Abstract
How should you decide what to do when you’re uncertain about basic normative principles? A natural suggestion is to follow some “second-order” norm: e.g., obey the most probable norm or maximize expected choiceworthiness. But what if you’re uncertain about second-order norms too—must you then invoke some third-order norm? If so, any norm-guided response to normative uncertainty appears doomed to a vicious regress. This paper aims to rescue second-order norms from the threat of regress. I first elaborate and defend the claim some philosophers have made that the regress problem forces us to accept normative externalism, the view that at least one norm is incumbent on all agents regardless of their normative beliefs. But, I then argue, we need not accept externalism about first-order norms, thus closing off any question of how agents should respond to normative uncertainty. Rather, we can head off the threat of regress by ascribing external force to a single second-order norm: the enkratic principle.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Versions of MFT are defended by Gracely (1996) and Gustafsson and Torpman (2014). MFO is considered and rejected by Lockhart (2000) (under the name “PR2”), Gustafsson and Torpman (2014), and MacAskill and Ord (2020). Expectational views are defended by Lockhart (2000), Ross (2006), Sepielli (2009), MacAskill and Ord (2020), and Riedener (2020), among others. Other metanorms have been proposed by Guerrero (2007), Nissan-Rozen (2012), MacAskill (2016), Tarsney (2018, 2019), and Greaves and Cotton-Barratt (2024), among others.
These notions should be understood very thinly. To say that an option is “permissible” just means that it’s possible for an agent who is in some relevant sense normatively ideal (e.g. fully rational and, in the case of objective norms, fully informed) to choose that option.
I will hereafter use “beliefs” to mean “beliefs and/or evidence”, remaining neutral on whether the true subjective norms are sensitive to an agent’s beliefs, her evidence, or some combination of the two. I will assume that an agent’s beliefs come in the form of credences (i.e., degrees of belief that satisfy the probability axioms) and that her evidence gives rise to either evidential probabilities or evidential constraints on credences. Again for concision, I will use “credences” to mean “either subjective credences, or evidential probabilities, or some combination of the two (e.g., subjective credences constrained by evidence)”.
An objective norm need not be entirely insensitive to the agent’s belief state. For instance, the fact that I am uncertain about some important proposition might be an objective reason to seek out information concerning it.
One might worry that, by framing the debate in terms of rationality, I am talking past opponents of metanormativism like Weatherson, Harman, and Hedden, who are often understood to be interested in properties like moral rightness rather than rational requirement. But I am talking past these philosophers only if they are prepared to concede metanormativism as a thesis about rationality. And it seems clear that they are not, because they deny the need for any kind of metanorms. For instance, Harman writes: “Because Uncertaintism [\(\approx\) metanormativism] is false, the puzzle we discussed above, about how to compare moral value between conflicting moral views, is not important. It may be interesting as a puzzle; but nothing normatively important hangs on solving it” (Harman,2015, p. 58). And Hedden writes: “There is no normatively interesting sense of ought in which what you ought to do depends on your uncertainty about (fundamental) moral facts” (Hedden, 2016, p. 104).
We could assign orders to subjective norms more elegantly by simply saying that for any subjective norm N, the order of N is the least ordinal greater than every order of subjective normative belief to which N is sensitive. But then we would lose the distinction between norms that are sensitive only to the agent’s empirical beliefs and those that are sensitive to her objective normative beliefs, and would classify norms in a way that does not match the standard usage of “first-order” and “second-order” in the normative uncertainty literature. So I have adopted unnecessarily clunky definitions in order to interface better with the existing debate.
The internalist/externalist distinction is borrowed from Weatherson (2014, 2019), though I characterize it somewhat differently than he does. (For Weatherson’s characterization, see in particular §1.3 of Weatherson (2019).) Other philosophers have recognized the same distinction in various terms. For instance, Broome endorses the view I am calling externalism when he says that some norms impose “strict liability” (e.g., in Broome (2013, pp. 91ff)). Bykvist (2013) endorses the same thesis when he writes: “[M]y tentative conclusion is that in cases of uncertainty of rational matters there is an answer to the question of what it is rational to prefer which is not sensitive to your own views about rationality” (p. 133). Lin (2014) endorses a different form of externalism based on the idea of “adaptive rationality”. And I take Elga (2010) to endorse externalism in the epistemic domain when he says that certain epistemic norms “must be dogmatic with respect to their own correctness” (p. 185).
This premise is meant to allow that A is permitted to choose O on the basis of her nth-order normative beliefs alone, even under nth-order normative uncertainty, so long as all the nth-order norms in which she has positive credence authorize O. First, since norms are merely sets of propositions closed under logical consequence, any disjunction of norms is itself a norm (as long as it implies at least one permissibility or impermissility judgement). The disjunction \(N_1 \vee N_2 \vee \ldots \vee N_n\) yields a norm (generally non-comprehensive) whose choice set in situation S is the intersection of the choice sets of norms \(N_1\)–\(N_n\). Thus, if A is uncertain between various nth-order norms, but assigns positive credence to at least one true norm, and all the nth-order norms to which she assigns positive credence assert that O is permissible, then there is a true nth-order norm to which she assigns credence 1 that authorizes O (namely, the disjunction of all the nth-order norms in which she has positive credence). Because she assigns that norm credence 1, she presumably meets the belief conditions that place O its domain. But second, even if A is not certain of any nth-order norm that authorizes O, P2 asserts that she must resort to a higher-order norm in order to permissibly choose O only when she assigns positive credence to some nth-order norm that deauthorizes O.
This presentation of the regress problem was originally inspired by remarks in Weatherson (2014), though I now take Weatherson to be making a different argument, related to the “Argument from Fallibility” discussed in the next section. Sepielli seems to have something like the preceding argument in mind in this passage: “We can imagine someone who is...uncertain at all levels [of subjective normativity]. Indeed, one would suspect that this blanket uncertainty is typical. For who among us is certain about morality, let alone such esoterica as 8th-order, or 1,000th-order, normative uncertainty? But recall what animated our Divider [someone who recognizes both objective and subjective ‘oughts’] in the first place: that we cannot guide our behavior by norms about which we are uncertain. It would seem to follow from this that someone who is uncertain ‘all the way up’ will be unable to guide her behavior by norms at all” (Sepielli, 2018b, p. 792).
Thanks to Owen Cotton-Barratt for this suggestion
Under these assumptions, convergence is guaranteed only at transfinite levels of the metanormative hierarchy. To guarantee convergence at finite levels, further substantial assumptions are needed.
For further discussion of Sepielli’s view, see Riedener (2015, pp. 25–30).
Spelling out the regress problem as we have in this section helps us identify several escape routes for the internalist that are not obvious at first glance. In addition to the two we have considered—denying P3 on the basis of convergence results or denying P2 by allowing that agents may act in the face of merely dispositional uncertainty—there are at least four other possibilities: First, we could deny P2 by proposing some threshold less than certainty at which an agent may permissibly choose an option O based on her nth-order normative beliefs: e.g., a “Lockean threshold” for full belief or a requirement that the probability assigned to nth-order norms that deauthorize O be “de minimis” or “rationally negligible” (Smith, 2014). Second, we could deny P3 by holding that (i) agents are rationally required to assign probability 1 to all subjective normative truths and (ii) agents who violate this requirement of epistemic rationality will be unable to satisfy the requirements of practical rationality. (Claim (i) bears some resemble to the “Fixed Point Thesis” defended in Titelbaum (2015), although Titelbaum only claims that rationality prohibits false belief about the requirements of rationality, not that it prohibits any positive credence in false norms of rationality.) Third, we could deny P4 and hold that even unbounded agents cannot fully satisfy the demands of rationality, unless they are endowed with a degree of normative omniscience that lets them escape the uncertainty demanded by P3. Fourth, we could simply deny the assumption that there is a univocal notion of rational requirement (or subjective ought) that identifies appropriate responses to my belief state as a whole. This would probably require us to understand the debate between internalism and externalism very differently than I have in this paper. For instance, we might hold that when an agent is uncertain about norms of every order, the true nth-order norm nevertheless applies to her and determines what she nth-order-ought to do, but the true \((n+1)\)-order norm also applies to her and determines what she \((n + 1)\)-order ought to do, and these norms can give conflicting prescriptions that are not resolved by any all-things-considered norm. (Sepielli (2014b) might be understood as describing a view of this kind, though one could also identify his notion of “global systemic rationality” with rationality simpliciter and interpret him as denying P4.) I don’t find these responses particularly promising, but I won’t try to evaluate them here.
Second-order norms, as defined in §2.3, can be sensitive to an agent’s beliefs about objective norms, her beliefs about first-order subjective norms, or both. But MEC in particular is sensitive only to an agent’s beliefs about objective norms. As characterized in §2.2, objective norms assess options in terms of choiceworthiness (i.e., degree of objective reason) while subjective norms assess options in terms of rationality and subjective reasons. Thus, varying an agent’s beliefs about objective norms in isolation (holding all other features of a choice situation fixed) can change the expected choiceworthiness of her options, but varying her beliefs about subjective norms cannot.
One might claim that MEC is comprehensive only when combined with a rule for comparing the strength of reasons posited by rival objective norms (i.e., a solution to the “problem of intertheoretic value comparisons”). Without such a rule, it is typically indeterminate (except in cases of dominance) whether a given option maximizes expected choiceworthiness. The solution to this worry, I think, is to adopt what MacAskill (2014) calls the “universal scale” approach to intertheoretic comparisons. On this approach, objective norms are understood as different assignments to options of a shared set of cardinal choiceworthiness properties. When we are uncertain how to make intertheoretic comparisons between two objective norms (say, total and average utilitarianism), this should be understood as uncertainty between different “amplifications” (affine transformations) of one or both objective norms. If objective norms are characterized in this way, as rival assignments on a shared cardinal scale, then we do not need to resolve the problem of intertheoretic comparisons to apply MEC. For further developement of the universal scale approach to intertheoretic comparisons, see MacAskill et al. (2020, pp. 141–147). For ideas in a similar spirit, see the discussions of “absolutism” in Riedener (2015, Ch. 3) and of “de dicto utilities” in Carr (2020).
Perhaps the most familiar formulation of EP is: “If A believes she ought to \(\varphi\), then she is rationally required to intend to \(\varphi\)”. I immediately substitute what I think is an improved formulation, to avoid distracting complications. I have no strong view on the debate between narrow- and wide-scope formulations of principles of practical rationality, but adopt the wide-scope formulation simply because it’s weaker. I omit the usual reference to intentions in the consequent of EP for reasons described in Reisner (2013).
For an extended defence of MEC in the spirit of what I have called “the enkratic conception of rationality”, see Wedgwood (2017a). MEC is also defended at length in MacAskill and Ord (2020), though they don’t associate it with EP. And I take Broome to endorse MEC, or something very much like it, e.g. in Broome (1991) and (2013). Broome (2013) defends a version of EP he calls Enkrasia that, apart from some complications that aren’t relevant for our purposes, resembles a wide-scope version of the standard principle: Rationality requires that, if an agent believes she ought to \(\varphi\), then she intends to \(\varphi\) (p. 170). But the “ought” Broome has in mind is “prospective” rather than objective, i.e., depends on the prospects of the options in a given choice situation (Ch. 3). And Broome says that “the value of a prospect is an expected value of some sort” (p. 41). As far as I can see, this makes Broome’s Enkrasia a version of MEC.
For my own part, I am inclined to favor not MEC but a formulation of EP in terms of stochastic dominance, holding that O is rationally prohibited if and only if there is another option P such that (i) for any degree of choiceworthiness, P is at least as likely as O to be at least that choiceworthy and (ii) for some degree of choiceworthiness, P is strictly more likely to be at least that choiceworthy. But I favor this principle in part because I believe that, under normal epistemic circumstances, it is in surprisingly close agreement with MEC (while better handling some standard problem cases for expectational decision theory). The reasoning behind these claims is set out in Tarsney (2020), but is too involved to reproduce here. So for simplicity, I will focus in this paper on the more familiar MEC.
Arguments in this spirit are made by Broome (2013, p. 93), Bykvist (2013, p. 133), and Weatherson (2014, pp. 156–157). Weatherson puts the point as follows: “There is a worry that externalism is not sufficiently action guiding, and can’t be a norm that agents can live by. But any philosophical theory whatsoever is going to have to say something about how to judge agents who ascribe some credence to a rival theory. That’s true whether the theory is the first-order theory that Jeremy Bentham offers, or the second-order theory that Andrew Sepielli offers. Once you’re in the business of theorising at all, you’re going to impose an external standard on an agent, one that an agent may, in good faith and something like good conscience, sincerely reject...[T]he objector who searches for a thoroughly subjective standard is going to end up like Ponce de Leon”.
For a similar observation regarding beliefs about moral requirements, see the discussion of ProbWrong in Weatherson (2014, p. 146).
The same worry arises, to only a slightly lesser degree, if the threshold t is lower. At the extreme, suppose the internalist only asserts that an agent cannot be rationally permitted to choose an option that she is certain is rationally impermissible, or prohibited from choosing an option that she is certain is permissible. Then certainty in rational permissions or prohibitions is self-fulfilling and self-justifying. If I become certain that an option O is rationally permissible (or impermissible), no matter how irrationally, my belief is guaranteed to be correct, and I am justified in maintaining it in the face of any new evidence or arguments.
For defences of first-order externalism, see Weatherson (2014, 2019), Harman (2015), and Hedden (2016). For metanormativist replies, see Sepielli (2016, 2018a), Johnson-King (2018), and MacAskill and Ord (2020), among others. I give my own defense of metanormativism and reply to the first-order externalists in Tarsney (2017, Chs 2–3).
Is enkratic externalism vulnerable to an analogue of Podgorski’s argument, showing that it recommends options that are certainly irrational? No—at least, not any argument with the same force as Podgorski’s. First, a rational norm can imply that an agent is sometimes rationally required to choose an option that she’s correctly certain is less choiceworthy than an available alternative (as first-order externalist norms do, in Podgorski’s example). But no rational norm can imply that an agent is rationally required to choose an option that she’s correctly certain is irrational. (Or rather, any norm that implied this would be inconsistent, and MEC is clearly consistent.) So enkratic externalism cannot require an agent to act against her correct certainties either about choiceworthiness or about rationality. Second, enkratic externalism can plausibly explain why acting against your false certainties about rationality is unproblematic: The false certainty that O is irrational need not count against it at all from your perspective (since it is not in itself a belief about objective reasons, i.e., considerations that count for/against options); nor need it be accompanied by objective reason beliefs that count decisively against it from your perspective (since you may believe that you have no reason, or only pro tanto reason, to be rational). Third, constructing an argument analogous to Podgorski’s would require that, just as the agent in Podgorski’s example is uncertain which option is more choiceworthy according to a particular objective norm (because she is uncertain what outcome each option will have), an agent can be uncertain which of two options is rationally required according to a given subjective norm. But this may not be possible, if subjective norms are sensitive exclusively to an agent’s explicit beliefs or other aspects of her mental state to which she has perfect introspective access.
References
Broome, J. (1991). Desire, belief and expectation. Mind, 100(2), 265–267.
Broome, J. (2007). Is rationality normative? Disputatio, 2(23), 161–178.
Broome, J. (2013). Rationality through reasoning. Wiley-Blackwell.
Bykvist, K. (2013). Evaluative uncertainty, environmental ethics, and consequentialism. In A. Hiller, R. Ilea, & L. Kahn (Eds.), Consequentialism and environmental ethics. Routledge.
Carr, J. R. (2020). Normative uncertainty without theories. Australasian Journal of Philosophy, 98(4), 747–762.
Elga, A. (2010). How to disagree about how to disagree. In R. Feldman & T. A. Warfield (Eds.), Disagreement (pp. 175–186). Oxford University Press.
Gracely, E. J. (1996). On the noncomparability of judgments made by different ethical theories. Metaphilosophy, 27(3), 327–332.
Greaves, H., & Cotton-Barratt, O. (2024). A bargaining-theoretic approach to moral uncertainty. Journal of Moral Philosophy, 21, 127–169. https://doi.org/10.1163/17455243-20233810
Guerrero, A. A. (2007). Don’t know, don’t kill: Moral ignorance, culpability, and caution. Philosophical Studies, 136(1), 59–97.
Gustafsson, J. E., & Torpman, O. (2014). In defence of My Favourite Theory. Pacific Philosophical Quarterly, 95(2), 159–174.
Hájek, A. (2003). Waging war on Pascal’s wager. Philosophical Review, 112(1), 27–56.
Harman, E. (2015). The irrelevance of moral uncertainty. In R. Shafer-Landau (Ed.), Oxford studies in metaethics. (Vol. 10). Oxford University Press.
Hedden, B. (2016). Does MITE make right? On decision-making under normative uncertainty. In R. Shafer-Landau (Ed.), Oxford studies in metaethics. (Vol. 11). Oxford University Press.
Johnson-King, Z. A. (2018). Trying to act rightly. Ph.D. thesis, University of Michigan. https://zoe-johnsonking-blhb.squarespace.com/s/Trying-to-Act-Rightly-Doctoral-Dissertation-Zoe-Johnson-King-Rackham-formatting.pdf
Kiesewetter, B. (2017). The normativity of rationality. Oxford University Press.
Kolodny, N. (2005). Why be rational? Mind, 114(455), 509–563.
Lin, H. (2014). On the regress problem of deciding how to decide. Synthese, 191(4), 661–670.
Lockhart, T. (2000). Moral uncertainty and its consequences. Oxford University Press.
Lord, E. (2018). The importance of being rational. Oxford University Press.
MacAskill, W. (2014). Normative uncertainty. Ph.D. thesis, University of Oxford. https://philpapers.org/rec/MACNU
MacAskill, W. (2016). Normative uncertainty as a voting problem. Mind, 125(500), 967–1004.
MacAskill, W., Bykvist, K., & Ord, T. (2020). Moral uncertainty. Oxford University Press.
MacAskill, W., & Ord, T. (2020). Why maximize expected choice-worthiness? Noûs, 54(2), 327–353.
Nissan-Rozen, I. (2012). Doing the best one can: A new justification for the use of lotteries. Erasmus Journal for Philosophy and Economics, 5(1), 45–72.
Podgorski, A. (2020). Normative uncertainty and the dependence problem. Mind, 129(513), 43–70.
Reisner, A. (2013). Is the enkratic principle a requirement of rationality? Organon F, 20(4), 436–462.
Riedener, S. (2015). Maximising expected value under axiological uncertainty: An axiomatic approach. Ph.D. thesis, University of Oxford. https://uzh.academia.edu/StefanRiedener
Riedener, S. (2020). An axiomatic approach to axiological uncertainty. Philosophical Studies, 177(2), 483–504.
Robinson, P. (2022). Is normative uncertainty irrelevant if your descriptive uncertainty depends on it? Pacific Philosophical Quarterly, 103(4), 874–899.
Ross, J. (2006). Acceptance and practical reason. Ph.D. thesis, Rutgers University—New Brunswick. https://philpapers.org/rec/ROSAAP-2
Sepielli, A. (2009). What to do when you don’t know what to do. Oxford Studies in Metaethics, 4, 5–28.
Sepielli, A. (2010). ‘Along an imperfectly-lighted path’: Practical rationality and normative uncertainty. Ph.D. thesis, Rutgers University Graduate School—New Brunswick. https://philpapers.org/rec/SEPAAI
Sepielli, A. (2012). Subjective normativity and action guidance. In Timmons, M. (Ed.), Oxford studies in normative ethics (Vol. II). Oxford University Press.
Sepielli, A. (2014). Should you look before you leap? The Philosophers’ Magazine, 66, 89–93.
Sepielli, A. (2014). What to do when you don’t know what to do when you don’t know what to do... Noûs, 48(3), 521–544.
Sepielli, A. (2016). Moral uncertainty and fetishistic motivation. Philosophical Studies, 173(11), 2951–2968.
Sepielli, A. (2018). How moral uncertaintism can be both true and interesting. In M. Timmons (Ed.), Oxford studies in normative ethics (Vol. 7, pp. 98–116). Oxford University Press.
Sepielli, A. (2018). Subjective and objective reasons. In D. Star (Ed.), The Oxford handbook of reasons and normativity (pp. 784–799). Oxford University Press.
Smith, N. J. J. (2014). Is evaluative compositionality a requirement of rationality? Mind, 123(490), 457–502.
Tarsney, C. (2017). Rationality and moral risk: A moderate defense of hedging. Ph.D. thesis, University of Maryland. https://philarchive.org/archive/TARRAM-7
Tarsney, C. (2018). Moral uncertainty for deontologists. Ethical Theory and Moral Practice, 21(3), 505–520.
Tarsney, C. (2019). Normative uncertainty and social choice. Mind, 128(512), 1285–1308.
Tarsney, C. (2020). Exceeding expectations: Stochastic dominance as a general decision theory. Global Priorities Institute Working Paper Series.
Titelbaum, M. G. (2015). Rationality’s fixed point (or: In defense of right reason). In T. S. Gendler & J. Hawthorne (Eds.), Oxford studies in epistemology (Vol. 5, pp. 253–294). Oxford University Press.
Trammell, P. (2021). Fixed-point solutions to the regress problem in normative uncertainty. Synthese, 198(2), 1177–1199.
Weatherson, B. (2014). Running risks morally. Philosophical Studies, 167(1), 141–163.
Weatherson, B. (2019). Normative externalism. Oxford University Press.
Wedgwood, R. (2013). Akrasia and uncertainty. Organon F, 20(4), 484–506.
Wedgwood, R. (2017). Must rational intentions maximize utility? Philosophical Explorations, 20(sup2), 73–92.
Wedgwood, R. (2017). The value of rationality. Oxford University Press.
Worsnip, A. (2021). Fitting things together: Coherence and the demands of structural rationality. Oxford University Press.
Acknowledgements
For helpful feedback on earlier versions of this paper, I am grateful to Joseph Carlsmith, Owen Cotton-Barratt, Hilary Greaves, Daniel Kokotajlo, Dan Moller, Natasha Oughton, Eric Pacuit, Miriam Schoenfield, Teruji Thomas, Elliott Thornley, Philip Trammell, and audiences at the University of Richmond, the National University of Singapore, the Philosophy Days conference at Uppsala University, the “Hard Cases and Rational Choice” conference at the University of Bern, the REAPP Workshop on Moral and Rational Uncertainty at the University of Reading, and the “Ethics and Uncertainty” workshop at the Hebrew University of Jerusalem.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares that there is no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Tarsney, C. Metanormative regress: an escape plan. Philos Stud (2024). https://doi.org/10.1007/s11098-024-02114-w
Accepted:
Published:
DOI: https://doi.org/10.1007/s11098-024-02114-w