A post by Neil Van Leeuwen

Is imagination epistemically safe?

In some intuitive sense of that phrase, the answer is yes. One can imagine outlandish and false things without risk of being wrong. I can imagine that I am a walrus without the epistemic failure that would obtain if I believed that I am a walrus. Sure, the content is false, but that’s okay: I’m only imagining it. The difference in attitude, even holding content fixed, makes a difference to epistemic risk.

The view just expressed is natural enough that one might even appeal to “epistemic safety” for the very purpose of characterizing what it is to have an imaginative attitude. Here’s Peter Langland-Hassan’s initial characterization of attitude imagining: “Attitude imaginings (or A-imaginings) are, again, cases of rich, elaborated, epistemically safe thought about the possible, pretended, unreal, and so on.”[1] Langland-Hassan’s subsequent theoretical task is to articulate the psychological structure of the mental states that fit that broad specification; if he does this well, he’s given a theory of attitude imagining.

Now, I happen to disagree[2] with the psychological details of Langland-Hassan’s more particular theory—his attempt to cash out attitude imagining entirely in terms of what he deems more “basic” mental states. But that is consistent with my agreeing with him that epistemic safety (in some sense of that term) differentiates imagining that p (whatever that turns out to be in terms of psychological processing) from believing that p. Here’s how I articulate the idea in somewhat different terms in a recent piece about cognition of fiction (where Ba<s-i->p means agent A believes that the story i p obtains and Iap means agent A imagines that p):

Ba<s-i->p can be epistemically correct or incorrect in a way that Iap cannot be. A belief with false or inaccurate contents is in and of itself an epistemic failing, whereas just imagining inaccurate things . . . is not in and of itself an epistemic failing. Imagining falsely becomes an epistemic flaw only when it leads to believing falsely—and then only because it so leads.[3]

In other words, imagining is epistemically safe. Though I myself didn’t use it, the word “safe” feels appropriate to express the idea in question for the following reason. If I believe that p, there are all sorts of things that a cruel and merciless world can do to turn my mental state into an epistemic failing—all sorts of ways to make p false, the majority of which (yikes!) are beyond my control. But if I just imagine that p, let the world do what it will in terms of turning that proposition false—my mental state will be safe from thereby being an epistemic failing (hooray!).

So epistemic safety, in the sense just developed, is a normative property of imagining (something like immunity from being an epistemic failing just in virtue of having false contents) that gives us a desideratum on descriptive psychological theories of what imagining is in terms of functional processing.

Epistemic safety desideratum: a psychological theory of what imagining is should help make sense of why imagining is epistemically safe in the way that belief is not.

This should be useful common ground in debates about the psychological nature of imagination: however, much we disagree about the psychological nature of imagination, we can agree that our theory of it should satisfy that desideratum.

Yet I worry that this line of thought may occasion some grumbling from more traditional epistemologists. That’s because they are wont to use the phrase “epistemic safety” in a rather different way from what we have just seen, such that we imagination theorists might be accused of terminological appropriation.[4]

For Ernest Sosa[5] and quite a few others, “safety” is what differentiates true beliefs that are knowledge from true beliefs that aren’t. The general idea is to rule out beliefs that came out true by luck or accident: e.g., I look at a broken clock at just the right time and happen to form a true belief about what time it is (e.g., that it is 10:45 AM). That belief is true but not an instance of knowledge because its manner of formation is faulty. And so on. Epistemic safety, on this construal, is meant to be a (the?) notion that replaces justification in post-Gettier attempts to articulate what else a true belief must have in order to be an instance of knowledge.

Accordingly, a cottage industry in epistemology has sprung up that aims at making precise this intuitive notion of epistemic safety, effectively to pressure test it to see if it can play the knowledge-characterizing role set out for it. Here’s Sosa’s original version: “Call a belief by S that p ‘safe’ iff: S would believe that p only if it were so that p.”[6] (Apply this to the clock case: looking at the clock I would still believe it’s 10:45 AM even if it were not so, so even in the case where that belief is true it is not safe in the relevant sense.) Of course, as one might expect, there are examples and counterexamples that make it hard to state precisely what “safety” amounts to in a way that makes it satisfactory for helping to get the extension of “knowledge” correct. Hence we see the cottage industry in question, which includes such notable epistemologists as Timothy Williamson, Duncan Pritchard, and Jennifer Nagel, among quite a few others.

So “epistemic safety” in one sense is a normative property that differentiates imagining from belief. But “epistemic safety” in another sense is a descriptive (modal) property that characterizes the formation processes of true beliefs that rise to the level of knowledge.

I started with a question and want to end with a few.

Should we conclude that there are two distinct notions here? Or is there a deeper sense that unifies the different notions of epistemic safety? In a trivial way, mental states that have either form of epistemic safety are “safe” from being false beliefs. But they are safe in different ways—one because the state is not being a belief at all and the other because the formation process of the belief has certain properties. If one wanted to claim unity, one would hope for something more than that (I think). So: Is there enough unity there to say the two notions of epistemic safety are ultimately deep down about the same thing, or not? And if so, what is it?

* * * 

Acknowledgement: Thanks to Daniel Munro for feedback on an earlier draft of this blog.

Notes:

[1] Langland-Hasssan (2020: 58); my italics.

[2] Van Leeuwen (forthcoming).

[3] Van Leeuwen (2021: 646).

[4] Or at least Peter might be.

[5] Sosa (1999).

[6] Sosa (1999: 142).

References

Langland-Hassan, P. (2020). Explaining Imagination. New York, NY: Oxford University Press.

Sosa, E. (1999). How to defeat opposition to Moore. Philosophical Perspectives 13, 137-149.

Van Leeuwen, N. (forthcoming). Pennywise Parsimony: Langland-Hassan on Imagination. Analysis

Van Leeuwen, N. (2021). Imagining Stories: Attitudes and Operators. Philosophical Studies 178, 639-664. https://doi.org/10.1007/s11098-020-01449-4