Abstract
In this paper, I explicate and discuss performance-similarity reasoning as a strategy for mechanism schema evaluation, understood in Lindley Darden’s sense. This strategy involves inferring hypotheses about the mechanism responsible for cognitive capacities from premises describing the performance of those capacities; performance-similarity reasoning is a type of Inference to the Best Explanation, or IBE. Two types of such inferences are distinguished: one in which the performance of two systems is compared, and another when the performance of two systems under intervention is compared. Both types of inferences are illustrated with examples taken from cognitive science. I conclude that performance-similarity reasoning is an important strategy for evaluating mechanistic hypotheses.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The turn to mechanisms has not gone uncontested, however: in the field of cognitive science, for instance, Stepp et al. (2011) have argued that mechanistic explanations are insufficient, while in biology, Bechtel (2011) has criticized the basic account of mechanisms as too simplistic to understand the dynamic behavior of certain complex systems. For all these criticisms however, there are probably few philosophers who would argue that the mechanistic movement is flawed at heart (Bickle comes close when he says that his ‘intervene cellularly/molecularly and track behaviorally’ account of reduction better captures the explanatory practices of neuroscience than the analysis of the new mechanists, 2006, p. 431). What is disputed however, is that it is the aim of models in cognitive science to describe mechanisms. Weiskopf (2011) for example, argues that psychological models are not mechanistic, analyzing capacities in terms of semantic and intentional (or more generally representational) states and processes, rather than entities and activities. Others have argued that in fact, Weiskopf’s functional analyses are to be understood as mechanism sketches (Povich 2015). While I do believe that Cummins-style functional analyses have their place in cognitive science, there is little doubt that many models in cognitive science are committed to describing mechanisms (see the examples of models of face recognition in Sect. 6). In any case, a functionalist might disagree with the approach I take in this paper, yet still find performance-similarity reasoning a valuable tool—though valuable for evaluating cognitive models rather than mechanism schemas.
Throughout this article, I shall use Craver and Darden’s account of discovering mechanisms, as presented in their (2013), as the backdrop against which I will analyze new strategies for hypothesis evaluation. This is not because it is the only such account (cf. Bechtel 2006), but because it is a particularly comprehensive and recent one.
Thus, I follow Craver and Darden in assuming what they term (2013, p. 10) a ‘garden-variety realism’: there are target mechanisms in the world, and it is possible to obtain evidence about those mechanisms in light of which we adjust our degree of belief in our hypotheses about those target mechanisms.
Craver and Darden do not mention IBE in their (2013); the present paper constitutes and attempt to extend their strategies for schema evaluation to that topic.
The inference schemas presented here are loosely adapted from Thagard (2014). There are differences: Thagard provides a list of inference steps, where the last step states that the proposed architecture explains the capacity in question. In my inferences, a conclusion about the character of the responsible mechanism is inferred from the capacity, and the premise that instantiating a mechanism of this character explains the capacity. Also, the premise referring to alternative hypotheses is not present in (Thagard 2014).
One could also see these inferences as hypothetico-deductive: if the theory that people have the relevant connectionist mechanism entails that they are able to recognize faces, then the presence of this ability confirms that theory, while its absence disconfirms it. I will stick to IBE, however.
See Craver and Darden (2013, ch. 2) for a listing and description of many features of mechanisms.
There is an important sense in which PS reasoning is connected with extrapolation. Steel (2008) asks the following question: how do we know that a model is suitable for extrapolation to a particular target? He argues that for any account of extrapolation that wants to go beyond ‘simple induction’ (simply inferring that a given causal relationship will also hold in the target population), there are two problems to solve. First, there is the extrapolator’s circle: in order to extrapolate, one needs to know enough about the target population for there to be no need for extrapolation in the first place. Second, how can one justify extrapolation when it is almost inevitable that there will be causally relevant differences between the populations? Steel’s own answer following a procedure he calls ‘comparative process training’ (cf. 2008, ch. 5), which need not concern us here, since it requires information about the target mechanism (albeit limited information). PS reasoning does not in principle require any information about the mechanism of the target mechanism (hence it avoids the extrapolator’s circle), but gives us a preliminary indication of mechanistic similarity. It will often (but not always) be used prior to any in depth comparison of mechanisms, suggesting where to look for a model system rather than giving an outright justification of its use for extrapolation.
Of course, depending on the situation, one of these sub-virtues can take precedence over others, but in general, an increase in one of them is an increase in overall performance.
As a cautionary note, it is important not to exaggerate the rivalry. Some (e.g., Barnden and Srivinas 1992) hold that the two models are actually compatible, and that while the mind is a neural net, from a more abstract viewpoint, symbolic processing is implemented by this net. Although comparing connectionism and classical computationalism was ‘hot’ in the 1980s and early 1990s (Dennett 1993, p. 269 called it “a major industry in academia”), it is not as if one of the two emerged as ‘the winner’. The upshot of all this is a point that I have mentioned before, namely that we should not consider PS inferences as any kind of ‘proof’ for the respective conclusions they draw about the mechanism responsible for the cognitive capacity of interest. Rather, they serve as heuristic tools that help us evaluate our hypotheses. Indeed, the inferences discussed below do not ‘establish the architecture of the human mind’. However, they were presented as arguments for computational and connectionist theories of cognition, which is enough to serve our present purpose.
This argument is part of the larger ‘systematicity debate’. See Symons and Calvo (2014) for a recent overview of the debate.
Some will reject this claim. However, I consider the arguments as they appeared in the literature at the time. Whether they ultimately stood the test of time is another matter.
This observation may itself be used for a PS inference drawing a negative conclusion about a classic computationalist architecture: when implemented with neurons, the computationalist architecture would be much slower, so that the fact that a computationalist schema matches human speed actually counts against the computationalist schema. Of course, we can accordingly amend the inference, so that S stands for a neuron-implemented computational architecture, so that we have a straightforward PS inference drawing a negative conclusion from a dissimilarity in performance.
References
Adini Y, Moses Y, Ullman S (1997) Face recognition: the problem of compensating for changes in illumination direction. IEEE Trans Pattern Anal Mach Intell 19:721–731
Barnden J, Srivinas K (1992) Overcoming rule-based rigidity and connectionist limitations through massively-parallel case-based reasoning. Int J Man Mach Stud 36:221–246
Bechtel W (2006) Discovering cell mechanisms: the creation of modern cell biology. Cambridge University Press, Cambridge
Bechtel W (2011) Mechanism and biological explanation. Philos Sci 78:533–557
Bickle J (2006) Reducing the mind to molecular pathways: explicating the reductionism implicit in current cellular and molecular neuroscience. Synthese 151:411–434
Bouton ME (2007) Learning and behavior: a contemporary synthesis. Sinauer, Sunderland
Bruyer R (1991) Covert face recognition in prosopagnosia: a review. Brain Cogn 15:223–235
Craver CF (2006) When mechanisms explain. Synthese 153:355–376
Craver CF (2007) Explaining the brain. Clarendon Press, Oxford
Craver CF, Darden L (2013) In search of mechanisms: discoveries across the life sciences. University of Chicago Press, Chicago
Cummins R (2000) “How does it work?” versus “What are the laws?” Two conceptions of psychological explanations. In: Keil F, Wilson R (eds) Explanation and cognition. MIT Press, Cambridge, pp. 117–145
Dailey MN, Cottrell GW, Busey TA (1999) Facial memory is kernel density estimation (almost). In: Kearns MS, Solla SA, Cohn DA (eds) Advances in neural information processing, vol 11. MIT Press, Cambridge, pp. 24–30
Darden L (2009) Discovering mechanisms in molecular biology. Finding and fixing incompleteness and incorrectness. In: Meheus J, Nickles T (eds) Models of discovery and creativity. Springer, Dordrecht, pp. 34–55
Dennett DC (1993) Consciousness explained. Penguin Books, London
Farah MJ, O’Reilly RC, Vecera SP (1993) Dissociated overt and covert recognition as an emergent property of a lesioned neural network. Psychol Rev 100:571–588
Feldman JA, Ballard DH (1982) Connectionist models and their properties. Cogn Sci 6:205–254
Fodor J (1975) The language of thought. Thomas Y. Cromwell, New York
Fodor J (1980) Methodological solipsism considered as a research strategy in cognitive psychology. Behav Brain Sci 3:63–73
Fodor J (1987) Psychosemantics. MIT Press, Cambridge
Fodor J, Pylyshyn Z (1988) Connectionism and cognitive architecture: a critical analysis. Cognition 28:3–71
Hancock, P. J. B., Bruce V, Burton MA (1998) A comparison of two computer-based face identification systems with human perception of faces. Vision Res 38:2277–2288
Joormann J (2004) Attentional bias in dysphoria: the role of inhibitory processes. Cogn Emot 18:125–147
Kalocsai P, Zhao W, Elagin E (1998) Face similarity as perceived by humans and artificial systems. Proceedings of the Third International Conference on Automatic Face and Gesture Recognition: 177–180. IEEE Computer Society, Los Alamitos
Lamberts K, Goldstone R (2004) Handbook of cognition. Sage, London
Machamer PK, Darden L, Craver CF (2000) Thinking about mechanisms. Philos Sci 57:1–25
Mayr S, Buchner A (2007) Negative priming as a memory phenomenon—a review of 20 years of negative priming research. J Psychol 21: 35–51
Mitchell CJ, De Houwer J, Lovibond PF (2009) Behav Brain Sci 32:183–246
Neill WT, Valdes LA (1992). The persistence of negative priming: steady-state or decay? J Exp Psychol 18:565–576
Pinker S, Prince A (1988) On language and connectionism: analysis of a parallel distributed processing model of language acquisition. Cognition 28:73–194
Povich M (2015) Mechanisms and model-based functional magnetic resonance imaging. Philos Sci 82:1035–1046
Rumelhart DE (1989) The architecture of mind: a connectionist approach. In: Posner MI (ed) Foundations of cognitive science. MIT Press, Cambridge, pp. 133–159
Steel P (2008) Across the boundaries: extrapolation in biology and social science. Oxford University Press, New York
Stepp N, Chemero A, Turvey MT (2011) Philosophy for the rest of cognitive science. Top Cogn Sci 3:425–437
Symons J, Calvo P (2014) Systematicity, an overview. In: Calvo P, Symons J (eds) The architecture of cognition: rethinking fodor and pylyshyn’s systematicity challenge. MIT Press, Cambridge, pp. 3–30
Thagard, P. (2014). Cognitive Science. The Stanford Encyclopedia of Philosophy (Fall 2014 Edition), Edward N. Zalta (ed.), http://plato.stanford.edu/archives/fall2014/entries/cognitive-science/. Accessed 29 May 2015
Tranel D, Damasio AR (1988) Nonconscious face recognition in patients with face agnosia. Behav Brain Res 30:235–249
Weiskopf DA (2011) Models and mechanisms in psychological explanation. Synthese 183:233–258
Acknowledgements
The author acknowledges support from the Research Fund Flanders (FWO).
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Gervais, R. Performance-Similarity Reasoning as a Source for Mechanism Schema Evaluation. Topoi 39, 69–79 (2020). https://doi.org/10.1007/s11245-017-9507-3
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11245-017-9507-3