Representing the Self in Predictive Processing

Elmarie Venter — PhD can­did­ate, Department of Philosophy, Ruhr-Universität Bochum

Who do you think you are? Or, less con­front­a­tion­ally, what ingredi­ents (e.g. memor­ies, beliefs, desires) go into the mod­el of your self? In this post, I explore dif­fer­ent con­cep­tions of how the self is rep­res­en­ted in the pre­dict­ive pro­cessing (PP) frame­work. At the core of PP is the notion that the brain is in the busi­ness of mak­ing pre­dic­tions about the world, and that the brain is primar­ily an organ that func­tions to min­im­ize pre­dic­tion error (i.e. the dif­fer­ence between pre­dic­tions about the state of the world and the observed state of the world) (Clark, 2017, p.727). Predictive pro­cessing neces­sit­ates mod­el­ing the causes of our sens­ory per­turb­a­tions and since agents them­selves are also such causes, a self-model is required under PP. The intern­al mod­els of the self will include “…rep­res­ent­a­tions of the agent’s own body and its tra­ject­or­ies and inter­ac­tions with oth­er causes in the world” (Hohwy & Michael, 2017, p.367).

In this post I will dis­cuss accounts of how the self is mod­elled under two PP camps: Conservative PP and Radical PP. Broadly speak­ing, Conservative PP holds that the mind is infer­en­tially secluded from the envir­on­ment — the body also forms part of the extern­al envir­on­ment. All pre­dic­tion error min­im­iz­a­tion occurs behind an ‘evid­en­tiary bound­ary’ which implies that the brain recon­structs the state of the world (Hohwhy, 2016, p.259). In con­trast, Radical PP holds that rep­res­ent­a­tions of the world are a mat­ter of embod­ied and embed­ded cog­ni­tion (Dolega, 2017, p.6). Perceiving my self, oth­er agents, and the world, is not a pro­cess of recon­struc­tion but rather a coupled pro­cess between per­cep­tion and action. How does the view of a self-model align with these ver­sions of pre­dict­ive pro­cessing? I will argue that Radical PP’s account of self-modelling is prefer­able because it avoids two key con­cerns that arise from Conservative PP’s mod­el­ing of the self.

On the side of Conservative PP, Hohwy & Michael (2017) con­ceive of the self-model as one that cap­tures “…rep­res­ent­a­tions of the agent’s own body…” as well as hid­den, endo­gen­ous causes, such as “…char­ac­ter traits, biases, reac­tion pat­terns, affec­tions, stand­ing beliefs, desires, inten­tions, base-level intern­al states, and so on” (Hohwy & Michael, 2017, p.369). On this view, the self is just anoth­er set of causes that is modeled in order to min­im­ize pre­dic­tion error. This view likens the mod­el of the self to mod­els of the envir­on­ment and oth­er people (and their men­tal states), and is in line with the Conservative PP account advoc­ated by Hohwy (2016) under which there is an ‘evid­en­tiary bound­ary’ between mind and world, behind which pre­dic­tion error min­im­iz­a­tion takes place. Any parts of our body “…that are not func­tion­ally sens­ory organs are bey­ond the bound­ary… [and are] just the kinds of states that should be modeled in intern­al, hier­arch­ic­al mod­els of a (pre­dic­tion error min­im­iz­a­tion) sys­tem.” (Hohwy, 2016, p.269).

As I see it, Conservative PP’s self-modeling (as described by Hohwy & Michael (2017)) is prob­lem­at­ic in two ways:

1) Our access to inform­a­tion about our own body is neg­lected by Conservative PP. Agents typ­ic­ally have access to cer­tain inform­a­tion about their body that is immune to error through misid­en­ti­fic­a­tion; this immunity does not extend to inform­a­tion about the world and oth­er agents.

2) Conservative PP ignores the marked dif­fer­ence in how we rep­res­ent ourselves and oth­er agents. Other agents can only enter our inten­tion­al states as part of the con­tent, where­as we ourselves can also enter our inten­tion­al states in anoth­er way.

In deal­ing with these con­cerns I pro­pose that the self is rep­res­en­ted along two dimen­sions: as-subject and as-object (a dis­tinc­tion that can be traced back to Wittgenstein’s Blue Book, and which can be fleshed out by appeal to debates on ref­er­ence and inten­tion­al­ity). The fun­da­ment­al idea here is that there is a cer­tain kind of error — in identi­fy­ing the per­son that some­thing is true of (e.g. a bod­ily pos­i­tion or a men­tal state) — that can occur when identi­fy­ing the self as-object which can­not occur in identi­fy­ing the self as-subject (Longuenesse, 2017, p.20; Evans 1982). Imagine that I per­ceive a cof­fee mug in front of me, and once I have seen it I reach out my hand to grasp the mug in order to drink from it. Now envi­sion a sim­il­ar situ­ation, in which I am act­ing like this while at the same time look­ing at myself in a mir­ror. In the lat­ter situ­ation I have two sources of inform­a­tion for obtain­ing know­ledge about myself grasp­ing the cup of cof­fee. One source of inform­a­tion is proprio­cept­ive and kin­es­thet­ic, and there­fore provides me with inform­a­tion about myself from the inside. The oth­er source of inform­a­tion is visu­al, and provides me with inform­a­tion from the out­side. The lat­ter source could provide me with inform­a­tion about the actions of oth­er agents as well, where­as the former can only be a source of inform­a­tion about my own self.

Since I am rep­res­en­ted in the con­tent of my visu­al exper­i­ence in the mir­ror scen­ario, I can mis­rep­res­ent myself as the inten­tion­al object of that very visu­al exper­i­ence. I could be mis­taken with respect to whom I am see­ing in the mir­ror grasp­ing the cof­fee mug; I may mis­takenly believe that I am in fact observing someone else grasp­ing the cup. No such mis­take is pos­sible in the con­trast case, in which I gain inform­a­tion about grasp­ing the mug from a proprio­cept­ive and kin­es­thet­ic source. A more rad­ic­al example of this dis­tinc­tion between self as-object and as-subject comes from indi­vidu­als with soma­to­pa­raphrenia. Such indi­vidu­als do not identi­fy some parts of their body as their own, e.g. they may believe that their arm belongs to someone else, but they are not mis­taken about who is identi­fy­ing their arm as belong­ing to someone else (Kang, 2016; Vallar & Ronchi, 2009). Recanati (2007, pp.147–148) spells out this dif­fer­ence by dis­tin­guish­ing between the con­tent and mode of an inten­tion­al state: “The con­tent is a rela­tiv­ized pro­pos­i­tion, true at a per­son, and the intern­al mode determ­ines the per­son rel­at­ive to which that rela­tiv­ized con­tent is eval­u­ated: myself”. With this dis­tinc­tion in mind, the prob­lems with Conservative PP becomes clear: the agent and their body are not rep­res­en­ted in the same way as any oth­er distal state in the world. Instead of the agent and their body only form­ing part of the con­tent of an inten­tion­al state (as Hohwy & Michael’s account would imply), they enter the state through the mode of per­cep­tion as well.

Clark (2017, p.729) provides an ana­logy that illus­trates the first prob­lem with self-modeling under Conservative PP: “The pre­dict­ing brain seems to be in some­what the same pre­dic­a­ment as the imprisoned agents in Plato’s “allegory of the cave”.” That is, under Conservative PP, distal states can only be inferred by the secluded brain, just as the pris­on­ers in the cave can only infer what the shad­ows on the walls are shad­ows of. The con­sequence of this is that we have no dir­ect (and, there­fore, error-immune) access to our own bod­ies. However, as has been illus­trated above, the self enters inten­tion­al states through mode (per­ceiv­ing, ima­gin­ing, remem­ber­ing, etc.) as well as con­tent, and this provides us with cer­tain inform­a­tion that is immune from error. In con­trast, Radical PP does not con­ceive of the body as a distal object. Instead, the agent’s body plays an act­ive role in determ­in­ing the sens­ory inform­a­tion that we have access to; it plays a fun­da­ment­al role in how we sample, and act in, the world. This act­ive role is such that cer­tain inform­a­tion is avail­able to us error free – even if I am mis­taken about anoth­er agent grasp­ing the cup, I can­not be mis­taken that it is me that is see­ing someone grasp the cup. In this sense, Radical PP provides us with a prefer­able story about how whole embod­ied agents are mod­els of the envir­on­ment and min­im­ize pre­dic­tion error through a vari­ety of adapt­ive strategies (Clark, 2017,  p.742).

 The two dimen­sions of self can also shed light on the second con­cern with Conservative PP because this dis­tinc­tion illus­trates how we per­ceive and inter­act with oth­er agents. As dis­cussed above, the self as-object enters inten­tion­al states as part of the con­tent, and the self as-subject enters such states through mode. The world, includ­ing oth­er agents and their men­tal states, only ever form part of the con­tent of our inten­tion­al states. Referring back to the example spelled out above: anoth­er agent can only ever play the same role in per­cep­tion as I do in the mir­ror case, i.e. as con­tent of the inten­tion­al struc­ture. I do not have access to oth­er agents “from the inside,” how­ever. For instance, I do not have the same access to the reas­ons behind oth­ers’ actions (are they grasp­ing the cup to drink from it, to clear it from the table, to see if there is still cof­fee in it?), nor do I have access to wheth­er the oth­er agent will suc­cess­fully grasp the mug (is their grip wide enough, do they have enough strength in their wrist?). There is thus a dimen­sion of the self to which one has priv­ileged access. We only have access to oth­er agents through per­cep­tu­al infer­ence (i.e. by observing their beha­vi­or and infer­ring its causes), where­as we have both per­cep­tu­al and act­ive infer­en­tial access to our own beha­viours. Though Conservative PP pro­ponents main­tain that the secluded brain only has per­cep­tu­al infer­en­tial access to our own body (Hohwy, 2016, p.276), there is some­thing markedly dif­fer­ent in what enables us to mod­el the causes of our own beha­vi­or and men­tal states to that of oth­er agents. I have proprio­cept­ive, kin­es­thet­ic, and intero­cept­ive access to inform­a­tion about myself; I only have extero­cept­ive inform­a­tion about oth­er agents.

 For Conservative PP, the body (and by exten­sion, the self) is just anoth­er object in the world that receives com­mands to act in ser­vice of pre­dic­tion error min­im­iz­a­tion. I have high­lighted two con­cerns about this view: the body is treated as a distal object, and the body (and self) placed on the same side of the evid­en­tiary bound­ary as oth­er agents. This means that the dimen­sion of self which is immune to error through misid­en­ti­fic­a­tion is not accom­mod­ated, and the marked dif­fer­ence in our access to inform­a­tion about our own states and those of oth­er agents is ignored. Radical PP, how­ever, avoids both con­cerns by tak­ing into account the two rep­res­ent­a­tion­al dimen­sions of the self and employ­ing an embod­ied approach to cog­ni­tion. The Radical PP account there­fore provides a more refined ver­sion of self-modeling. My beliefs, desires, and bod­ily shape can all be inferred in the mod­el of self-as-object, but self-as-subject cap­tures the part of the self that is not inferred: it con­tains inform­a­tion about me and my body from the inside, which is an essen­tial part of who we think we are.


Clark, A., 2017. Busting Out: Predictive Brains, Embodied Minds, and The Puzzle of The Evidentiary Veil. Noûs, 51(4): 727–753.

Dolega, K., 2017. Moderate Predictive Processing. In T. Metzinger & W. Wiese (Eds.) Philosophy and Predictive Processing. Frankfurt Am Main: MIND Group.

Evans, J., 1982. The Varieties of Reference. Oxford: Clarendon Press.

Friston, K. J. and Stephan, K. E., 2007. Free-energy and the Brain. Synthese, 159(3): 417–458.

Hohwy, J., 2016. The Self-Evidencing Brain. Noûs, 50(2): 259–285.

Hohwy, J. and Michael, J., 2017. Why Should Any Body Have A Self? In F. de Vignemont & A. Alsmith (Eds.) The Subject’s Matter: Self-Consciousness And The Body. Cambridge, Massachusetts: MIT Press.

Kang, S. P., 2016. Somatoparaphrenia, the Body Swap Illusion, and Immunity to Error through Misidentification. The Journal of Philosophy, 113(9): 463–471.

Longuenesse, B., 2017. I, Me, Mine: Back To Kant, And Back Again. New York: Oxford University Press.

Michael, J. and De Bruin, L., 2015. How Direct is Social Perception. Consciousness and Cognition, 36: 373–375.

Recanati, F., 2007. Perspectival Thought: A Plea For (Moderate) Relativism. Clarendon Press.

Thompson, E. and Varela, F. J., 2001. Radical Embodiment: Neural Dynamics and Consciousness. Trends in Cognitive Sciences, 5(10): 418–425.

Vallar, G. and Ronchi, R., 2009. Somatoparaphrenia: A Body Delusion. A Review of the Neuropsychological Literature. Experimental Brain Research, 192(3): 533–551.

Wittgenstein, L. 1960. Blue Book. Oxford: Blackwell.