Skip to main content

Content and misrepresentation in hierarchical generative models

Abstract

In this paper, we consider how certain longstanding philosophical questions about mental representation may be answered on the assumption that cognitive and perceptual systems implement hierarchical generative models, such as those discussed within the prediction error minimization (PEM) framework. We build on existing treatments of representation via structural resemblance, such as those in Gładziejewski (Synthese 193(2):559–582, 2016) and Gładziejewski and Miłkowski (Biol Philos, 2017), to argue for a representationalist interpretation of the PEM framework. We further motivate the proposed approach to content by arguing that it is consistent with approaches implicit in theories of unsupervised learning in neural networks. In the course of this discussion, we argue that the structural representation proposal, properly understood, has more in common with functional-role than with causal/informational or teleosemantic theories. In the remainder of the paper, we describe the PEM framework for approximate Bayesian inference in some detail, and discuss how structural representations might arise within the proposed Bayesian hierarchies. After explicating the notion of variational inference, we define a subjectively accessible measure of misrepresentation for hierarchical Bayesian networks by appeal to the Kullbach–Leibler divergence between posterior generative and approximate recognition densities, and discuss a related measure of objective misrepresentation in terms of correspondence with the facts.

This is a preview of subscription content, access via your institution.

Notes

  1. We also here include Fodor’s (1990) solution to the disjunction problem, based on asymmetries among nomic relations.

  2. Of course, we do not mean to rule out that parts may also be structured and function themselves as structural representations. Indeed, this is likely the case in hierarchically organized systems like those considered in this paper, but we lack space to consider this issue here.

  3. This way of putting things may go to the heart of some of the more radical formulations of the implications of PEM-style accounts for the nature of the mind-world relation—for example, claims that perception is “controlled hallucination” (Grush 2004). Similarly, Geoff Hinton (one of the originators of contemporary models of perceptual inference involving generative models) claims in essence that the contents of mental states are hypothetical worlds (Hinton 2005).

  4. There may be ways of resisting the conclusion in the case of misrepresentation by distinguishing types of misrepresentation, as discussed in Sect. 5 below. However, the point seems difficult to sidestep with respect to imagination.

  5. It should be noted that there are ‘wide’ versions of functional role semantics as well; see Harman (1973).

  6. We thank an anonymous reviewer for an earlier version of this paper for pressing this crucial point, as well as the issue concerning internalism just considered.

  7. This is a decidedly Kantian reading of these ideas, but we believe it would be more procrustean to attempt to defend the opposite view according to which all representation in imagination is really somehow representation of one’s physical environment. This is true even though actual sources of sensory input play an indispensable explanatory role within the PEM framework, and in fact are etiologically necessary to get any kind of representation off the ground.

  8. O’Brien and Opie (2004) distinguish strictly between functional role semantics and their preferred version of structural representation theory on the grounds that the former appeals to causal relations among vehicles while the latter appeals to physical relations. It is not obvious, however, why the latter category should preclude the former.

  9. Gładziejewski and Miłkowski (2017) draw a similar conclusion for different reasons.

  10. It is sometimes claimed that this notion of representation is also too liberal. The “exploitability” constraint mentioned earlier goes some way toward mitigating this. Also note that interesting, human-like cases at least are hard to come by: a system must, as a matter of empirical fact, be quite complex before it is able to structurally represent deeply hidden environmental causes.

  11. There is thus reason to think that perceptual systems employ generative models based on considerations about learning alone, in addition to the considerations about contextually biased interpretation of stimuli mediated by extra-classical receptive field effects.

  12. Though of course the system need not begin by representing such distinctions, for reasons discussed in Sect. 2.3.

  13. It should be stressed that the construction of the reliable information channel and the development of meaningful representations are constitutively related. These are two ways of describing the process whereby causal structure (which representation exploits) is set up within cortical hierarchies.

  14. We thank an anonymous reviewer for pressing us to clarify this distinction, which is also drawn by Gładziejewski (2016), who uses the “X”-on-a-map example referred to here.

  15. Clearly, if each generative model represents only the hypothetical world whose causal structure is isomorphic to it, there can be no misrepresentation, and thus arguably no genuine representation either. We need an independent standard of comparison to define misrepresentation, but note that this target need not be the actual world: it could be one specified by a fictional description, for example.

  16. In Friston’s model (2005), only the top-down connections introduce nonlinearities, but this is just a further way in which the two models diverge while still sharing structure. The whole point, of course, is that the recognition model is a (to some degree crude) approximation of the posterior under the generative model.

  17. The distribution c of course specifies the relevant natural scene statistics. NSS are therefore important in principle to understanding Bayesian models of perception, as Orlandi claims (see discussion in Sect. 2.3), even if such models are interpreted in representational terms.

References

  • Adams, R. A., Huys, Q. J. M., & Roiser, J. P. (2015). Computational psychiatry: Towards a mathematically informed understanding of mental illness. Journal of Neurology, Neurosurgery & Psychiatry, 87(1), 53–63.

    Google Scholar 

  • Allen, M., & Friston, K.J. (2016). From cognitivism to autopoiesis: towards a computational framework for the embodied mind. Synthese. doi:10.1007/s11229-016-1288-5.

  • Anderson, M., & Chemero, A. (2013). The problem with brain GUTs: Conflation of different senses of ‘prediction’ threatens metaphysical disaster. Behavioral & Brain Sciences, 36, 204–205.

    Article  Google Scholar 

  • Apps, M. A. J., & Tsakiris, M. (2014). The free-energy self: A predictive coding account of self-recognition. Neuroscience & Biobehavioral Reviews, 41, 85–97.

    Article  Google Scholar 

  • Barrett, L. F. (2016). The theory of constructed emotion: An active inference account of interoception and categorization. Social Cognitive and Affective Neuroscience, 12, 1.

    Google Scholar 

  • Bastos, A. M., Usrey, W. M., Adams, R. A., Mangun, G. R., Fries, P., & Friston, K. J. (2012). Canonical microcircuits for predictive coding. Neuron, 76(4), 695–711.

    Article  Google Scholar 

  • Bengio, Y. (2009). Learning deep architectures for AI. Foundations and Trends in Machine Learning, 1(2), 1–127.

    Article  Google Scholar 

  • Bengio, Y., Courville, A. C., & Vincent, P. (2012). Unsupervised feature learning and deep learning: A review and new perspectives. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(8), 1798–1828.

    Article  Google Scholar 

  • Bishop, C. M. (2007). Pattern recognition and machine learning. Cordrecht: Springer.

    Google Scholar 

  • Block, N. (1994). Advertisement for a semantics for psychology. In S. P. Stich & T. Warfield (Eds.), Mental representation: A reader. Oxford: Blackwell.

    Google Scholar 

  • Bogacz, R. (2017). A tutorial on the free-energy framework for modelling perception and learning. Journal of Mathematical Psychology, 76(Part B), 198–211.

    Article  Google Scholar 

  • Brandom, R. (1994). Making it explicit. Cambridge: Harvard University Press.

    Google Scholar 

  • Bruineberg, J. (2016). The anticipating brain is not a scientist: The free-energy principle from an ecological-enactive perspective. Synthese. doi:10.1007/s11229-016-1239-1.

  • Burr, C., & Jones, M. (2016). The body as laboratory: Prediction-error minimization, embodiment, and representation. Philosophical Psychology, 29(4), 586–600.

    Article  Google Scholar 

  • Carreira-Perpiñán, M. A., & Hinton, G. E. (2005). On contrastive divergence learning. In Proceedings of the tenth international workshop on artificial intelligence and statistics.

  • Clark, A. (2012). Dreaming the whole cat: Generative models, predictive processing, and the enactivist conception of perceptual experience. Mind, 121(483), 753–771.

    Article  Google Scholar 

  • Clark, A. (2013). Whatever next? Predictive brains, situated agents, and the future of cognitive science. Behavioral & Brain Sciences, 36(3), 181–204.

    Article  Google Scholar 

  • Clark, A. (2016). Surfing uncertainty: Prediction, action, and the embodied mind. New York: Oxford University Press.

    Book  Google Scholar 

  • Colombo, M., & Seriés, P. (2012). Bayes in the brain–On Bayesian modelling in neuroscience. The British Journal for the Philosophy of Science, 63, 697–723.

    Article  Google Scholar 

  • Colombo, M., & Wright, C. (2017). Explanatory pluralism: An unrewarding prediction error for free energy theorists. Brain and Cognition, 112, 3–12.

    Article  Google Scholar 

  • Corlett, P. R., & Fletcher, P. C. (2012). The neurobiology of schizotypy: Fronto-striatal prediction error signal correlates with delusion-like beliefs in healthy people. Neuropsychologia, 50(14), 3612–3620.

    Article  Google Scholar 

  • Cummins, R. (1994). Interpretational semantics. In S. Stich & T. Warfield (Eds.), Mental representation: A reader. Oxford: Blackwell.

    Google Scholar 

  • Dayan, P., & Abbott, L. F. (2001). Theoretical neuroscience. Cambridge, Mass: MIT Press.

    Google Scholar 

  • Dayan, P., Hinton, G. E., Neal, R. M., & Zemel, R. S. (1995). The Helmholtz machine. Neural Computation, 7(5), 889–904.

    Article  Google Scholar 

  • Dretske, F. (1981). Knowledge and the flow of information. Cambridge, MA: MIT Press.

    Google Scholar 

  • Eliasmith, C. (2000). How neurons mean: A neurocomputational theory of representational content. Ph.D., Washington University in St.Louis.

  • Farennikova, A. (2014). Perception of absence and penetration from expectation. Review of Philosophy and Psychology, 6, 1–20.

    Google Scholar 

  • Field, H. (1977). Logic, meaning and conceptual role. Journal of Philosophy, 74(69), 379–409.

    Google Scholar 

  • Fletcher, P. C., & Frith, C. D. (2009). Perceiving is believing: A Bayesian approach to explaining the positive symptoms of schizophrenia. Nature Reviews Neuroscience, 10(1), 48–58.

    Article  Google Scholar 

  • Fodor, J. A. (1975). The language of thought. Cambridge: Harvard University Press.

    Google Scholar 

  • Fodor, J. A. (1990). A theory of content and other essays. Cambridge, Mass: MIT Press.

    Google Scholar 

  • Friston, K. J. (2005). A theory of cortical responses. Philosophical Transactions: Biological Sciences, 369(1456), 815–836.

    Article  Google Scholar 

  • Friston, K. (2010). The free-energy principle: A unified brain theory? Nature Reviews Neuroscience, 11(2), 127–138.

    Article  Google Scholar 

  • Friston, K. (2013). Life as we know it. Journal of The Royal Society Interface, 10(86), 1–12.

    Article  Google Scholar 

  • Friston, K., Levin, M., Sengupta, B., & Pezzulo, G. (2015). Knowing one’s place: A free-energy approach to pattern regulation. Journal of The Royal Society Interface, 12(105), 20141383.

    Article  Google Scholar 

  • Gallagher, B. (2006). Matching structure and semantics: A survey on graph-based pattern matching. In AAAI fall symposium on capturing and using patterns for evidence detection (pp. 45–53). American Association for Artificial Intelligence.

  • Gallagher, S., & Allen, M. (2016). Active inference, enactivism and the hermeneutics of social cognition. Synthese. doi:10.1007/s11229-016-1269-8.

  • Gładziejewski, P. (2016). Predictive coding and representationalism. Synthese, 193(2), 559–582.

    Article  Google Scholar 

  • Gładziejewski, P., & Miłkowski, M. (2017). Structural representations: Causally relevant and different from detectors. Biology and Philosophy. doi:10.1007/s10539-017-9562-6.

  • Godfrey-Smith, P. (1996). Complexity and the function of mind in nature. Cambridge: Cambridge University Press.

    Book  Google Scholar 

  • Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep learning. Cambridge, MA: MIT Press.

  • Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., et al. (2014). Generative adversarial nets. In Z. Ghahramani, M. Welling, C. Cortes, N. D. Lawrence & K. Q. Weinberger (Eds.), Advances in neural information processing systems (Vol. 27, pp. 2672–2680).

  • Grush, R. (2004). The emulation theory of representation: Motor control, imagery, and perception. Behavioral and Brain Sciences, 27, 377–442.

    Google Scholar 

  • Harman, G. (1973). Thought. Princeton: Princeton University Press.

    Google Scholar 

  • Harman, G. (1999). Reasoning, meaning and mind. Oxford: Oxford University Press.

    Book  Google Scholar 

  • Harrison, L. M., Stephan, K. E., Rees, G., & Friston, K. J. (2007). Extra-classical receptive field effects measured in striate cortex with fMRI. NeuroImage, 34(3), 1199–1208.

    Article  Google Scholar 

  • Hinton, G. E. (2005). What kind of graphical model is the brain? In International joint conference on artificial intelligence 2005, Edinburgh.

  • Hinton, G. E. (2007). Learning multiple layers of representation. Trends in Cognitive Sciences, 11(10), 428–434.

    Article  Google Scholar 

  • Hinton, G. E., & Sejnowski, T. J. (1983). Optimal perceptual inference. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.

  • Hinton, G. E., & Sejnowski, T. J. (1999). Unsupervised learning: Foundations of neural computation. In G. E. Hinton & T. J. Sejnowski (Eds.), Unsupervised learning: Foundations of neural computation. Cambridge, MA: MIT Press.

    Google Scholar 

  • Hohwy, J. (2004). Top–down and bottom–up in delusion formation. Philosophy, Psychiatry and Psychology, 11(1), 65–70.

    Article  Google Scholar 

  • Hohwy, J. (2011). Phenomenal variability and introspective reliability. Mind & Language, 26(3), 261–286.

    Article  Google Scholar 

  • Hohwy, J. (2013). The predictive mind. Oxford: Oxford University Press.

    Book  Google Scholar 

  • Hohwy, J. (2015a). The neural organ explains the mind. In T. Metzinger & J. M. Windt (eds.) Open MIND (pp. 1–23). Frankfurt am Main: MIND Group.

  • Hohwy, J. (2015b). Prediction error minimization, mental and developmental disorder, and statistical theories of consciousness. In R. Gennaro (Ed.), Disturbed consciousness: New essays on psychopathology and theories of consciousness (pp. 293–324). Cambridge, MA: MIT Press.

    Chapter  Google Scholar 

  • Hohwy, J. (2016a). Prediction, agency, and body ownership. In A. Engel, K. Friston, & D. Kragic (Eds.), Where is the action? The pragmatic turn in cognitive science. Cambridge, MA: MIT Press.

    Google Scholar 

  • Hohwy, J. (2016b). The self-evidencing brain. Noûs, 50(2), 259–285.

    Article  Google Scholar 

  • Hohwy, J. (2017). Priors in perception: Top-down modulation, Bayesian perceptual learning rate, and prediction error minimization. Consciousness and Cognition, 47, 75–85.

    Article  Google Scholar 

  • Hohwy, J., & Michael, J. (2017). Why would any body have a self. In F. Vignemont & A. Alsmith (Eds.), The body and the self. Cambridge, MA: MIT Press.

    Google Scholar 

  • Hosoya, T., Baccus, S. A., & Meister, M. (2005). Dynamic predictive coding by the retina. Nature, 436(7047), 71.

    Article  Google Scholar 

  • Hutto, D. (2017). Getting into the great guessing game: Bootstrap heaven or hell? Synthese. doi:10.1007/s11229-017-1385-0.

  • Hutto, D., & Myin, E. (2013). Radicalizing enactivism: Basic minds without content. Cambridge, MA: MIT Press.

    Google Scholar 

  • Kirchhoff, M. (2016). Autopoiesis, free energy, and the life–mind continuity thesis. Synthese. doi:10.1007/s11229-016-1100-6.

  • Klein, C. (2016). What do predictive coders want? Synthese. doi:10.1007/s11229-016-1250-6.

  • Le, Q. V., Ranzato, M., Monga, R., Devin, M., Chen, K., Corrado, G. S., et al. (2012). Building high-level features using large scale unsupervised learning. In Proceedings of the 29th international conference on machine learning, Edinburgh.

  • Limanowski, J., & Blankenburg, F. (2013). Minimal self-models and the free energy principle. Frontiers in Human Neuroscience, 7, 1–12.

    Article  Google Scholar 

  • Loughlin, V. (2016). Jakob hohwy: The predictive mind. Phenomenology and the Cognitive Sciences. doi:10.1007/s11097-016-9479-6.

  • Lupyan, G. (2015). Cognitive penetrability of perception in the age of prediction: Predictive systems are penetrable systems. Review of Philosophy and Psychology, 6(4), 547–569.

    Article  Google Scholar 

  • Macpherson, F. (2017). The relationship between cognitive penetration and predictive coding. Consciousness and Cognition, 47, 6–16.

    Article  Google Scholar 

  • Mathys, C. D., Lomakina, E. I., Daunizeau, J., Iglesias, S., Brodersen, K. H., Friston, K. J., et al. (2014). Uncertainty in perception and the Hierarchical Gaussian Filter. Frontiers in Human Neuroscience, 8, 825.

    Article  Google Scholar 

  • Metzinger, T., & Wiese, W. (Eds.). (2017). Philosophy and predictive processing. MIND Group: Frankfurt am Main.

    Google Scholar 

  • Millikan, R. (1984). Language, thought, and other biological categories. Cambridge, MA: MIT Press.

    Google Scholar 

  • Millikan, R. (1989). Biosemantics. The Journal of Philosophy, 86(6), 281–291.

    Article  Google Scholar 

  • Neal, R. M., & Hinton, G. E. (1998). A view of the EM algorithm that justifies incremental, sparse, and other variants. NATO ASI Series D Behavioural and Social Sciences, 89, 355–370.

    Google Scholar 

  • O’Brien, G., & Opie, J. (2004). Notes toward a structuralist theory of mental representation. In H. Clapin, P. Staines, & P. Slezak (Eds.), Represenation in mind: New approaches to mental representation. Oxford: Clarendon Press.

    Google Scholar 

  • Orlandi, N. (2014). The innocent eye: Why vision is not a cognitive process. Oxford: Oxford University Press.

    Book  Google Scholar 

  • Orlandi, N. (2016). Bayesian perception as ecological perception. Philosophical Topics, 44(2), 327–351.

    Article  Google Scholar 

  • Palmer, C. J., Lawson, R. P., & Hohwy, J. (2017). Bayesian approaches to autism: Towards volatility, action, and behavior. Psychological Bulletin, 143(5), 521–542.

    Article  Google Scholar 

  • Papineau, D. (1984). Representation and explanation. Philosophy of Science, 51(4), 550–572.

    Article  Google Scholar 

  • Rao, R., & Ballard, D. (1999). Predictive coding in the visual cortex: A functional interpretation of some extra-classical receptive-field effects. Nature Neuroscience, 2, 79–87.

    Article  Google Scholar 

  • Sellars, W. (1953). Inference and meaning. Mind, 62(247), 313–338.

    Article  Google Scholar 

  • Sellars, W. (2007). In the space of reasons. Cambridge: Harvard University Press.

    Google Scholar 

  • Seth, A. K. (2013). Interoceptive inference, emotion, and the embodied self. Trends in Cognitive Sciences, 17(11), 565–573.

    Article  Google Scholar 

  • Seth, A. K. (2014). A predictive processing theory of sensorimotor contingencies: Explaining the puzzle of perceptual presence and its absence in synesthesia. Cognitive Neuroscience, 5(2), 97–118.

    Article  Google Scholar 

  • Seth, A. K., Suzuki, K., & Critchley, H. D. (2012). An interoceptive predictive coding model of conscious presence. Frontiers in Psychology, 2, 1–16.

    Article  Google Scholar 

  • Shea, N. (2014). Exploitable isomorphism and structural representation. Proceedings of the Aristotelian Society Supplementary, 114(2), 123–144.

    Article  Google Scholar 

  • Sims, A. (2016). A problem of scope for the free energy principle as a theory of cognition. Philosophical Psychology, 29(7), 967–980.

    Article  Google Scholar 

  • Trappenberg, T. (2010). Fundamentals of computational neuroscience. Oxford: Oxford University Press.

    Google Scholar 

  • Van de Cruys, S., Evers, K., Van der Hallen, R., Van Eylen, L., Boets, B., Lee de-Wit, L., et al. (2014). Precise minds in uncertain worlds: Predictive coding in autism. Psychological Review, 121(4), 649–675.

    Article  Google Scholar 

  • Vetter, P., & Newen, A. (2014). Varieties of cognitive penetration in visual perception. Consciousness and Cognition, 27, 62–75.

    Article  Google Scholar 

Download references

Acknowledgements

We wish to thank Michael Kirchhoff and two anonymous reviewers for comments. JH is supported by Australian Research Council Grants FT100100322 and DP160102770, and by the Research School Bochum and the Center for Mind, Brain and Cognitive Evolution, Ruhr-University Bochum.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Jakob Hohwy.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Kiefer, A., Hohwy, J. Content and misrepresentation in hierarchical generative models. Synthese 195, 2387–2415 (2018). https://doi.org/10.1007/s11229-017-1435-7

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11229-017-1435-7

Keywords

  • Problem of content
  • Misrepresentation
  • Functional role semantics
  • Structural resemblance
  • Prediction error minimization
  • Generative model
  • Recognition model
  • Kullbach–Leibler divergence
  • Variational Bayesian inference
  • Unsupervised learning