Why is Cognition Question-Sensitive?
Juan Murillo Vargas (MIT)
Author’s note: I no longer believe most of the paper’s claims. But I find the question interesting and worth thinking about, even if I think the answer I gave for it is no longer workable. I’m curious to hear what readers think!
It’s become trendy in certain corners to think the mind is question-sensitive: many of the ways we think and talk are sensitive to the questions we’re posed, not just the information we have available. On this trend the mind is best understood as an active inquirer rather than as a passive information-processor. Call this the zetetic picture of the mind.
One central motivation for the zetetic picture of the mind comes from the success of question- sensitive models of how we think and talk.1 Here’s an example that motivates such models. Compare two discourses:
(1) a. Jim: is it raining outside?
b. Tim: yes, it’s raining outside.
(2) a. Jim: is it raining outside?
b. Tim: # Wheaties is the breakfast of Champions.
Tim’s reply in (1b) is fine. His reply in (2b)…not so much. Why? Because whether an assertion is felictious depends on the question at hand (Groenendijk and Stokhof, 1984; Roberts, 1996, 2012). It’s raining outside is a great answer to is it raining outside? whereas Wheaties is the breakfast of Champions is not. Hence the contrast in felicity. Similar patterns are used to motivate the idea that thinking (Holgu´ın, 2022), guessing (Dorst and Mandelkern, 2021), knowing (Schaffer and Szabo´, 2014), believing (Yalcin, 2018; Hoek, 2022), intending (Beddor and Goldstein, 2023), and more are question-sensitive.2
If this question-centric way of thinking about the mind is right, presumably there’s a reason why so much of how we think and talk is question-sensitive. Hence the titular question: why is cognition question-sensitive?
A prevailing intuition is that questions in some sense make life easier—it is easier to navigate the world question-by-question than it otherwise would be. My paper’s aim was to flesh out this intuition. To that end I defended what I called the questions-tractability hypothesis.3
Questions-Tractability Hypothesis
One of questions’ main functions is to aid computational tractability without sacrificing flexibility.
On this view questions make life easier in a very literal way: they help us save computational resources without giving up flexibility. This is valuable since we are computationally limited agents who can’t always sacrifice flexibility.
How to go about defending this hypothesis? First I argued that we only see question-sensitivity in what Fodor (1983) called central cognition: the suite of cognitive processes that are holistic, flexible, and have access to information from anywhere in cognition. Then I combined work in natural language semantics and computational complexity theory to argue that questions could in principle reduce computational costs. Let me sketch each argument in turn.
To see why it seems plausible that question-sensitivity seems to only be present in central cognition, it’s best to just talk through example. Communicating, believing, intending, etc. are all paradigm central cognitive processes. Quine’s metaphor of a “web” of beliefs—where each belief is sensitive to every other belief—is a good way of thinking about it. Whether I believe it’s raining on the basis of a weather report depends on a web of other beliefs, e.g., whether I believe the weather report is reliable. The same goes for communicating and intending. And as we saw above, cognitive processes that have this “web-y” character have been claimed to be question-sensitive.
By contrast, processes that have been claimed to not be part of central cognition—perception and syntax, just to name two—don’t seem to be question-sensitive. I can’t make the Müller-Lyer illusion go away by asking you the right question.
Figure 1: Muller-Lyer Illusion
And (3) just isn’t grammatical regardless of what question I ask you.
- * Himself ate the sandwich. (Meaning: he ate the sandwich.)
Thus we have a generalization. Candidates for central cognition are candidates for being question- sensitive. Candidates for not being part of central cognition are not candidates for being question-sensitive. Hunch: questions’ function has something to do with central cognition’s holistic character— hence why question-sensitivity seems to only arise in central cognition.
To see how questions might aid tractability, we need just a bit of formalism. Following some natural language semanticists (see e.g., Groenendijk and Stokhof, 1984) let’s think about questions as partitions of logical space: they pick out the sets of propositions that answer the question. (4) picks out the set in (5) for example.
- Is it raining outside?
- {it is raining outside, it is not raining outside}
Now we need to think about what sorts of factors affect computational tractability. Let’s focus on two: input size—how much information that computation is processing—and scaling—how computational costs increase as the amount of information increases (see e.g., Garey and Johnson, 1979; Van Rooij et al., 2019; Brooke-Wilson, 2023).
Questions can aid with both. The question in (4) shrinks the amount of information a process needs to compute: instead of computing over all of logical space, it need only compute over the partition picked out by the question. Crucially, questions can also aid with scaling. One factor that affects scaling is how many different kinds of information a process needs to consider. Questions can reduce that too. For instance (6) reduces the kinds of information a process needs to consider—it only needs to look at locations, rather than (say) locations and people.
- Where is London?
- {London is in Ontario, London is in England, … }
Thus questions can aid with tractability in multiple fronts.
Why would questions be helpful for central cognition in particular? It’s worth comparing them to other alleged ways of saving computational costs.
Some folks think that imposing architecturally built-in restrictions on the amount of information a cognitive process has to consider can aid tractability by shrinking input size (for the locus classicus see e.g., Fodor, 1983). Even if that’s right, that won’t work for central cognition since central cognition is web-y. But the kinds of information restrictions questions impose needn’t be built-in in this way. Different questions restrict different sets of information, and which question is at hand depends a great deal on what the context is. This makes questions great for central cognition, which needs ways of staying tractable without having built-in restrictions on information.
Some other folks—most notably Kahneman and Tversky—think that the mind features heuristics: it stays tractable by using faster, more tractable procedures to approximate slower, less tractable procedures.4 Nothing we’ve said here rules this out. But questions work differently from heuris- tics. Heuristics are meant to be automatic. Questions are not: the agent needs to figure out which question at hand—there is no single question she automatically defaults to. This again makes questions helpful for central cognition, which needs ways of staying tractable while staying flexible and holistic.
Summing up: the Questions-Tractability Hypothesis might explain why question-sensitivity is a feature of central cognition in particular, given the particular way questions reduce computational costs. This would in turn give us an explanation for why the zetetic picture of the mind would be on the right track.
Endnotes
1I should note: this line of work has by and large not engaged with a similar line of work coming from developmen- tal psychology. I find this unfortunate. See e.g., Schulz (2012); Gopnik (2012); Xu (2019).
2For slightly different arguments see also (Koralus and Mascarenhas, 2013; Koralus, 2014, 2023).
3Credit to Kareem Khalifa for the name!
4I should note: this is not the only way of thinking about heuristics. What I say here arguably doesn’t generalize, but it hopefully illustrates why questions would be helpful for central cognition.
Acknowledgments: this project will probably never see the light of day, so I wanted to thank everyone who helped me with it. Thanks to audiences at the first Inquiry Workshop in Nashville, the Eastern APA, and SSPP. I’d also like to thank Kareem Khalifa and Adrienne Prettyman for being awesome commentators. In addition I’d like to thank (including but not limited to) Yong Xin Hui, Kenneth Black, Gareth Norman, Philipp Mayr, Katie Zhou, Selina Guter, Adrian Liu, Akshan deAlwis, Jared Milson, Julia Staffel, Dennis Whitcomb, David Thorstad, Peter van Elswyk, Justin Khoo, Matthias Michel, Kevin Dorst, David Barack, and Eric Mandelbaum.
References
Beddor, B. and Goldstein, S. (2023). A Question-Sensitive Theory of Intention. The Philosophical Quarterly, 73(2):346–378.
Brooke-Wilson, T. (2023). How Is Perception Tractable? Philosophical Review, 132(2):239–292. Dorst, K. and Mandelkern, M. (2021). Good Guesses. Philosophy and Phenomenological Research,
page phpr.12831.
Fodor, J. A. (1983). The modularity of mind an essay on faculty psychology.
Garey, M. R. and Johnson, D. S. (1979). Computers and intractability: a guide to the theory of NP- completeness. A series of books in the mathematical sciences. W. H. Freeman, New York.
Gopnik, A. (2012). Scientific Thinking in Young Children: Theoretical Advances, Empirical Re- search, and Policy Implications. Science, 337(6102):1623–1627.
Groenendijk, J. and Stokhof, M. (1984). Studies on the Semantics of Questions and the Pragmatics of Answers. PhD Thesis, University of Amsterdam.
Hoek, D. (2022). Questions in Action. The Journal of Philosophy, 119(3):113–143. Holgu´ın, B. (2022). Thinking, Guessing, and Believing. Philosophers’ Imprint, 22(1).
Koralus, P. (2014). The Erotetic Theory of Attention: Questions, Focus and Distraction: The Erotetic Theory of Attention. Mind & Language, 29(1):26–50.
Koralus, P. and Mascarenhas, S. (2013). The Erotetic Theory of Reasoning: Bridges Between Formal Semantics and the Psychology of Deductive Inference*. Philosophical Perspectives, 27(1):312–365.
Koralus, P. E. (2023). Reason and inquiry: the erotetic theory. Oxford University Press, Oxford.
Roberts, C. (1996). Information Structure in Discourse: Towards an Integrated Formal Theory of Pragmatics. In Ohio State University Working Papers in Linguistics, volume 49. Ohio State University.
Roberts, C. (2012). Information structure in discourse: Towards an integrated formal theory of pragmatics. Semantics and Pragmatics, 5.
Schaffer, J. and Szabo´, Z. G. (2014). Epistemic comparativism: a contextualist semantics for knowl- edge ascriptions. Philosophical Studies, 168(2):491–543.
Schulz, L. (2012). The origins of inquiry: inductive inference and exploration in early childhood.
Trends in Cognitive Sciences, 16(7):382–389.
Van Rooij, I., Blokpoel, M., Kwisthout, J., and Wareham, T. (2019). Cognition and Intractability: A Guide to Classical and Parameterized Complexity Analysis. Cambridge University Press, 1 edition.
Xu, F. (2019). Towards a rational constructivist theory of cognitive development. Psychological Review, 126(6):841–864.
Yalcin, S. (2018). Belief as Question-Sensitive. Philosophy and Phenomenological Research, 97(1):23–47.