail to make sense of certain other perceptual phenomena (see, for example, Block & Siegel (2013).
This feature of the predictive processing model may entail that thinkers consistently violate epistemic norms regarding appropriate responsiveness to evidence during the production of action. (See, for example, Bruineberg et al. (2018).
The reader may remain skeptical over the prospects for an adequate predictive processing theory of desire and motivation. The reader can consult, for example, Klein (forthcoming) for a sustained argument that the predictive processing model cannot adequately account in principle for the phenomenon of desire.
Clark, in a personal correspondence with the author, says that he recognizes ‘…a distinction between the practical and the theoretical…’. Indeed, in his (2015) Surfing Uncertainty, Clark develops a predictive processing approach that appeals to reinforcement learning and other deep learning models that posit sui generis value-functions. This all suggests that the result pursued here – namely, that proponents of predictive processing are committed to a reduction of the faculty of practical reason to a species of theoretical reason – is no triviality, but rather a substantive claim that is inconsistent with some presentations of predictive processing from the literature.
What kind of epistemic norms would practical norms reduce to for the radical cognitivist? Of course, it will depend on the practical norm in question. But let’s consider a few representative examples. (The interested reader should consult my ‘Radical Cognitivism about Practical Reason’ (forthcoming) and ‘Practical Reason as Theoretical Reason’ (MS) for a far more thorough discussion of the details of radical cognitivism). Consider the norm of means-end coherence governing intention – that is, the requirement to form the intention to take the necessary means to your intended ends. According to the radical cognitivist, your intentions to act are just certain predictions about what you are going to do. Given this, it makes sense that the practical norm of means-end coherence, on the radical cognitivist’s reduction, would be the epistemic norm of modus ponens – that is, the theoretical requirement to form the belief that q if you believe both that p and that if p then q. Consider also the practical norm against intending to perform actions that you believe to be incompatible. The radical cognitivist can, very plausibly, hold that this norm reduces to, or is to be explained in terms of, the epistemic norm against believing contradictory propositions (Ross, 2009). Lastly, consider the reasons for action governing the formation of your intentions to act. As we just saw, according to the radical cognitivist, your intentions are just certain predictions about what you are going to do. And your reasons for action, by their nature, are just those considerations that count in favor of your forming an intention to act. Consequently, the radical cognitivist will hold that your reasons for action reduce to whatever considerations count in favor of the formation of those mental states – namely, certain predictions about what you are going to do – to which your intentions to act reduce on the radical cognitivist’s scheme. What (epistemic) considerations count in favor of your forming predictions about what you are going to do? Well, epistemic reasons to believe (or evidence) that you will act in some certain way, of course. These, for the radical cognitivist, are the consideration to which your reasons for action reduce. (Again, for a far more detailed discussion, the interested reader should consult my ‘Radical Cognitivism about Practical Reason’ (forthcoming) and ‘Practical Reason as Theoretical Reason’ (MS).
As I explain in my papers ‘Radical Cognitivism about Practical Reason’ (forthcoming) and ‘Practical Reason as Theoretical Reason’ (MS), radical cognitivism about practical reason has a number of important theoretical virtues. Most pertinently, it brings us parsimony, both in our psychological theory and our normative theory. After all, why posit two fundamental faculties of reason – one theoretical in nature, the other practical – when we can make do with the resources entailed by just one of them? This reduction unifies and streamlines our theory of the mind: it promises to explain behavior through appeal to just one kind of mental state, theoretical attitudes, animated by one aim (to believe the truth), rather than by reference to a plurality of such states, regulated by different aims. And our normative theory is likewise unified and economized with no loss of explanatory power. In particular, your reasons for action turn out to be a variety of evidence regarding your future actions, entities already posited by our normative theory. Secondly, as I attempt to demonstrate in these other papers, it looks to vindicate (limited) forms of moral rationalism and prudential rationalism – the doctrines, respectively, that we have (some) reasons to be moral and prudent that are independent of our desires (Shafer-Landau, 2003). And it does this in a novel way – quite unlike Kantian or Realist strategies for defending these conclusions (Korsgaard, 1996).
The Humean theory of reasons has it that reasons are desire-dependent. For Humeans, S has a reason to F just when, and because, S has a desire that would be satisfied by her F-ing, or would have such a desire if she soundly deliberated from her already existing set of motivations (Williams, 1979; Schroeder, 2007).
Anti-Humeans about reasons deny that reasons are, by their nature, desire-dependent. For the anti-Humean, an agent S can have a reason to F even though she lacks any desire that would be satisfied (even partially) by her doing F and even if she could not acquire that desire through sound deliberation from her pre-existing motivational set (Parfit, 2011). The key difference with the anti-Humean normative functionalist analysis of reasons is this: for the anti-Humean, if S has a reason to F, and S is aware of this fact, then S is rationally required to form a desire to F if she doesn’t already hold such a desire and even if she could not acquire that desire through sound deliberation from her prior motivational set.
Of course, many philosophers hold that normative reasons are metaphysically fundamental and cannot be defined in prior terms (Enoch, 2011; Scanlon, 2013). Such thinkers will therefore reject both normative functionalism and my argument developed here that is based upon it. However, my second argument that the predictive processing model entails a radical cognitivism about practical reason, the argument from parsimony, that I develop next, works best when assuming that normative properties are metaphysically fundamental.
By ‘a subjective epistemic reason’, I mean an epistemic reason by your lights. In other words, that it appears to you that you have an epistemic reason to believe so-and-so. I could have spoken here instead of ‘being required by epistemic rationality’: you are required by epistemic rationality, if you believe that you will keep yourself alive and that you can only keep your alive by eating food, to believe you will eat food. But I don’t take this to be a particular significant difference. Indeed, Kolodny (2005) analyzes the normativity of such rational requirements in terms of the normativity of it appearing to you that you have a reason to form such-and-such an attitude.
Other than – perhaps – the desire to minimize prediction-error, for those predictive processors who hold that our talk of the mind being animated by a goal/aim/imperative/drive/desire to minimize prediction-error should be taken literally and as corresponding to an actual desire – for example: Sun and Firestone (2020).
References
Adler, J. (2002). Belief’s own Ethics. MIT Press.
Botvinick, M., & Toussaint, M. (2012). Planning as inference. Trends in Cognitive Science, 16(10), 485–488.
Buckley, C., Chang, S., McGregor, S., & Seth, A. (2017). The free energy principle for action and perception: A mathematical review. Journal of Mathematical Psychology, 81, 55–79.
Clark, A. (2013a). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral and Brain Sciences, 36, 3: 181–204.
Clark, A. (2013b). Expecting the World: Perception, Prediction, and the origin of human knowledge. The Journal of Philosophy, 15, 9: 469–496.
Clark, A. (2015). Surfing uncertainty: Prediction, Action, and the embodied mind. Oxford University Press.
Clark, A. (2019). Beyond Desire? Agency, Choice, and the predictive mind. Australasian Journal of Philosophy. 1–15.
Dennett, D. (1991). Real patterns. Journal of Philosophy, 87, 27–51.
Downey, A. (2018). Predictive Processing and the representation wars: A victory for the Eliminativist (via Fictionalism) (pp. 5115–5139). Synthese.
Friston, K. (2005). A theory of cortical responses. Philos Trans R Soc Lond B Biol Sci Apr, 29(1456), 815–836.
Friston, K. (2011). What is Optimal About Motor Control? Neuron 72(3): 488–498.
Friston, K. (2012). Free-Energy Minimization and the Dark-Room Problem. Frontiers in Psychology, 2012, 3: 130.
Friston, K. (2013). Active inference and Free Energy: Commentary on Andy Clark’s ‘Predictive brains, situated Agents, and the future of Cognitive Science’. Behavioral and Brain Sciences, 36, 3: 212–213.
Friston, K., & Stephan, K. (2007). Free energy and the brain. Synthese, 159, 417–458.
Hieronymi, P. (2005). The wrong Kind of reason. Journal of Philosophy, 102(9), 437–457.
Hohwy, J. (2013). The predictive mind. Oxford University Press.
Jackson, F. (1998). From Metaphysics to Ethics: A defence of conceptual analysis. Clarendon Press.
Jackson, F., & Pettit, P. (1995). Moral Functionalism and Moral Motivation. The Philosophical Quarterly, 45(178).
Kirchhoff, M., & Robertson, I. (2018). Enactivism and Predictive Processing: A non-representational view. Philosophical Explorations, 21(2).
Klein, C. (2018). What do Predictive Processors Want? Synthese, 95(6): 2451–2557.
Klein, C., Forthcoming. A Humean Challenge to Predictive Coding. In The Philosophy and Science of Predictive Processing (eds. Steven Gouveia, Dina Mendonca, & Manuel Curado), Bloomsbury Press.
Korsgaard, C. (1996). The sources of Normativity. Cambridge University Press.
Leary, S. (2017). In defense of practical reasons for belief. Australasian Journal of Philosophy, 95(3), 529–542.
Parfit, D. (2011). On what matters, volume two. Oxford University Press.
Ratoff, W. Forthcoming. Radical cognitivism about practical reason. Journal of Ethics and Social Philosophy.
Ratoff, W., & Manuscript. Practical Reason as Theoretical Reason.
Ross, J. (2009). How to be a cognitivist about practical reason. Oxford Studies in Meta-Ethics, 4, 243–281.
Scanlon, T. M. (2014). Being realistic about reasons. Oxford University Press.
Schroeder, M. (2007). Slaves of the Passions. Oxford University Press.
Setiya, K. (2007). Reasons without rationalism. Princeton University Press.
Shafer-Landau, R. (2003). Moral Realism: A Defense. Oxford University Press.
Shah, N. (2006). A New Argument for Evidentialism. Philosophical Quarterly, 56(225), 481–498.
Solway, A., & Botvinick, M. (2012). Goal-directed decision making as probabilistic inference: A computational framework and potential neural correlates. Psychological Review, 119(1), 120–154.
Sun, Z., & Firestone, C. (2020). The Dark Room Problem. Trends in Cognitive Science, 24, 346–348.
Van de Cruys, S., Friston, K., & Clark, A. (2020). Controlled optimism: Reply to Sun and Firestone on the Dark Room Problem. Trends in Cognitive Sciences, 24(9), 680–681.
Velleman, J. D. (1985). Practical reflection. The Philosophical Review, 94(1), 33–61.
Velleman, J. D. (1989). Practical reflection. Chicago University Press.
Wallace, J. (2001). Normativity, Commitment, and instrumental reason. Philosophers Imprint, 1, 1–26.
Wallace, D. (2003). Everett and structure. Studies in History and Philosophy of Modern Physics, 34, 87–105.
Way, J. (2016). Two arguments for Evidentialism. Philosophical Quarterly, 66(265), 805–818.
Wiese, W. (2017). Action is enabled by systematic misrepresentations. Erkenntnis, 82, 1233–1252.
Williams, B. (1979). Internal and external reasons. In H. Ross (Ed.), Rational action (pp. 101–113). Cambridge University Press.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflicts of interest
The author reports that there are no conflicts of interest associated with this paper.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Ratoff, W. How should predictive processors conceive of practical reason?. Synthese 202, 140 (2023). https://doi.org/10.1007/s11229-023-04354-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s11229-023-04354-5