Imaginative Constraints and Generative Models

ABSTRACT

How can imagination generate knowledge when its contents are voluntarily determined? Several philosophers have recently answered this question by pointing to the constraints that underpin imagination when it plays knowledge-generating roles. Nevertheless, little has been said about the nature of these constraints. In this paper, I argue that the constraints that underpin sensory imagination come from the structure of causal probabilistic generative models, a construct that has been highly influential in recent cognitive science and machine learning. I highlight several attractions of this account, and I favourably contrast it with Peter Langland-Hassan’s account of sensory imagination in terms of the forward models exploited in sensorimotor control.

KEYWORDS:

• imagination
• constraints
• knowledge
• learning
• generative model
• predictive coding

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 There is also extensive empirical evidence attesting to the centrality of imagery in such cases, which suggests that this impression is veridical (see Hegarty [2004] and Moulton and Kosslyn [2009]).

2 I thank an anonymous reviewer for raising this important point.

3 When Williamson [2016] considers the cognitive processes underlying this process of imaginative inference, however, he draws an analogy to logical inferences over sentence-like representations, which is different from the account of sensory imagination developed here.

4 In this way the account developed here is consistent with Jackson’s [2018: 222] ‘recreativist’ solution to the Up-To-Us Challenge, in which ‘imagination conforms to whatever structural regularities govern perception’.

5 I am grateful to an anonymous reviewer for raising this point.

6 I would like to thank those attending the ‘Mental Imagery and Bayesian Models in Philosophy and Cognitive Science’ workshop at the Centre for Philosophical Psychology, University of Antwerp, in May 2019 for helpful comments and discussion, especially Bence Nanay and Jakob Hohwy. I would also like to thank two anonymous reviewers for helpful comments and suggestions.

Additional information

Funding

This work was generously supported by the Fonds voor Wetenschappelijk Onderzoek (FWO) Odysseus grant [G.0020.12N] and the Fonds voor Wetenschappelijk Onderzoek (FWO) research grant [G0C7416N], as well as by a Research Fellowship from Corpus Christi College, University of Cambridge.