Symposium Introduction
Human beings are limited creatures. They (we!) have short lives, short attention spans, intermittently accurate sensory input, semi-reliable memories and means of communication, and modest capacities for valid or statistically appropriate inference. Moreover, our attempts to acquire and disseminate evidence have costs. Books and subscriptions to all the journals are expensive. And then there’s the simple fact that even when we have it all, the mass of complicated information can be crushing to our intellects. Theories of bounded rationality proceed from the thought that given our limited capacities and resources and the difficulty of the job of figuring things out, there are more and less rational approaches to apportioning those resources to get the best result we can expect.
David Thorstad’s Inquiry Under Bounds is a contribution to the program of epistemology in light of the bounded rationality program. Thorstad has five core insights for theories of epistemic rationality, given the bounded features of human rationality. First, that bounds of limited abilities and costs should figure into evaluations of epistemic rationality. Second, that these evaluations of epistemic rationality should focus on processes of inquiry instead of doxastic results. Third, that given the relevant limitations of humans, fast-and-frugal heuristics for beliefs are rational. Fourth, that the rationality of frugality and heuristics depends on the environments in which they are deployed. And fifth, that these approaches result in significant revisions of how traditional epistemic approaches (what Thorstad calls the “Standard Picture”) evaluate epistemic performances along purely epistemic standards, without practical considerations intervening. Thorstad takes the result, a significant break with the standard practice in epistemology and critical thinking, to be a vindicatory epistemology, arguing that what we’ve thought were violations of good epistemic practice, on the Standard Picture, are in fact rational for humans under bounds. Anchoring biases can be rational, given limited information and few opportunities for gathering further evidence; avoidance of cognitive clutter can be a reason to be averse to pursuing further evidence.
Thorstad closes the book with an account of how his approach merges with the zetetic movement in epistemology. Norms of inquiry, in short, should be indexed to bounds on exercises of rationality, and these norms are not only ones bearing on doxastic attitudes but the practical preferences (and dispreferences) for how we pursue our inquiries and upon where our critical attention should turn.
This is certainly a challenging research program, and we have here collected a variety of voices to contribute to the developing discussion. This, I believe, is certainly worth the investment of time, attention, and effort for those curious about practical issues bearing on norms of cognitive rationality.
5.20.26 |
Response
The Many Norms of Inquiry
It is my pleasure to respond to David Thorstad’s excellent book. Thorstad skillfully integrates the empirical literature on bounded rationality with the recent, burgeoning philosophical literature on inquiry, and I learned a great deal from reading it. In this response, however, I will argue that Thorstad’s commitment to a demanding consequentialist account of the norms of inquiry undermines the prospects for vindicating the rationality of apparently biased inquiry, and that his account would benefit from positing different kinds of important, joint-carving rational norms of inquiry.
Let me begin by recapping Thorstad’s vindicatory ambitions and several key features of his account of rational inquiry.
Thorstad writes, “The program of vindicatory epistemology aims to recast apparent instances of biased cognition as the result of rational inquiry or metacognition” (35). In Chapter 7, Thorstad applies his account of rational inquiry to further this vindicatory program, while aiming to avoid the extreme “Panglossian” view on which humans rarely or never think and act irrationally. While I will omit many details in my discussion here for the sake of space, Thorstad argues that, contrary to initial appearances, the method of anchoring and adjustment—in which agents “estimate a quantity by beginning with an anchor value, then iteratively adjusting their estimate to incorporate new items of information” (137)—is often an efficient way of testing hypotheses or a rational way to trade off accuracy and effort. Similarly, Thorstad argues that apparent irrationalities in logical (especially conditional) reasoning—paradigmatically, in the Wason selection task’s logic puzzle—often reflect rational evidence-gathering processes and good probabilistic reasoning because they result in expected informational gains about probabilistic relationships between propositions.
In Chapter 3, Thorstad defends a maximizing, reasons-responsive, global consequentialist account of the norms governing agency (including, but not limited to, actions). The following principle is central to Thorstad’s account:
Reason-Evaluative Bridge: For all agents S, times t, and features of agency X, S has sufficient possessed reason at t to X just in case S’s X-ing at t is best. (49)
Thorstad specifies this view by arguing that it is right for S to X at t iff S has sufficient possessed reason at t to X, and by defending an information-sensitive (or evidence-relative) account of possessed reasons. Moreover, Thorstad argues, S’s X-ing at t is (ex post) rational iff S correctly responds to S’s possessed reasons to X at t (e.g., by X-ing on the basis of those possessed reasons). Applying this account to inquiry, it is right and (ex post) rational for S to perform inquiring action Y at t iff S’s Y-ing at t is best, and S’s Y-ing at t is (ex post) rational iff S correctly responds to S’s possessed reasons to Y at t.
Thorstad endorses a “rich axiology” on which epistemic goods—such as understanding, knowledge, and accurate belief—are among the relevant goods that ought to be promoted (61–62). Accordingly, Thorstad grants that we can speak of epistemic reasons for performing inquiring actions that promote acquiring epistemic goods. (Perhaps Thorstad would be willing to likewise countenance prudential reasons for inquiring actions that promote prudential goods and instrumental reasons for inquiring actions that facilitate achieving one’s ends.) Importantly, though, in Chapter 4 Thorstad denies that there are any “nontrivial” (71–72) or “joint-carving” (83–84) epistemic norms governing inquiry or that there is a nontrivial, joint-carving sense in which agents’ inquiries can be assessed as epistemically rational or irrational. The only important, joint-carving rational norm governing inquiry, Thorstad seems to hold, is the all-things-considered, global consequentialist norm. One of Thorstad’s arguments against positing nontrivial epistemic norms on inquiry is that such norms do not play any necessary, important theoretical roles (78–81). Thorstad thus invites proponents of epistemic norms of inquiry to explain “the roles that these norms play in our thought, our talk, and our lives” (88).
In what follows, I will accept Thorstad’s invitation by arguing that epistemic norms of inquiry—as well as instrumental and prudential norms of inquiry—can in fact play an important, useful theoretical role. In particular, I propose, they offer resources for furthering Thorstad’s vindicatory program by making space to say that processes of inquiry that are all-things-considered irrational can nonetheless be rational in an important sense.
It’s first worth noting that maximizing consequentialist views like Thorstad’s are famously demanding (especially when construed not only as a theory of morality but also as a theory of rationality). Of course, the fact that Thorstad takes our cognitive bounds to constrain how much rationality requires of us mitigates his view’s demandingness to some degree. Nonetheless, it seems far more demanding than what Thorstad calls the “Standard Picture” of rationality—on which rationality is primarily a matter of having coherent credences and preferences—against which theorists of bounded rationality are reacting. While it may be plausible that agents generally inquire in ways that are instrumentally rational relative to the credence function and utility function that they actually have, it’s a much taller task to show that agents inquire in ways that are rational relative to the credence function and utility function that they have most reason to have. If Thorstad’s global consequentialism is combined with a fairly standard axiology (e.g., one on which the welfare of sentient creatures is of great significance), it seems to imply that, even given our cognitive bounds, agents are rationally forbidden from inquiring about the fairly trivial matters many of us spend our time on (such as the latest celebrity gossip). Rather, agents are rationally required to inquire about how they can do the most good (such as by reading Thorstad’s own excellent blog, “Reflective Altruism”). If we accept Thorstad’s maximizing consequentialist account of rational inquiry, it appears difficult to avoid the verdict that most agents are highly irrational inquirers.
Consequently, it is not clear that Thorstad succeeds in vindicating the claim that most inquirers are generally rational when evaluated with respect to the all-things-considered standards of his global consequentialist account of rationality. Let us grant that Thorstad has shown that anchoring and adjusting can facilitate making fairly accurate estimates with fairly little effort, and that subjects performing the Wason selection task often gather evidence resulting in informational gains. However, this does not demonstrate that agents undertaking these processes of inquiry are all-things-considered responding correctly to their reasons. Instead, it indicates that such agents are boundedly epistemically rational if their inquiries lead them to form fairly accurate credences, boundedly instrumentally rational if their inquiries facilitate achieving their cognitive and practical goals fairly well, or boundedly prudentially rational if their inquiries tend to promote their own well-being. Indeed, other bounded rationality theorists engaged in this vindicatory project (to my knowledge) typically appear to be primarily interested in showing that agents’ processes of inquiry tend to be rational with respect to epistemic, instrumental, prudential, or coherence standards. Consequently, they evaluate agents with respect to a much less demanding standard of rationality than Thorstad’s. And it is because these standards are much less demanding than the standards Thorstad endorses that their vindicatory claims are plausible.
In sum: Thorstad’s arguments make a strong case for the claim that many apparently irrational heuristics and processes of inquiry can in fact be epistemically, instrumentally, or prudentially rational. But on Thorstad’s demanding maximizing consequentialist view, engaging in such processes of inquiry is rational in an important, joint-carving sense only if doing so maximizes expected value. Our inquiries rarely maximize expected value. Thus, Thorstad’s vindicatory project is in trouble.
Of course, it’s compatible with Thorstad’s view to posit trivial, non-joint-carving epistemic, prudential, or instrumental norms of inquiry. But I take it that the vindicatory program aims to vindicate the rationality of many processes of inquiry not merely in a trivial sense but in an important, joint-carving sense. Positing nontrivial epistemic norms of inquiry would provide Thorstad with additional resources for vindicating the claim that agents who engage in reliable, truth-conducive processes of inquiry about questions for which they lack sufficient all-things-considered reason to inquire are nonetheless rational in an important, joint-carving sense. The same kinds of considerations support positing important, joint-carving prudential and instrumental norms of inquiry. So, I submit, agents whose processes of inquiry facilitate acquiring epistemic goods and avoiding epistemic bads are epistemically rational inquirers. Agents whose processes of inquiry promote their own well-being are prudentially rational inquirers. And agents whose processes of inquiry facilitate achieving their cognitive and/or practical goals are instrumentally rational inquirers.
To be clear, I have not denied that there is a demanding yet important all-things-considered normative standard governing inquiry with respect to which most inquirers fare poorly. But I have argued that positing different kinds of norms can help to vindicate some important sense in which agents’ processes of inquiry tend to be rational: namely, that they can be epistemically, prudentially, or instrumentally rational. Rather than accepting a view on which the only important, joint-carving sense of rationality is Thorstad’s all-things-considered global consequentialist sense, then, proponents of the vindicatory program have reason to embrace a plurality of important, joint-carving rational norms of inquiry.
-
5.20.26 |
Reply
A Response to Z Quanbeck
1. Introduction
I would like to thank Quanbeck for a thoughtful discussion of the demandingness objection to the reason-responsive consequentialist view. In this paper, I distinguish two forms of the demandingness objection (section 2) and examine potential responses to each (sections 3–4). I also discuss the attempt to reduce demandingness through rationality pluralism (section 5) and conclude with brief reflections on the demandingness objection (section 6).
2. Two Types of Demandingness
The Demandingness Objection to a normative theory N holds that:
(Demandingness Objection) N is an objectionably demanding normative theory.
It is important to distinguish two forms of the Demandingness Objection.
2.1 Demandingness as Compliance Difficulty
The reason-responsive consequentialist view is a maximizing normative theory. Formulated in terms of reasons, maximizing normative theories hold:
(Maximizing Ought-Reason Link) For all agents S, times t and options o:
(Maximizing Permissions) S may o at t just in case for every alternative o’ to o, S has no more reason at t to o’ than she has at t to o.
(Maximizing Requirements) S must o at t just in case o is the uniquely permissible option for S at t.
The Maximizing Ought-Reason Link implies that agents will often act wrongly. Faced with small sets of options whose merits are easily assessed, they may easily identify the options they have most reason to perform. But faced with larger sets of options whose merits are more difficult to assess, it will be difficult for them to identify the options they have most reason to perform.
Suppose we then interpret demandingness along the following lines:
(Demandingness as Compliance Difficulty) A normative theory is demanding to the degree that it is difficult for agents to determine how to act permissibly according to the theory.
The discussion above suggests that not only maximizing consequentialism, but in fact all maximizing normative theories face the Demandingness Objection given Demandingness as Compliance Difficulty.
2.2 Demandingness as Excessive Sacrifice
There is a second sense in which consequentialist theories have been claimed to be unduly demanding. This is:
(Demandingness as Excessive Sacrifice) A normative theory is demanding to the degree that it often requires agents to make personal sacrifices in order to act permissibly.
On this conception, maximizing consequentialism is often held to be more demanding than other maximizing normative theories. This is because many consequentialists do not countenance agent-centered prerogatives (Scheffler 1994), agent-relative value (Dreier 1993), or other strategies for permitting agents to privilege their own interests over the interests of others.
How should consequentialists respond to the Demandingness Objection?
3. Responding to Demandingness as Compliance Difficulty
The traditional consequentialist response to the Demandingness Objection given Demandingness as Compliance Difficulty is to seek companions in guilt. We saw in section 2.1 that all maximizing normative theories face the Demandingness Objection on this conception. So, too, do many other normative theories. For example, impermissive theories of rational credence require agents to hold precisely one of the continuum-many possible credences toward each proposition they are opinionated about. Agents typically fall short of this standard.
Consequentialists then offer solutions to the Demandingness Objection, while emphasizing that they are not the only ones on the hook for developing solutions. A first move they make is to note that while it is difficult for agents to identify the best options, it is not always difficult for agents to identify good options. Indeed, one of the main projects of Inquiry under Bounds is to show how bounded agents often manage to act well. While this does not salvage the deontic claim that agents can easily identify permissible options in difficult decision problems, it may salvage the axiological claim that they can easily identify good options. We could salvage an analog of the deontic claim if we were willing to adopt scalar theories of rationality and rightness on which acts can be more or less right and rational according to their values (Norcross 1997). Then the claim that agents can often identify good options would imply that agents can often identify acts which are fairly right and rational.
4. Responding to Demandingness as Excessive Sacrifice
Consequentialists will need to make a different response to the Demandingness Objection given Demandingness as Excessive Sacrifice. One traditional option is to hold the line and insist that rightness and rationality are demanding. A second option is to modify consequentialism to reduce its demandingness. While this is not my preferred view, readers who wish to adopt non-maximizing normative theories (Slote 1984), incorporate agent-relative value (Portmore 2011), or agent-centered prerogatives (Scheffler 1994) are welcome to do so and read Inquiry under Bounds with these modifications in mind. A final option is to argue that given Demandingness as Excessive Sacrifice, consequentialism is often less demanding than it may appear (Railton 1984).
Here the case study of longtermism may be instructive. When I arrived at the Global Priorities Institute in 2020, a very earnest community of young longtermists thought that it was overwhelmingly important to determine the best course for humanity’s long-term future. Moments not spent inquiring into these matters were deeply regretted. Large expenses including business-class flights and downtown offices were justified as a way of improving the productivity of inquirers. Some went so far as to argue that every trivial decision, including what to eat for breakfast, should be made primarily on the basis of its expected impact on the long-term future of humanity (Cotton-Barratt 2021).
A theme of my work has been that these longtermist practices may be a mistake. I have argued that longtermists overestimate both the magnitude of future risks (Thorstad 2024a, c) and the importance of addressing them (Thorstad 2023, 2024b). I have also argued that even if we grant the importance of acting in some domains with a focus on improving the long-term future, it does not follow that these concerns should steer all aspects of our lives (Thorstad forthcoming). It is my hope that these reflections will help some members of the longtermist community to make substantive changes to the inquiries that they conduct and the role of those inquiries in their lives.
I argue in Chapter 6 that the reason-responsive consequentialist view has a related virtue. Namely, it grounds principles of clutter avoidance (Friedman 2018; Harman 1986), including a prohibition against engaging in trivial inquiries such as counting grass. Many theorists studying the epistemology of inquiry have taken clutter avoidance to be a plausible principle and thought that it is essential for a theory of rational inquiry to ground principles of clutter avoidance. I am not sure what this implies for celebrity gossip. Likely, the answer depends on the knowledge and entertainment derived from reading celebrity gossip. But certainly it is no defect in a theory of inquiry that the theory prohibits trivial inquiries, whatever those turn out to be.
5. Rationality Pluralism
Quanbeck suggests that we can reduce the demandingness of rational inquiry by adopting rationality pluralism. Rationality pluralists posit distinctive practical, moral, and instrumental species of rationality alongside the all-things-considered demands of rationality simpliciter. I do not think that the Demandingness Objection motivates rationality pluralism.
On Demandingness as Compliance Difficulty, there is no clear reason to expect that it should be much easier for agents to determine which of a given set of options is best-supported by their practical, moral, or instrumental reasons than by their all-things-considered reasons. Even if we hold, for example, that an agent’s practical reasons are a subset of her all-things-considered reasons, it does not follow that these reasons are significantly smaller in number, nor that assessing the balance of the smaller set of reasons is easier than assessing the balance of the larger set of reasons. Sometimes, it may be quite clear what an agent has most all-things-considered reason to do, but less clear what she should do on posited species of rationality that remove some of her strongest all-things-considered reasons.
On Demandingness as Excessive Sacrifice, positing indexed species of rationality is likely to make the Demandingness Objection worse. Now it will turn out that there are often indexed species of rationality on which agents are required to act against the balance of their all-things-considered reasons. This is always a sacrifice, and in many cases it may be quite a large sacrifice. I could learn more if I followed Derek Parfit’s habit of eating cold meals and reading philosophy even while brushing my teeth. On this basis, it might be held that I am required by epistemic rationality to do these things. For my own part, I would rather eat a good meal and brush my teeth in peace.
6. Conclusion
This paper distinguished two forms of the Demandingness Objection: Demandingness as Compliance Difficulty and Demandingness as Excessive Sacrifice. In response to Demandingness as Compliance Difficulty, we saw that consequentialists can seek companions in guilt and offer scalar notions of value or even perhaps scalar notions of rationality and rightness. In response to Demandingness as Excessive Sacrifice, we saw that consequentialists can hold the line, modify their view, or argue that consequentialism demands less sacrifice than typically thought.
More importantly, we saw that neither form of the Demandingness Objection provides a strong motivation for rationality pluralism. While there are many traditional solutions to the Demandingness Objection, rationality pluralism is unlikely to resolve Demandingness as Compliance Difficulty and probably exacerbates Demandingness as Excessive Sacrifice. This suggests that the Demandingness Objection, while an important objection to consequentialism, does not motivate rationality pluralism
References
Cotton-Barratt, Owen. “Everyday Longtermism.” EA Forum, 2021. https://forum.effectivealtruism.org/posts/3PmgXxBGBFMbfg4wJ/everyday-longtermism.
Dreier, James. “Structures of Normative Theories.” The Monist 76, no. 1 (1993): 22–40.
Friedman, Jane. “Junk Beliefs and Interest-Driven Epistemology.” Philosophy and Phenomenological Research 97, no. 3 (2018): 568–83.
Harman, Gilbert. Change in View. Cambridge, MA: MIT Press, 1986.
Norcross, Alastair. “Good and Bad Actions.” Philosophical Review 106, no. 1 (1997): 1–34.
Portmore, Douglas. Commonsense Consequentialism: Wherein Morality Meets Rationality. Oxford: Oxford University Press, 2011.
Railton, Peter. “Alienation, Consequentialism, and the Demands of Morality.” Philosophy and Public Affairs 13, no. 2 (1984): 134–71.
Scheffler, Samuel. The Rejection of Consequentialism. Oxford: Oxford University Press, 1994.
Slote, Michael. “Satisficing Consequentialism.” Proceedings of the Aristotelian Society, Supplementary Volume 58 (1984): 139–63.
Thorstad, David. “High Risk, Low Reward: A Challenge to the Astronomical Value of Existential Risk Mitigation.” Philosophy and Public Affairs 51, no. 4 (2023): 373–412.
Thorstad, David. “Against the Singularity Hypothesis.” Philosophical Studies (2024a).
Thorstad, David. “Mistakes in the Moral Mathematics of Existential Risk.” Ethics 135, no. 1 (2024b): 122–50.
Thorstad, David. “What Power-Seeking Theorems Do Not Show.” Global Priorities Institute Working Paper 27, 2024c. https://globalprioritiesinstitute.org/what-power-seeking-theorems-do-not-show-david-thorstad/.
Thorstad, David. “The Scope of Longtermism.” Forthcoming in Australasian Journal of Philosophy.
5.25.26 |
Response
Inquiry Under Bounds and Epistemic Purism
Abstract
David Thorstad’s Inquiry Under Bounds offers a compelling defense of consequentialist zetetic norms. This paper explores how his view, which endorses consequentialism for guiding inquiry while aiming to screen off doxastic norms from radical pragmatic and moral encroachment, can be further supported by an appeal to the nature of belief. I argue that the mind-to-world direction of fit inherent in belief introduces special fittingness considerations that justify purely evidential norms for belief. However, I then contend for a moderate degree of pragmatic encroachment on epistemic predicates, suggesting that while the first-order content of justification remains purely evidential, second-order considerations—such as how much evidence is “enough”—can be influenced by pragmatic or moral reasons within different contexts of inquiry. This approach allows for a rich axiology to inform inquiry without jeopardizing the distinctively epistemic character of justification and knowledge.
Introduction
David Thorstad has given us a robust defense of consequentialist zetetic norms: i.e., consequentialism for guiding inquiry. The view is very appealing. In particular, Thorstad succeeds in his aim of showing us how to argue that, when cognitive bounds are considered, many common deviations from standard, formal theories of rationality are rationally defensible heuristics without being stuck saying that all such deviations are defensible.
If zetetic norms are value-maximizing, then this threatens doxastic norms with radical pragmatic and moral encroachment, contrary to widely endorsed purely evidential norms for epistemic predicates. Accordingly, Thorstad aims to screen off the doxastic from the broader consequentialism he endorses for the zetetic. I want to explore how this screening off can be argued for from the nature of belief, then defend a moderate degree of pragmatic encroachment on epistemic predicates.
Consequentialisms
A direct consequentialist about X applies the standard of value-maximizing directly to X. An indirect consequentialist about X applies the standard of value-maximization to some Y, then evaluates X in relation to Y. For example, a kind of rule consequentialist applies the standard of value-maximizing directly to rules and indirectly to acts; an act is right insofar as it follows the optimific rule. The idea is that consequentialism applies directly at one level, indirectly at others.
A global consequentialist is a direct consequentialist about everything. Because everything is directly normed by the good, accordingly, it makes sense to say “focal points” rather than “levels.” This view implies that there will often be tension between levels/focal points. Confronted with a manifest case of hate speech that causes harmful distress, creates an atmosphere that encourages violence, and makes no contribution to good-faith, alethic discourse, the Millian magistrate cannot perform both the most optimific act and follow the most optimific general principle of justice. Is level tension a feature or a bug? Answers may depend on whether one expects a moral theory to disclose or resolve dilemmas inherent in the human condition.
Zetetic and Doxastic Norms
Thorstad endorses an attenuated version of global consequentialism (Thorstad 2024, sec. 3.2). The view restricts consequentialism to predicates that are under agentive control and not involving special considerations of fittingness. As an example of a predicate that is restricted by considerations of fittingness, Thorstad offers “blameworthiness.” We shouldn’t assign blame when it is not fitting, even if assigning blame is optimific; e.g., we shouldn’t blame the innocent to appease a mob.
Inquiry is the act of gathering evidence with the goal of reaching conclusions. Acts are under agentive control, if anything is, and there are no special fittingness considerations that restrict the range of values used to norm acts. Accordingly, the acts involved in inquiry are subject to being directly normed by the good, and with a sufficiently rich axiology we can obtain reflective equilibrium on cases typically discussed in the inquiry literature. Belief, on the other hand, is characterized as a representational mental state with a mind-to-world direction of fit, which introduces special fittingness considerations that may be used to refine Thorstad’s argument. Just as blame is only fitting for the guilty, belief is only fitting for evidentially supported propositions.
Thorstad references two theoretical roles for belief to support distinctively epistemic norms for belief. Direction of fit supports the “role of epistemic justification in the analysis of knowledge” (Thorstad 2024, sec. 4.6.1) and concisely connects to the fittingness consideration (Thorstad 2024, sec. 3.2.2). On this view, the actions involved in inquiry are normed by a complete axiology, but because of the special fittingness considerations for beliefs there are purely epistemic norms for belief, focused on evidential justification linking mind to world.
Encroachment
The view that the zetetic is normed by a rich axiology but that the doxastic is normed by specifically alethic values supports a sharp line against pragmatic and moral encroachment on or full-blown pragmatism about belief. Calling a belief justified on account of non-evidential reasons, because it supports some practical good, is like calling a gutter ball a good shot because the opponent’s delighted fans are more numerous. Their delight is good, but it’s not fitting to call a shot a good shot for reasons that don’t pertain to its hitting its mark. Belief’s mind-to-world direction of fit gives it a mark to hit, so to speak, and the justification factor of knowledge involves the agent possessing evidential reasons for thinking that the belief succeeds.
Let’s call “full-blown pragmatism” the view that there can be non-evidential reasons for belief. According to this view, if believing you’ll hit a home run makes it more likely that you will, then you have a reason to believe that you will—even if it is still not especially likely. Opponents of this view will say that this sort of reason may be a reason to act in ways that foster useful delusion, but that is not the same as a reason for a belief. Clearly, if the reason the delusional belief is useful is that it fosters confidence, and if confidence without delusion is psychologically possible, then we could prefer swagger without delusion. For many philosophers, these reflections defease the intuitive case for full-blown pragmatism.
If doxastic norms are subsidiary to zetetic norms and zetetic norms are directly consequentialist, then there would seem to be another path to full-blown pragmatism. Thorstad finds this sort of indirect consequentialism about rational belief in Herbert Simon (Simon 1976). Accordingly, a belief is rational if and only if it is the product of a rational process of inquiry. If this is correct and zetetic norms are consequentialist, then a rational process of inquiry that takes into consideration our cognitive bounds may produce a rational belief, even when it tells us to halt inquiry before considering all of the evidence available to us, since fully consequentialist norms for inquiry may imply that the cost of continuing to gather evidence is not worth the benefit (Thorstad 2024, sec. 9.4).
Contrary to Simon’s indirect consequentialist approach, global consequentialism with fittingness-restricted axiology for belief norms allows us to endorse pragmatic and moral, as well as epistemic, norms for inquiry without jeopardizing the distinctively epistemic character of justification, knowledge, etc. Reflecting on direction of fit and fittingness bolsters the case Thorstad makes for this view. These considerations are core enough to belief that something like “has enough evidence” should be a necessary condition for justified belief. The first-order content of justification is purely evidential, so to speak.
But there remain second-order considerations: How much is “enough”? Infallibilists think the answer is “enough to rule out any doubt.” Fallibilists worry that then we are left with a concept that is not very useful, one that has a contingently empty (or very small) extension. We want our epistemic predicates to do certain things for us, even in the process of inquiry. For example, we want to identify justified believers as knowing informants. We also use knowledge as input and output conditions for inquiry. Isaac Levi identified knowledge as a standard for serious possibility in inquiry (Levi 1980), and when we take ourselves to know that p, having obtained “enough” evidence, we consider ourselves unreasonable to continue to inquire whether p.
We can agree that only evidence counts as a first-order reason to believe while maintaining that in different contexts, including different contexts of inquiry, we may have pragmatic or moral reasons to give different answers to “How much is enough?” in different contexts. The idea here is very similar to Ralph Wedgwood’s suggestion that there are a range of precisifications of norms for belief that respect the core of the concept of rational belief (Wedgwood 2002, 275).
In the context of a physics seminar, for example, the fecundity of presupposing the temporal and spatial uniformity of nature gives us a reason to give a less demanding answer to the enough question than we might give in an epistemology seminar. The point isn’t that the fecundity of a proposition is a reason that it’s true—only evidence provides alethic reasons—but rather that it’s a reason not to adopt an overly strict understanding of “enough.” Familiar cases from the pragmatic encroachment literature, such as Fantl and McGrath’s train case, may be framed as varying the stringency of “enough” and likewise for moral encroachment (Fantl and McGrath 2002, 67).
Summing up, my proposal is that Thorstad’s argument, bolstered by consideration of mind-to-world direction of fit, does block full-blown pragmatism, even while allowing consequentialism about inquiry. Nonepistemic reasons are not first-order reasons to believe, accordingly, but fallibilist conditions for justification raise the second-order question “How much is enough?” and how we answer this question may be subject to pragmatic and moral reasons, yielding a moderate impurism consistent with Thorstad’s overall project.
References
Fantl, Jeremy, and Matthew McGrath. 2002. “Evidence, Pragmatics, and Justification.” Philosophical Review 111: 67–94.
Levi, Isaac. 1980. The Enterprise of Knowledge: An Essay on Knowledge, Credal Probability, and Chances. Cambridge, MA: MIT Press.
Simon, Herbert A. 1976. “From Substantive to Procedural Rationality.” In 25 Years of Economic Theory: Retrospect and Prospect, 65–86. Boston: Springer.
Thorstad, David. 2024. Inquiry Under Bounds. Oxford: Oxford University Press.
Wedgwood, Ralph. 2002. “The Aim of Belief.” Philosophical Perspectives 16: 267–97. https://doi.org/10.1111/1468-0068.36.s16.10.
-
5.25.26 |
Reply
A Response to Jeremy Shipley
1. Introduction
I would like to thank Shipley for a thoughtful response to Inquiry under Bounds. As Shipley notes, this response is fully compatible with the view defended in Inquiry under Bounds.
2. My View
Inquiry under Bounds defends a reason-responsive consequentialist view of rational inquiry. This view has three main components. The first is a global consequentialist theory of rightness, on which:
(Global Consequentialism) For all features of agency X:
(Deontic-Evaluative Bridge) X is right if and only if X is best.
(Promotion) The value of X is determined by the goodness promoted by X.
The second is a reason-responsiveness theory of rationality, which links rightness to rationality through the relation Z of reason-responsiveness:
(Rationality-Rightness Bridge) For all agents S, times t and features of agency X, S’s X-ing at t is rational if and only if:
(RB1) S’s X-ing at t is right, and
(RB2) S’s X-ing at t bears Z to the normative reasons she possesses at t.
The third is an information-sensitive theory on which an agent’s reasons are assessed not against her beliefs or against the facts, but rather against her total evidence.
3. Alternative Views
As Shipley notes, the reason-responsive consequentialist view does not force us to revise traditional views of rational belief so long as we adopt a strict level separation between questions about rational belief and rational inquiry. However, as Shipley notes, the reason-responsive consequentialist view also does nothing to discourage us from revising traditional views of rational belief on other grounds.
In particular, Shipley notes that the reason-responsive consequentialist view is consistent with:
(Full-Blown Pragmatism) Some reasons for belief are non-evidential.
The reason-responsive consequentialist view is also consistent with Shipley’s preferred view, which we might call Second-Order Pragmatism:
(Second-Order Pragmatism)
(First-Order Evidentialism) All reasons for belief are evidential.
(Second-Order Encroachment) The strength of reasons needed to support belief in a proposition can be affected by non-evidential factors.
Shipley offers two examples of Second-Order Encroachment. First, the strength of reasons needed to support belief in a proposition may increase with the cost of error. Second, the strength of reasons needed to support belief in a proposition may decrease with its explanatory fecundity.
4. Evaluating Second-Order Pragmatism
The dialectic surrounding Second-Order Pragmatism is by now well known. I would invite readers to make up their minds about Second-Order Pragmatism by working through traditional arguments for the view as well as traditional objections.
On stakes-sensitivity, critics of Second-Order Pragmatism have objected that it is difficult to formulate a conception of stakes that preserves the pragmatist’s desired connection between rational and practically adequate belief (Anderson and Hawthorne 2019). They have argued that stakes-sensitivity leads the rationality of belief to be shifted in seemingly counterintuitive ways (Eaton and Pickavance 2015). And they have argued that practical stakes bear not on the rationality of belief, but instead on the rationality of related attitudes such as acceptance (Van Fraassen 1984), endorsement (Fleisher 2018), or hypothesis (Palmira 2020).
On explanatory fecundity, some have likewise argued that the explanatoriness of a scientific theory provides reasons for acceptance rather than belief (Van Fraassen 1980), although this proposal is more controversial (Horwich 1991). Others have argued that explanatory fecundity provides evidence for the truth of a proposition (McCain and Poston 2023).
There are well-known replies to these concerns, and it is not my intention to resolve the debate over Second-Order Pragmatism here. Those who find themselves attracted to Second-Order Pragmatism will find their views fully compatible with the views expressed in Inquiry under Bounds. They are welcome to bring these views to their reading of the book.
References
Anderson, Charity, and John Hawthorne. “Knowledge, Practical Adequacy, and Stakes.” Oxford Studies in Epistemology 6 (2019): 234–57.
Eaton, Daniel, and Timothy Pickavance. “Evidence against Pragmatic Encroachment.” Philosophical Studies 172, no. 12 (2015): 3135–43.
Fleisher, Will. “Rational Endorsement.” Philosophical Studies 175, no. 10 (2018): 2649–75.
Horwich, Paul. “On the Nature and Norms of Theoretical Commitment.” Philosophy of Science 58, no. 1 (1991): 1–14.
McCain, Kevin, and Ted Poston. “Explanation and Evidence.” In Routledge Handbook of the Philosophy of Evidence, edited by Maria Lasonen-Aarnio and Clayton Littlejohn, 329–41. New York: Routledge, 2023.
Palmira, Michele. “Inquiry and the Doxastic Attitudes.” Synthese 197, no. 11 (2020): 4947–73.
Van Fraassen, Bas. The Scientific Image. Oxford: Clarendon Press, 1980.
Van Fraassen, Bas. “Belief and the Will.” Journal of Philosophy 81, no. 5 (1984): 235–56.
Alexandra Peláez
Response
What Makes Inquiry Distinctive
David Thorstad’s Inquiry Under Bounds offers an ambitious and thought-provoking account of rational inquiry, challenging a widely held assumption in epistemology: that inquiry, like belief, is governed by epistemic norms. Given the state of the literature and the zetetic turn in epistemology, one of the book’s most striking contributions is its argument that inquiry is best understood as a practical decision-making process rather than as an epistemic activity with its own distinct aims and norms. Thorstad develops this view by emphasizing the role of bounded rationality and procedural reasoning in shaping inquiry.
Bounded Rationality—Rational inquiry must account for human cognitive limitations, such as time constraints, limited memory, and effort costs.
Procedural Rationality—Rationality is not just about having correct attitudes (e.g., true beliefs) but also about making good procedural choices given these constraints.
From these foundations, Thorstad develops a consequentialist approach to rational inquiry, arguing that it is not constrained by distinctively epistemic norms but instead governed by all-things-considered rationality.
Thorstad defends epistemic nihilism about inquiry—the view that epistemic norms apply only to belief, not to inquiry itself. His case for this view unfolds through a series of arguments rejecting standard motivations for positing epistemic norms of inquiry. First, he argues that inquiry, unlike belief, is an activity that is often driven by non-epistemic reasons (72–75). For example, one might inquire about the weather to gain knowledge, but also to plan a picnic. Since inquiry can be pursued for practical reasons, he suggests that epistemic norms are unnecessary for explaining inquiry’s structure. Second, he appeals to linguistic evidence, noting that we often evaluate inquiries in pragmatic terms. For example, we might say something like, “You shouldn’t look into that conspiracy theory, it’s a waste of time.” Because evaluations of inquiry tend to reference practical rather than epistemic reasons, Thorstad takes this as evidence that inquiry is constrained by practical rather than epistemic norms (75–78). Third, Thorstad argues that epistemic norms for inquiry do no essential theoretical work (78–81). Since we can fully explain how inquiry functions (e.g., how it contributes to doxastic deliberation) without appealing to epistemic norms, we lack good reasons for positing their existence. The idea is that if we already have a framework that accounts for why agents engage in inquiry, adjust their beliefs, and deliberate responsibly, there is no explanatory gap that epistemic norms would need to fill.
More generally, it seems important to Thorstad that inquiry lacks a constitutive aim. He argues against views that claim it does. He considers the view that inquiry aims at settling questions by acquiring knowledge but argues that this model does not explain how bounded agents should handle tradeoffs in inquiry (100). If instead inquiry had knowledge as its constitutive aim, it would yield the wrong pattern of tradeoffs (101). Similarly, Thorstad argues that a knowledge-aimed model struggles to explain why agents should sometimes rely on heuristics, which may fall short of knowledge but are often the most efficient means of inquiry given real-world constraints (101). Epistemic norms for belief seem to stem from a constitutive epistemic aim, like accuracy. Since no such aim structures inquiry, Thorstad concludes, inquiry is not subject to epistemic norms.
Thorstad offers a compelling account of how practical considerations shape inquiry, but it leaves open an important question: What—if anything—makes inquiry distinctive? What differentiates it from other kinds of practical reasoning?
Even if rational inquiry involves tradeoffs between epistemic, cognitive, and practical considerations, it does not follow that inquiry lacks distinctive norms or internal structure. One way to see this is by considering what happens once inquiry is underway. An agent might have good practical reasons to start or abandon an inquiry—investigating a conspiracy theory might be a waste of time, or continuing research on a topic might not be worth the effort. But once an inquiry is underway, it is subject to constraints that are not merely practical. Consider someone who does decide to investigate a conspiracy theory. Inquiring may be a mistake all-things-considered; however, the inquiry itself can still be meaningfully described as good or bad depending on, e.g., how well the inquirer tracks reliable sources, weighs counterevidence, and applies reasoning standards. These constraints govern how inquiry itself is evaluated, independent of whether it is a good idea to inquire in the first place.
Suppose an agent is trying to determine which cereal to buy. It is natural to say that their inquiry ought to be structured according to external constraints—time constraints, importance of arriving at the right answer, cognitive resources. It also seems right to say that their inquiry should be structured according to internal constraints, which do not depend on practical (or external) considerations. For example, if they are considering which scissors to buy, their inquiry has gone wrong, not due to practical inefficiencies, but because it violates an internal constraint of relevance. Inquiry is structured by relations between questions and possible answers—by standards of evidence and logic. Failing to follow those relations or adhere to those standards seems to be a failure in inquiry, not just a failure in decision-making. The fact that inquirers often rationally rely on heuristics and approximations does not mean that inquiry lacks distinctive norms or internal structure—it simply means that bounded agents rationally settle for good-enough answers.
Thorstad’s Inquiry Under Bounds raises rich and productive questions about the nature of inquiry, the role of epistemic aims, and how rationality operates under constraints. Thorstad’s model captures an important part of inquiry: it is sensitive to practical considerations and often shaped by real-world constraints. However, a reader might be left asking what makes an inquiry successful or unsuccessful as an inquiry—not just as a rational decision. If inquiry is just a process of managing tradeoffs, how can it be distinctive? This remains an open and fruitful question—one that Inquiry Under Bounds helps bring into sharper focus.
5.18.26 | David Thorstad
Reply
A Response to Alexandra Peláez
1. Introduction
I would like to thank Pelaez for a thoughtful review that raises important questions about the relationship between epistemic and constitutive norms. In this paper, I clarify what Inquiry under Bounds does and does not claim about epistemic normativity (section 2) and the relationship between practical and theoretical inquiry (section 3). I conclude with discussions of the relationship between constitutive and epistemic norms (section 4.1) and the grounds on which the comparative merits of constitutivist and non-constitutivist views may be assessed (section 4.2).
2. Clarifying the Dialectic
Chapter 4 of Inquiry under Bounds argues that we lack adequate reason to posit a distinctively epistemic sense in which inquiries count as rational, justified, or as inquiries that agents should or ought to engage in. Pelaez disagrees.
On my best reading, Pelaez makes three proposals.
(Constitutive Aim) Inquiry is an activity with a constitutive aim.
(Constitutive Norm) Inquiry is an activity governed by constitutive norms.
(Attributive Value) Inquiries can be good or bad qua inquiry in a sense that does not reduce to the application of a domain-general notion of value to the activity of inquiry.
To assess whether these proposals are in tension with my view, we must be careful to specify exactly what is being claimed. While I do not hold any of the three views proposed by Pelaez, as stated, many or all of them may be compatible with the view defended in Inquiry under Bounds.
3. Constitutive Aim
To say that inquiry has a constitutive aim is not yet to say anything about rationality, justification, or how agents should inquire. On a popular personal reading of Constitutive Aim, the claim is that:
(Constitutive Aim—Personal Reading) There is some goal G such that:
(Agents Aim) All inquiring agents aim at G.
(Aims Constitute Inquiry) Aiming at G is part of what constitutes an agent’s activity as an act of inquiry.
The personal reading of Constitutive Aim is no more in tension with my view than is the popular claim that all inquiring agents must hold one of a special class of interrogative attitudes (Friedman 2017).
My view may be in tension with an alternative metaphysical reading of Constitutive Aim on which:
(Constitutive Aim—Metaphysical Reading) There is some goal G such that:
(Activities Aim) It is in the nature or essence of inquiry to aim at G.
While the immediate statement of the metaphysical reading is not in tension with my view, some interpretations of Activities Aim may imply normative claims that will be in tension with my view. This leads us to our next claim.
4. Constitutive Norm
To say that inquiry is an activity governed by constitutive norms or internal standards is not yet to make a claim about rationality, justification, or how agents should or ought to inquire. It is not yet to make a definite normative claim of any kind. Constitutive Norm can be further specified by indicating the type of normative claim being made.
We will see below that some ways of cashing out Constitutive Norm, for example as claims about attributive value, are not in tension with my view. Strictly speaking, Constitutive Norm would not be in tension with my view even if it were interpreted as the claim that there is an inquiry-specific sense of rationality, justification, or how agents should or ought to act, which answers only to the constitutive aim of inquiry. My view could still be read as an account of rationality, rather than the newly posited inquiry-specific species of rationality, and my view could be used in all applications to which it is put in Inquiry under Bounds. However, I have elsewhere argued that many views of this sort struggle to recover plausible normative claims about rationality (Thorstad 2023), and I argue in Chapter 5 of Inquiry under Bounds that they struggle to account for plausible normative claims about bounded rationality.
5. Attributive Value
To say that inquiry may be good or bad qua inquiry is to make an axiological claim. Inquiry under Bounds is committed to only one strong axiological claim, namely the consequentialist dictum that:
(Promotion) For all features of agency X, the value of X is determined by the goodness promoted by X.
Promotion is a claim about the domain-general notion of value. Because attributive value is not determined by applying the domain-general notion of value to the activity of inquiry, Attributive Value is compatible with Promotion. Many constitutivists within epistemology even interpret constitutive norms so that they satisfy an analog of promotion:
(Attributive Promotion) For all activities X and tokens x of X, the value of x qua X is determined by the degree to which x promotes achievement of the aim of X.
Attributive Promotion is not necessary to restore compatibility with my view, but Attributive Promotion is a welcome point of consensus for those willing to adopt it.
6. Practical and Theoretical Inquiry
Traditionally, the division between practical and theoretical inquiry is a division between two types of activities. These activities take different inputs and return different outputs. Theoretical inquiry takes questions as inputs and outputs doxastic attitudes toward possible answers to its focal question. Practical inquiry takes option sets as inputs and outputs an intention to take one option in the set.
When Pelaez glosses my view as the claim that theoretical inquiry is a species of practical inquiry, I think that Pelaez has a different distinction in mind. On this view, the distinction between practical and theoretical inquiry is driven by an underlying normative distinction between practical and theoretical norms. Any activity governed by practical norms is classified as practical inquiry, even if it begins with a question and concludes with doxastic attitudes toward possible answers to that question. I do not prefer to gloss my view as practical in this sense either, since on my view the norms governing inquiry are all-things-considered and respond to the value of traditional epistemic concerns such as knowledge and understanding. However, the main point to note here is that the gloss is nontraditional.
This distinction matters because one of Pelaez’s main examples concerns an agent trying to determine which cereal to buy. On the traditional gloss, this agent is engaged in practical inquiry. Although dissent from orthodoxy is possible (Singer and Aronowitz 2022), it is not typically held that norms of practical inquiry, whether constitutive or not, are epistemic norms. Hence the claim that practical inquiry is governed by constitutive norms such as relevance does not provide evidence that these norms are epistemic. If anything, it provides evidence for the opposite conclusion.
7. Constitutivism and Epistemic Norms
There is a large literature on constitutive norms not only within the epistemology of inquiry (Friedman 2024; Kelp 2021a, b), but also within broader discussions in ethics (Korsgaard 2009; Velleman 2000) and epistemology (Maguire and Woods 2020; Kelp and Simion 2021). In particular, while some have defended constitutivist accounts of inquiry (Kelp 2021a, b), others have argued against them (Friedman 2024). For the most part, I think that existing arguments will be more comprehensive and persuasive than anything I can say in a short discussion.1 At the same time, two points are worth emphasizing.
7.1. Constitutivism and Epistemic Normativity
The first point is that there is no clear path from constitutivism to the existence of a distinctive class of epistemic norms governing inquiry. Constitutivists could hold that:
(Constitutive Implies Epistemic) All constitutive norms of theoretical inquiry are epistemic norms.
However, Constitutive Implies Epistemic is not plausible. After all, theoretical inquiry is an activity. Most constitutivists hold that activities are governed by constitutive norms such as the categorical imperative (Korsgaard 2009) or encountering and overcoming resistance to one’s ends (Katsafanas 2013), which are not epistemic norms in any traditional sense.
It may turn out that some of the constitutive norms governing theoretical inquiry are the same norms that theorists would like to call epistemic norms. That is not my view. But more to the point, if we deny that Constitutive Implies Epistemic, then the fact that these norms are constitutive of inquiry does not give us any reason to posit a distinctively epistemic species of normativity of which they are a part. This is particularly the case when, as we saw in the example of cereal-buying, similar constitutive norms are claimed to govern practical inquiry. Then it seems more appropriate to treat constitutivists as having made an argument for the existence of constitutive norms rather than epistemic norms.
7.2. Constitutivism and Explanatory Adequacy
To a large extent, I think that the debate between constitutivists and non-constitutivists should be settled on explanatory grounds. The leading constitutivist account of inquiry, due to Christoph Kelp (2021a, b), puts forward a number of explanatory consequences of the constitutivist view. For example, Kelp holds that constitutivism responds to traditional skeptical arguments, explains the distinctive value of knowledge and understanding, and provides an attractive picture on which all of epistemology is the theory of inquiry. Kelp proposes that readers judge the merits of constitutivism by the plausibility and fruitfulness of its explanatory consequences.
Inquiry under Bounds also calls for the reason-responsive consequentialist view to be judged on its explanatory merits. There and elsewhere, I have argued among other things that the approach grounds and clarifies traditional views about bounded rationality (Thorstad 2024b, c), resolves longstanding paradoxes in the theory of bounded rationality (Thorstad 2022), allows us to retain the existence and importance of traditional epistemic norms (Thorstad 2021, 2024b), and gives plausible treatments of tradeoffs between coherence and competing goals such as accuracy (Thorstad 2024a) and the benefits of complex cognition (Thorstad 2025). I would invite readers to judge my view by their assessment of the plausibility and fruitfulness of its explanatory consequences.
References
Friedman, Jane. “The Aim of Inquiry?” Philosophy and Phenomenological Research 108, no. 2 (2024): 506–23.
Friedman, Jane. “Why Suspend Judging?” Noûs 51, no. 2 (2017): 302–36.
Katsafanas, Paul. Agency and the Foundations of Ethics: Nietzschean Constitutivism. Oxford: Oxford University Press, 2013.
Kelp, Christoph. Inquiry, Knowledge, and Understanding. Oxford: Oxford University Press, 2021a.
Kelp, Christoph. “Theory of Inquiry.” Philosophy and Phenomenological Research 103, no. 2 (2021b): 359–84.
Kelp, Christoph, and Mona Simion. Sharing Knowledge: A Functionalist Account of Assertion. Cambridge: Cambridge University Press, 2021.
Korsgaard, Christine. Self-Constitution: Agency, Identity, and Integrity. Oxford: Oxford University Press, 2009.
Maguire, Barry, and Jack Woods. “The Game of Belief.” Philosophical Review 129 (2020): 211–49.
Singer, Daniel J., and Sara Aronowitz. “What Epistemic Reasons Are For: Against the Belief-Sandwich Distinction.” In Meaning, Decision, and Norms: Themes from the Work of Allan Gibbard, edited by Billy Dunaway and David Plunkett, 74–94. Ann Arbor: Michigan, 2022.
Thorstad, David. “The Accuracy-Coherence Tradeoff in Cognition.” British Journal for the Philosophy of Science 75 (2024a): 695–715.
Thorstad, David. “Against the Newer Evidentialists.” Philosophical Studies 180 (2023): 3511–32.
Thorstad, David. “The Complexity-Coherence Tradeoff in Cognition.” Mind 134, no. 534 (2025): 422–57.
Thorstad, David. “Inquiry and the Epistemic.” Philosophical Studies 178 (2021): 2913–28.
Thorstad, David. Inquiry under Bounds. Oxford: Oxford University Press, 2024b.
Thorstad, David. “Reply to Commentators.” Forthcoming in Philosophical Studies.
Thorstad, David. “Two Paradoxes of Bounded Rationality.” Philosophers’ Imprint 22 (2022): 1–16.
Thorstad, David. “Why Bounded Rationality (in Epistemology)?” Philosophy and Phenomenological Research 108 (2024c): 396–413.
Velleman, David. The Possibility of Practical Reason. Oxford: Oxford University Press, 2000.
See Thorstad (forthcoming) for a slightly more comprehensive treatment.↩