【Michael Stuart】Epistemology of Experimental Imagination

Epistemology of Experimental Imagination
Michael Stuart

This article is a book commentary on Epistemic Uses of Imagination, a recently published volume edited by Chris Badura and Amy Kind (Routledge 2021).

Thought experiments are everywhere, from theoretical physics to TED talks. They’re challenging little stories that are sometimes so powerful, they demolish entire research traditions. But other than that, it’s hard to say exactly what they are. They’re more than mere arguments, but less than fully worked-out fictions. And they’re not just examples either: the open-endedness of thought experiments demands a level of mental effort that examples don’t.

What excites us about thought experiments is the very same thing that makes us doubt them as faithful guides to the truth: the central and necessary role of imagination.

Christopher Badura and Amy Kind’s new edited collection, Epistemic Uses of Imagination, contains a three-chapter section on thought experiments (TEs). We begin with a chapter by Margherita Arcangeli called “Narratives and Thought Experiments: Restoring the Role of Imagination.” It is appropriate that Arcangeli starts off this section, since she was the first in the modern debate to emphasize the importance of imagination to the topic of TEs (Arcangeli 2010; for some history on imagination and TEs before the current debate took form, see Brecevic 2021; Greene 2021; Stuart 2021).

How do we learn from narratives? One popular idea is that when we encounter a narrative, we begin to play a game of make believe. In the game, we are encouraged to imagine certain things and not others. We are sanctioned to infer whatever is allowed by “rules” of the game, which are given, or assumed, by the author and audience. We learn from a narrative TE when the game directs us to imagine certain things about the real world, and the rules are such that what we find in the imagination, we also find in the world (Meynell 2014; 2018; Salis and Frigg 2020). A second idea, also popular, is that when we encounter a narrative, we construct a “mental model” of the world depicted. Mental models are representational structures formed in working memory. We manipulate mental models to produce some output, which is informative about the target system when the representations are accurate and the manipulations obey constraints similar to those that govern the target system (Nersessian 1992; 2007; 2018; Miščević 1992; 2007; Swirski 2007). Which of these two ideas is right? In his 2014 book Fiction and Narrative, Derek Matravers supports eliminating talk about (propositional) imagination for talk about mental models.

Arcangeli argues it would be a (category) mistake to use the concept of mental models instead of imagination. That’s because mental models are best understood as the content (or object, or vehicle of the content) of our imaginings, while imagination is the psychological attitude we take toward that content. No matter how we characterize the content of our thought, we still need to take some attitude toward it. And, Arcangeli points out, imagination fits the bill. By leaving out a discussion of psychological attitudes, any mental-models-only framework won’t be able to explain important attitudinal aspects of engagement with TEs, including the experience of “being there” we sometimes have when performing a TE. And these aspects can be epistemically relevant. Arcangeli concludes that we should not replace imagination with mental models in our theorizing, but just the opposite: imagination is “more promising” given its explanatory advantages (pp. 194-5).

I think Arcangeli is right that imagination and mental models are not the same type of thing. But in that case, why not reject the antagonism, and keep both? Indeed, Nersessian summarizes “the general hypothesis” behind the mental model framework as the claim that “thought experimenting is a species of reasoning rooted in the ability to imagine, anticipate, visualize, and re-experience from memory” (2018, 308). Imagination can be conceived of as a faculty (that enables mental modelling), a skill (that is exercised while mental modelling), or a psychological state (that we can have toward elements of mental models). In sum, there is lots of room for theoretical development if we keep both concepts. For example, we can ask which parts of a mental model are imagined, and which are believed, or supposed, and which kinds of imagination are relevant to which acts done in the process of mental modelling.

The next chapter is by Margot Strohminger, called “Two Ways of Imagining Galileo’s Experiment.” It focuses on Galileo’s falling bodies TE, which is without question the TE that has received the most attention in the literature. Strohminger provides a fresh new interpretation of the TE. Her account isn’t meant to capture all the historical details of the TE, but to reconstruct it as charitably as possible, and see how an improved version might work.

In Galileo’s TE, we imagine two stones of the same material “joined together” and dropped from a height. According to the standard interpretation of this TE, this overturns the Aristotelian theory that heavier bodies composed of the same material fall faster (in the same medium). It does this because the conjoined stones form an even heavier object, which should therefore fall faster. But on the other hand, the lighter stone should retard the heavier one, making the conjoined object fall more slowly. The same thing can’t fall at two different speeds, so the Aristotelian theory must be wrong (for more detailed reconstructions, see Gendler 1998; Gruszczyński 2020; El Skaf 2018). According to Strohminger, the tension at the heart of the TE only arises if the conjoined stones are really, at the same time, one object, and two objects. Strohminger argues that it wouldn’t be fair to attribute this belief to Aristotelians, because it would mean that Galileo’s TE could be run on all kinds of weird things. For example, we should be able to see Plato and his (attached) foot as two objects, the latter retarding the fall of the former, while also seeing Plato and his (attached) foot as one object. Do we really think that Aristotelians should have been committed to the view that a falling Plato would have fallen both faster and slower than a version of Plato who was missing a foot? No way!

Strohminger offers a new proposal. What happens is that the Aristotelian carries out the TE by reasoning through a counterfactual. Starting from the premise of the experimental set-up, they try to reason it out, and come to two, contradictory conclusions by means of different “routes in imagination” (p. 209). What are the different routes? Strohminger suggests that they might be the two processes identified by dual process theory, that is, one is “intuitive” (system 1) while the other is “reflective” (system 2). In other words, one route is arrived at via intuitive folk physics (the composite object falls slower than the heavy object), and the other is arrived at via reflective reasoning (the composite object falls faster). Strohminger leaves it as an open question whether the intuitive route or the reflective route provide knowledge.

I wonder whether we need to cut things up this way. In my (2019) I also argued that we could use dual process theory to understand the epistemology of TEs, and I tried to explain how each kind of imagination might lend justification to a TE. I agree with Strohminger that when performing Galileo’s TE we do seem to run the same set-up twice. But I would argue that both kinds of imagination could be used in each “branch” of the TE. Some aspects of the set-up, in both branches, will be “reflective,” while others will be “intuitive.” Which aspects these are might depend on the individual. A child encountering the TE for the first time might require more reflective thinking for both branches, while a trained physicist might proceed along both branches intuitively. The refreshing (or worrying) simplicity of the dual process model of cognition can lure us into forgetting how context-dependent, dynamic, and messy it is in application. Still, Strohminger is right that the framework is heavy with potential for the epistemology of imagination and TEs.

Here, we point out that with respect to self-experience, there are several ways in which memory and imagination differ. One of the key differences might be freedom or flexibility. Imagination not only offers counterfactual scenarios but also enables our selves to be variously presented in such experience—imagining oneself as someone else, imagining from a different perspective, and while adopting an observer perspective, identifying with different targets. It, however, remains to be seen whether some of these dimensions of variation in self-consciousness are in fact unique to imagination.

The final chapter is “Attention to Details: Imagination, Attention, and Epistemic Significance,” by Eric Peterson. Peterson agrees that for imagination to have epistemic import, it must be constrained. But this is not enough: we must also pay attention to the right things, in imagination. Peterson understands attention as a (voluntary or involuntary) mental phenomenon or mode of consciousness “whose role is to selectively focus on a certain content/phenomenon” (p. 220). To pay attention is to have one thing in the “center” of our conscious experience, relegating the rest to the periphery.

To show that attention is epistemically important for imagination, Peterson first shows that it is important for perception. To be epistemically successful in perceiving, we need to perceive the right things, or we’ll miss what matters. Attention lets us fix the content of perception by attributing properties to objects. In attending to something, we center its redness, or its roundness, etc. This role for attention is a condition for the possibility of having beliefs with perceptual content. Another role for attention is in making determinable objects determinate. This is important because while we can make justified claims about determinable objects, we can make more (and possibly more justified) claims about determinate objects. Finally, when something is centered in our attention, it becomes accessible as a possible reason for belief or action. In the video below, the person in the gorilla suit is not centered, so it is not part of our perceptual content, is not determinate, and is not available as a reason for action or belief.

Thus, attention is epistemically relevant for knowledge via perception. And if attention is generally relevant, then we should expect it to be relevant for imagination as well. If this is right, to execute a TE well, we must attend to the right things. In performing the Mary the colour scientist TE, we should focus on her knowledge of colour vision and her experience as she sees colours for the first time, rather than irrelevant things. Without the right attention, we don’t form the right beliefs about the described situation, we don’t make the right things determinate, and we don’t get access to the aspects of the situation that would justify our belief in the conclusion of the TE. Indeed, it’s possible that when different philosophers provide different interpretations of what we learn from this (or any) TE, it is because there are differences in what they attend to.

I think Peterson is right that attention is relevant for the epistemology of imagination, though I’m not sure this insight can’t be captured by those who focus on constraints. Afterall, we could understand Peterson’s chapter as drawing our attention to one more important kind of constraint on good reasoning, namely, that we should pay attention to the right things.

I also wonder whether everyone, should always pay attention to the same things when performing a particular TE. Couldn’t it be that sometimes it’s epistemically better not to pay attention, or to pay attention to different things? I think this is especially plausible if we define “epistemically better” as “having better epistemic consequences.” For example, Einstein’s clock in the box TE was about a way to violate Heisenberg’s uncertainty relation. But Bohr didn’t limit his attention to the possibility that Einstein emphasized. Instead, he focused intently on the details of the measurement apparatus that Einstein alluded to in the TE. Nevertheless, this was the right thing to do. This raises interesting questions: what are the “right” things to pay attention to in a TE, and how could we know? Doesn’t the “experimental” nature of TEs require extra leeway in how we interact with them in our imaginations?



Arcangeli, Margherita. 2010. “Imagination in Thought Experimentation: Sketching a Cognitive Approach to Thought Experiments.” In Model-Based Reasoning in Science and Technology: Abduction, Logic, and Computational Discovery, edited by Lorenzo Magnani, Walter Carnielli, and Claudio Pizzi, 571–87. Studies in Computational Intelligence. Berlin, Heidelberg: Springer. https://doi.org/10.1007/978-3-642-15223-8_32.

Brecevic, Char. 2021. “The Role of Imagination in Ernst Mach’s Philosophy of Science: A Biologico-Economical View.” HOPOS: The Journal of the International Society for the History of Philosophy of Science 11 (1): 241–61. https://doi.org/10.1086/712974.

Gendler, Tamar S. 1998. “Galileo and the Indispensability of Scientific Thought Experiment.” British Journal for the Philosophy of Science 49: 397–424.

Greene, Catherine. 2021. “Historical Counterfactuals, Transition Periods, and the Constraints on Imagination.” HOPOS: The Journal of the International Society for the History of Philosophy of Science 11 (1): 305–23. https://doi.org/10.1086/712937.

Gruszczyński, Rafał. 2020. “Parts of Falling Objects: Galileo’s Thought Experiment in Mereological Setting.” Erkenntnis, May. https://doi.org/10.1007/s10670-020-00263-y.

Matravers, Derek. 2014. Fiction and Narrative. Oxford University Press.

Meynell, Letitia. 2014. “Imagination and Insight: A New Acount of the Content of Thought Experiments.” Synthese 191 (17): 4149–68. https://doi.org/10.1007/s11229-014-0519-x.

———. 2018. “Images and Imagination in Thought Experiments.” In The Routledge Companion to Thought Experiments. Routledge. https://doi.org/10.4324/9781315175027.ch28.

Miščević, Nenad. 1992. “Mental Models and Thought Experiments.” International Studies in the Philosophy of Science 6 (3): 215–26. https://doi.org/10.1080/02698599208573432.

———. 2007. “Modelling Intuitions and Thought Experiments.” Croatian Journal of Philosophy VII: 181–214.

Nersessian, Nancy J. 1992. “In the Theoretician’s Laboratory: Thought Experimenting as Mental Modeling.” PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association 1992: 291–301.

———. 2007. “Thought Experiments as Mental Modelling: Empiricism without Logic.” Croatian Journal of Philosophy VII: 125–61.

———. 2018. “Cognitive Science, Mental Modeling, and Thought Experiments.” In The Routledge Companion to Thought Experiments. Routledge. https://doi.org/10.4324/9781315175027.ch17.

Salis, Fiora, and Roman Frigg. 2020. “Capturing the Scientific Imagination.” In The Scientific Imagination, edited by Arnon Levy and Peter Godfrey-Smith, 17–50. Oxford University Press.

El Skaf, Rawad. 2018. “The Function and Limit of Galileo’s Falling Bodies Thought Experiment.” Croatian Journal of Philosophy 18 (1): 37–58.

Stuart, Michael T. 2019. “Towards a Dual Process Epistemology of Imagination.” Synthese. https://doi.org/10.1007/s11229-019-02116-w.

———. 2021. “Telling Stories in Science: Feyerabend and Thought Experiments.” HOPOS: The Journal of the International Society for the History of Philosophy of Science 11 (1): 262–81. https://doi.org/10.1086/712946.

Swirski, Peter. 2007. Of Literature and Knowledge: Explorations in Narrative Thought Experiments, Evolution and Game Theory. 1st ed. Routledge. https://doi.org/10.4324/9780203965863.