I think one of the most helpful recent developments in philosophy has been the increased focus on understanding as an epistemic aim distinct from knowledge. For an overview, see Stephen Grimm’s new entry in the SEP. While I don’t specialize in epistemology, thinking about understanding has made some things clearer to me when it comes to some important issues in the philosophy of science.
The Nature of Understanding
One way to see the difference between knowledge and understanding is to harken back to the famous Wilfrid Sellars quote: “The aim of philosophy, abstractly formulated, is to understand how things in the broadest possible sense of the term hang together in the broadest possible sense of the term (Sellars, 1963, 1).” Knowledge is often concerned about particular “things”, but understanding is crucially about how things “hang together.” A paradigm case of knowledge is my knowing an isolated fact (there is one beer in the fridge), whereas understanding concerns explanatory relationships or dependencies (how the beer came to be in the fridge, or why there is just one left out of a sixpack). As Grimm’s article outlines, philosophers are exploring and debating the various differences between the two notions (notably with regard to the role of truth—this is discussed further below). In addition, some philosophers are advancing the view that understanding is especially important or primary. For example, Angela Potochnik argued in her recent book that “understanding is the epistemic aim of science (Potochnik, 2017, 91).”
This emphasis on understanding is a fairly recent development. In the philosophy of science, there is a rich history which explores what explanations are and how they work. There has been less interest in understanding itself, despite its close connection to explanation. Sometimes, as Grimm cites in the case of Carl Hempel, this may be because understanding was thought of merely in terms of the subjective feeling that may accompany it (see section 1.2 of the SEP article). Indirectly, of course, the nature and importance of a more robust notion of understanding was often implicit in discussions of explanation. A central case in point is in the debate over the so-called “ontic” vs. “epistemic” conceptions of explanation (a framing primarily due to Wesley Salmon—see Salmon, 1984, pp.15-20).
Under Salmon’s preferred ontic conception, to explain something is to exhibit how it fits into the causal structure of the world. This was offered in contrast to the account of explanation offered by Hempel’s deductive-nomological (or “covering-law”) approach, which was categorized as “epistemic” because of its reliance on arguments that demonstrate a phenomenon was to be expected.[1] Subsequently, a wide variety of critics have argued that the ontic conception of explanation falls short in a number of ways. Prominent among these was the complaint that, as William Bechtel put it: “It is important to note that offering an explanation is still an epistemic activity and that the mechanism in nature does not perform any explanatory work (Bechtel, 2006, 34).” It is the job of the scientist to craft an explanation using whatever tools and forms of representation that might be appropriate for the case at hand. And the final step, of course, is to communicate the explanation effectively to an audience so it can be grasped. This latter dimension of explanation has been emphasized recently by Potochnik: “Successful explanation explicitly depends on features of human psychology and cognition as much as it depends on features of the world (Potochnik, 2017, 20).” The psychological dimension of understanding on the part of the audience is understood here, not as an subjective feeling, but as a particular cognitive act or process.
So, to put it all together, understanding as an epistemic achievement has several ingredients. Following Grimm’s presentation, a theoretical framework for understanding includes the following:
- The “distinctive object” of understanding, involving explanatory connections or relations of some sort;
- A “distinctive psychology” (I would prefer to say a distinctive cognitive act or process);
- A “distinctive normative relation” between the two.
I think this is a helpful framework. The article goes on to summarize some of the approaches philosophers have taken toward filling in this schema. I am going to sketch my own view below, with a focus limited to elements which seem crucial to scientific understanding.
Scientific Understanding and Causation
As discussed above, understanding is concerned with explanatory relations. A scientific explanation seeks to provide a vehicle for understanding by representing a phenomenon in the context of a network or structure of these relations. In practice, these relations are most often causal relations. (There is an active debate regarding the status of various possible non-causal approaches to scientific explanation, but I will ignore this for present purposes). Here is what I think the three-part framework requires when a causal explanation is the vehicle used to achieve understanding.
First, the ultimate object or target of the explanation must be a part of the causal structure of the world. Importantly, this doesn’t mean that the explanation needs to be especially complete or even accurate in its representation of this target. In fact, the models or other representational devices used may feature idealizations, and the entities, properties and relations described may all be, strictly speaking, false. Here I follow the insights of recent thinkers such as Catherine Elgin[2] and Potochnik who have proposed that understanding does not require truth. I think Potochnik is right when she argues that idealized elements play a positive representational role and can “stand in for significant causal influences (Potochnik, 2017, 52).” While she is agnostic about the ultimate nature of this representational relation, the option I favor is a similarity-based approach where the idealized element plays a similar role in the explanation as a real causal element does in nature (my principal area of disagreement with Potochnik has to do with the way she further characterizes the object of understanding, but I will not discuss this here).
Second, the cognitive dimension of understanding requires that the audience use concepts and reasoning processes to successfully grasp the explanation. I don’t have any kind of complete notion of what is required, here, but for causal explanations, I conclude that this process must (obviously) employ a concept of causation.
Third and finally, given the discussion so far, we can identify what I take to be a minimal component that is required for the normative relation between the cognizer and the object of understanding: it is a fit between the cognizers’s causal concept and the actual causal nature of the worldly phenomenon. I think this requirement can be satisfied, and that there is a shared conception of causation at work: this shouldn’t be too surprising, since our cognitive capacities are plausibly shaped by our own participation in the natural causal network.
Much more can be said about this, and my dissertation focused on giving a positive account of the kind of causation that underlies explanation and understanding in the natural sciences. But here I want to highlight one controversial implication of the picture I have sketched: successful scientific understanding does not require scientific realism, but does require realism about causation.
The Realism we Need
Grimm, when discussing different theorists’ views about the objects of understanding (sec. 2.1), contrasts internalist and externalist views. The internalist sees the object as involving connections (e.g. logical or linguistic) among things within the subject (e.g. beliefs). An externalist sees the connections as metaphysical, involving “real, mind-independent relationships that obtain in the world.” It seems clear to me that understanding achieved in the natural sciences must be externalist in this latter sense. We seek to understand the world. We use various resources and tools to facilitate this, but it is connections within nature that we grasp when we are successful.
As mentioned above, this success depends on a fit that exists between a causal concept we deploy and actual causation in the world. If we achieve scientific understanding, then we have grasped causal relationships that exist in nature. This framework implies ontological realism about causation. Interestingly, this is a stance that philosophers of science (with some prominent exceptions like Nancy Cartwright[3]) are often reluctant to explicitly embrace.
Interestingly, understanding does not require scientific realism[4]. As Potochnik argues (persuasively in my view), idealization is ubiquitous in successful science. This makes it much harder to support the idea that truth generally characterizes the products of science. This insight reinforces more traditional skepticism about scientific realism (usually framed in terms of theories), which focuses on arguments from underdetermination by data or the pessimistic induction from the historical record of discarded scientific posits. Now, philosophers have put forward a whole host of partial or qualified versions of scientific realism that seek to accommodate some of these challenges, but I will leave those aside for now.
What we are left with is an intriguing picture. The specific representative elements of scientific explanations are either deliberately false (in the case of idealizations) or else can be judged as unlikely to be true. Yet, these same scientific explanations are a vehicle for achieving understanding. In other words, they enable us to grasp “how things hang together”, even when the “things” described (entities, properties, and relations) aren’t real. They do this because their false elements stand in for features of nature’s causal network, and as a result convey degrees of understanding about the world’s real causal structure.
References
Bechtel, W. (2006). Discovering Cell Mechanisms. Cambridge: Cambridge University Press.
Cartwright, N. (1989). Nature’s Capacities and Their Measurement. Oxford: Oxford University Press.
Chakravartty, Anjan. (2017). “Scientific Realism”, The Stanford Encyclopedia of Philosophy (Summer 2017 Edition), Edward N. Zalta (ed.): https://plato.stanford.edu/archives/sum2017/entries/scientific-realism/
Elgin, C. (2017). True Enough. Cambridge, Massachusetts; The MIT Press.
Grimm, S. (2021). “Understanding”, The Stanford Encyclopedia of Philosophy (Summer 2021 Edition), Edward N. Zalta (ed.): https://plato.stanford.edu/archives/sum2021/entries/understanding/
Hempel, C. G. (1966). Philosophy of Natural Science. Englewood Cliffs, NJ: Prentice-Hall.
Potochnik, A. (2017). Idealization and the Aims of Science. Chicago: University of Chicago Press.
Salmon, W. C. (1984). Scientific Explanation and the Causal Structure of the World. Princeton: Princeton University Press.
Sellars, W. (1963). Science, Perception, and Reality. New York: The Humanities Press.
[1] See Salmon (1984), p. 16, and Hempel (1966), p. 48.
[2] See Elgin’s 2017 book, True Enough.
[3] As in her classic Nature’s Capacities and their Measurement (1989).
[4] There are many ways to characterize scientific realism. Given the topic of this post, the version being gestured at here is an epistemological form, where the products of science are thought to yield knowledge (see section 1.2 of the SEP article on Scientific Realism by Anjan Chakravartty).