Non-rational action in the face of disagreement: an argument against (strong) non-conformism

A Correction to this article was published on 01 November 2018

This article has been updated


Recently there has been a surge of interest in the intersection between epistemology and action theory, especially in principles linking rationality in thought and rationality in action. Recently there has also been a surge of interest in the epistemic significance of perceived peer disagreement: what, epistemically speaking, is the rational response in light of disagreement with someone whom one regards as an epistemic peer? The objective of this paper is to explore these two issues—separately, but also in connection with one another. I turn first to the idea that the normative standing of our actions depends on the normative standing of our beliefs. I endorse this idea. More precisely, I endorse a principle according to which sufficiently high credence in success conditions for a given goal-directed action is a necessary condition on rational execution of that action. I then turn to the debate concerning the epistemic significance of perceived peer disagreement. The basic issue is whether such disagreement is always epistemically significant in the sense of serving as a defeater of the initial credences of the disagreeing parties. Conformists argue that this is so while non-conformists deny it. I present a new argument against a brand of non-conformism that I call “strong non-conformism”. The key premise is the principle that sufficiently high credence in success conditions for a given goal-directed action is a necessary condition on rational execution of that action. I argue that, given this principle, strong non-conformism fails to yield the verdict that the epistemic requirement on rational action is violated in a case where, intuitively, it is violated. This is because strong non-conformism has it that disagreement with a perceived peer does not act as a defeater in the relevant case. Conformism fares better.

This is a preview of subscription content, access via your institution.

Change history


  1. 1.

    Hawthorne and Stanley (2008) and Gerken (2011) likewise endorse the idea that there is an intimate connection between normativity in thought and action. However, although they converge on this idea, there are significant differences between them. Indeed, Gerken pitches his preferred principle—the “Warrant and Action Principle”—against the knowledge-based principles endorsed by Hawthorne and Stanley.

  2. 2.

    Should (i)–(iv) be qualified normatively, meaning that considerations regarding actions, evidence, urgency, and stakes are those that the subject can reasonably be expected to go through or entertain? In conversation Gerken has suggested that this kind of qualification might be apt if we are to avoid that, say, missing an obvious high stakes clue yields a deliberative context with low stakes when the CAG Principle is instantiated. The question concerning normativity raises a fundamental issue about how to understand the CAG Principle—namely, whether the requirement set on rational action by the principle is always tied to the subject’s de facto deliberative context. If deliberative contexts are normatively constrained, a subject’s de facto deliberative context need not be the one that is relevant to the CAG Principle. To see this suppose that deliberative contexts are normatively constrained and that Bob thinks that not much hangs on a certain goal-directed action that he performs. Now, suppose that he misses an obvious clue to the effect that a great deal hangs on the action. If so, Bob’s de facto deliberative context is not the one that determines the level of credence required for Bob’s action to be rational. Instead it is determined by a deliberative context in which the high stakes clue is taken into consideration and, thus, the level of credence required for rational action higher. I take Gerken to be inclined to take deliberative contexts to be normatively constrained. I’m inclined not to understand deliberative contexts in this way because I think that the constraint imposed by the CAG Principle should be based in the subject’s de facto deliberative context. I also think, however, that it is no trivial matter to settle which path is the better one to take. Thankfully, for the purposes of this paper, the normative qualifications can be left aside. Doing so does not impact the argument against strong non-conformism. The argument involves a specific case and the verdict that the CAG Principle delivers when combined with respectively strong non-conformism and conformism. Given the specifics of the case the CAG Principle will deliver the same verdict whether or not deliberative contexts are normatively construed.

  3. 3.

    In giving these remarks concerning invariantism, subject-sensitive invariantism, and contextualism I find myself in broad agreement with Gerken (2011).

  4. 4.

    Christensen (2007) and (2008), Feldman (2006) and (2007), Kelly (2005), Lackey (2010a) and (2010b). These authors follow Gutting (1982).

  5. 5.

    Incarnations of conformism with unequal assignments of relative weight involve greater deference to one of the two parties. Indeed, some incarnations are highly—but not completely—deferential to one party (e.g. assigning 0.95 to \(S_{1}\)’s initial credence and only 0.05 to that of \(S_{2})\). Some might find it odd that such assignments of relative weight fall under the rubric of conformism. A few comments. First, incarnations of (Conformism) are all conformist in the following sense: in the face of disagreement with a perceived peer, the disagreeing parties should adopt the same credence, and this credence is not identical to the initial credence of either party. Second, any brand of non-conformism involves the idea that there is some case in which at least one of the parties does not have to doxastically revise at all. This is a case with no weight being assigned at all to the view of some party. No incarnation of (Conformism) has this feature. Any incarnation of (Conformism) attributes non-zero weight to both parties. Third, in order to make finer-grained distinctions within the conformist camp we could describe assignments of weight as being more or less conformist relative to individual parties. E.g., the incarnation of (Conformism) that assigns 0.95 to \(S_{1}\)’s credence and 0.05 to \(S_{2}\)’s credence might be said to be only slightly conformist for \(S_{1}\) and highly conformist for \(S_{2}\). While both parties need to conform, \(S_{1}\) needs to revise doxastically to a much lower degree than \(S_{2}\).

  6. 6.

    Christensen (2007) and (2008) and Elga (2007). Feldman (2006, 2007) is likewise a conformist who thinks that each party should carry equal weight. However, Feldman endorses the view against the background of an all-or-nothing conception of doxastic attitudes rather than a graded one. Framed in this way the view says that \(S_{1}\) and \(S_{2}\) both ought to suspend judgment, i.e. neither believe p nor believe \(\sim \)p. For the purposes of this paper the graded version of the equal weight view works better, and so, I focus on this formulation of the view rather than the all-or-nothing version. As noted, equal weight is the most prominent version of (Conformism) in the literature. Other assignments of relative weight are possible, of course. However, see Elga (2007) for an influential argument to the effect that equal weight is the most compelling option.

  7. 7.

    Lackey (2010a, b), Rosen (2001) and Wedgwood (2007).

  8. 8.

    Kelly (2005) and Lackey (2010a, b). Lackey’s view satisfies both (Non-conformism\(^{\mathrm{SYM}})\) and (Non-conformism\(^{\mathrm{ASYM}})\).

  9. 9.

    This is somewhat similar to a credence counterpart of what Goldberg (2014, p. 168) calls a “doxastic defeater”.

  10. 10.

    I adopt some of this terminology from Kotzen (ms). Kotzen explores how one might transpose Pollock (1986)’s distinction between undercutting and rebutting defeaters from a binary framework to a credence framework, i.e. from a framework that works with beliefs as all-or-nothing states to one that operates with degrees of beliefs. This turns out not to be a trivial matter. See Kotzen (ms) for details. The distinction between undercutting and rebutting defeaters has been widely adopted in epistemology, as have various other defeater-related distinctions or taxonomies. For present purposes (Defeater) and (Defeater\(^{*})\) will do. What matters for the discussion to follow is whether perceived disagreement has an impact on what credence should be held by the parties involved. This issue can be discussed adequately on the basis of (Defeater) and (Defeater\(^{*})\).

  11. 11.

    As noted, given (Defeater) and (Defeater\(^{*})\) a defeater is any evidence that mandates doxastic revision, whether downwards or upwards. Here are two observations. First, given (Defeater) and (Defeater\(^{*})\) any evidence E for p that S acquires is a defeater for S’s prior credence in p. (This is because E is only evidence for p provided that \(\textit{cred}_{S}(p {\vert } E)>\) \(\textit{cred}_{S}(p)\).) Second, some people would only count cases of downwards revision as cases of defeat. My response to these two observations is as follows: it is correct that defeat goes in both directions within the present framework. It is also correct that some use ‘defeat’ only for downwards revision. (Defeater) and (Defeater\(^{*})\) nonetheless capture a proper notion of defeat. There may be others, including ones that are confined to cases of downwards revision. Here is the sense in which (Defeater) and (Defeater\(^{*})\) capture a proper notion of defeat: defeaters mandate downwards or upwards doxastic revision of a prior evidential credence. Since there is a rational mandate against sustaining the prior evidential credence, it loses its positive standing as a rational degree of belief. The prior evidential credence should be given up, and as such, there is a good sense in which its positive status as a rational degree of belief is defeated. For this reason the prior evidential credence can also be regarded as being epistemically downgraded by a defeater (although, again, this does not necessarily mean that downwards doxastic revision is called for—it may be that the defeater mandates upwards doxastic revision).

  12. 12.

    I understand disagreement as any case where two subjects have different credences in some target proposition p. Examples of disagreement often involve one subject’s having a high credence in p and another subject in \({\sim }{p}\). Since credences are subject to the axioms of probability (and so, \(prob(p) +prob( {\sim }{p}) = 1)\), this converts into an instance of disagreement as I have characterized it. One subject has a high credence in p, the other a low one.

  13. 13.

    Foley (2001, p. 108), Kelly (2005, pp. 179–180), Wedgwood (2007, pp. 260–261). The line of reasoning just presented violates what Christensen (2005, pp. 196–197) calls the “independence principle”. According to this principle a given party’s assessment of another party’s epistemic credentials must be independent of her own reasoning about the issue at hand. Those with conformist sympathies tend to endorse the principle while those with non-conformist sympathies tend to reject it.

  14. 14.

    In some respects (Disagreement) and its later extension, (Action), bear considerable resemblance to a case discussed by Christensen (2007, pp. 190–191).

  15. 15.

    I have just presented what I take to be the verdict that someone moved by the argument from epistemic egocentrism should be committed to. For the sake of generality, it is worth observing that other views also give a no-revision verdict for Dr. Fraenkel. At least they do so if (Disagreement) is slightly modified. Let me give two examples. If (Disagreement) is modified so as to give Dr. Fraenkel a symmetry breaker, the view presented in Lackey (2010a, b) will likewise sustain a no-revision verdict for Dr. Fraenkel. If, as part of (Disagreement), it is stipulated that Dr. Fraenkel has in fact reasoned correctly from E, Kelly (2005)’s Right Reasons View will likewise yield a no-revision verdict for Dr. Fraenkel.

  16. 16.

    As mentioned earlier, it may be that there are several necessary conditions on rational action and that a single action may fail several of them. Even so, it seems right that the epistemic features of (Disagreement) suffice to classify Dr. Fraenkel’s action as not being rational.

  17. 17.

    I claimed earlier that nothing hangs on my choice of 0.85 as the threshold for \(DC^\textit{ACTION}\)-adequacy. Let me substantiate this claim a bit. As seen, the credence that is rationally mandated by the equal weight view is 0.6. Thus, as long as the threshold for \(DC^\textit{ACTION}\)-adequacy is greater than 0.6, the equal weight view will deliver the right result—i.e. that Dr. Fraenkel’s action fails to be rational. Now, someone might suggest that the specifics of the case call for a higher threshold for \(DC^{\textit{ACTION}}\)-adequacy—say, some value greater than 0.9. In that case Dr. Fraenkel’s credence, specified to be 0.9, would be below the threshold for \(DC^{\textit{ACTION}}\)-adequacy. However, this does not undermine my argument. (Disagreement) and (Action) can just be modified so as to put Dr. Fraenkel’s credence above the greater-than-0.9 threshold. In that case, too, my argument will work for a wide range of threshold values. Given Dr. Fraenkel and Dr. Cohen’s initial credences the equal weight view will rationally mandate a credence below the greater-than-0.9 treshold.

  18. 18.

    Given (\(\textsc {ineq}^{*})\) it is also possible to put an upper bound on, say, Dr. Cohen’s credence. Reason as follows:

    $$\begin{aligned}&(\textsc {1}) \quad T(DC^{\textit{ACTION}})>w \cdot \textit{cred}_{F}(\textsc {right}{\vert } E) + (1 - w) \cdot \textit{cred}_{C}(\textsc {right}{\vert } E)\\&\Downarrow \\&(2) \quad T(DC^{\textit{ACTION}})- w \cdot \textit{cred}_{F} (\textsc {right}{\vert } E)> (1 - w) \cdot \textit{cred}_{C}(\textsc {right}{\vert } E)\\&\Downarrow \\&(3) \quad \frac{T(DC^{\textit{ACTION}})-w\cdot \textit{cred}_{F}(\textsc {right}\vert E)}{(1-w)} > \textit{cred}_{C}(\textsc {right}{\vert } E) \end{aligned}$$

    Given (3), knowledge of \(T(DC^{\textit{ACTION}})\), w, and \(\textit{cred}_{F}(\textsc {right}{\vert } E)\), we can calculate the maximum value that \(\textit{cred}_{C} (\textsc {right}{\vert } E)\) can take while still keeping the updated credence below \(T(DC^{\textit{ACTION}})\). Plugging in the values from (Above\(^{0.85-0.9}\)) we see that the upper bound for \(\textit{cred}_{C}(\textsc {right}{\vert } E)\) is 0.8:

    $$\begin{aligned} (3^{*}) \frac{0.85-0.5 \cdot 0.9}{(1-0.5)} > \textit{cred}_{C}(\textsc {right}{\vert } E) \Rightarrow (3^{**}) 0.8 > \textit{cred}_{C}(\textsc {right}{\vert } E) \end{aligned}$$

    What does this mean? It means the following: if we keep all other features of (Action) fixed and only vary Dr. Cohen’s initial credence, this credence has to be less than 0.8 in order for his post-disagreement credence to be below the threshold (0.85) for rational action.

  19. 19.

    Thanks to an anonymous reviewer for raising this issue.

  20. 20.

    See, e.g., Christensen (2007, pp. 190-191).


  1. Christensen, D. (2007). Epistemology of disagreement: The good news. Philosophical Review, 116, 187–217.

  2. Christensen, D. (2008). Disagreement as evidence: The epistemology of controversy. Philosophy Compass, 4(5), 756–767.

    Article  Google Scholar 

  3. Elga, A. (2007). Reflection and disagreement. Noûs, 41, 478–502.

    Article  Google Scholar 

  4. Feldman, R. (2006). Epistemological puzzles about disagreement. In S. Hetherington (Ed.), Epistemology futures (pp. 216–236). Oxford: Clarendon Press.

    Google Scholar 

  5. Feldman, R. (2007). Reasonable religious disagreements. In L. Anthony (Ed.), Philosophers without gods (pp. 194–214). Oxford: Oxford University Press.

    Google Scholar 

  6. Foley, R. (2001). Intellectual trust in oneself and others. New York: Cambridge University Press.

    Book  Google Scholar 

  7. Gerken, M. (2011). Warrant and action. Synthese, 178, 529–547.

    Article  Google Scholar 

  8. Goldberg, S. (2014). Disagreement, defeat, and assertion. In D. Christensen & J. Lackey (Eds.), The epistemology of disagreement (pp. 167–189). Oxford: Oxford University Press.

    Google Scholar 

  9. Gutting, G. (1982). Belief and religious skepticism. Notre Dame: University of Notre Dame Press.

    Google Scholar 

  10. Hawthorne, J., & Stanley, J. (2008). Knowledge and action. Journal of Philosophy, 105, 571–690.

    Article  Google Scholar 

  11. Kelly, T. (2005). The epistemic significance of disagreement. In J. Hawthorne & T. Gendler (Eds.), Oxford studies in epistemology (Vol. 1, pp. 167–196). Oxford: Oxford University Press.

    Google Scholar 

  12. Kotzen, M. (Ms). A formal account of defeat.

  13. Lackey, J. (2010a). A justificationist view of disagreement’s epistemic significance. In A. Haddock, A. Millar, & D. Prithcard (Eds.), Social epistemology (pp. 298–325). Oxford: Oxford University Press.

    Chapter  Google Scholar 

  14. Lackey, J. (2010b). What should we do when we disagree? In J. Hawthorne & T. Gendler (Eds.), Oxford studies in epistemology (Vol. 3, pp. 274–293). Oxford: Oxford University Press.

    Google Scholar 

  15. Pollock, J. (1986). Contemporary theories of knowledge. Towota, NJ: Rowman and Littlefield.

    Google Scholar 

  16. Rosen, G. (2001). Nominalism, naturalism, epistemic relativism. Philosophical Perspectives, 15, 60–91.

    Google Scholar 

  17. Wedgwood, R. (2007). The nature of normativity. Oxford: Oxford University Press.

    Book  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Nikolaj Jang Lee Linding Pedersen.

Additional information

I am grateful to the following people for helpful discussion: Kristoffer Ahlström-Vij, Jens Christian Bjerring, Sung Yong Bong, Evan Butts, Colin Caret, Adam Carter, Aiste Celkyte, David Christensen, Richard Dietz, Catherine Elgin, David Estlund, Axel Gelfert, Mikkel Gerken, Allan Hazlett, Todd Jones, Jesper Kallestrup, Jinho Kang, Klemens Kappel, Jiwon Kim, Kihyeon Kim, Joohan Lee, Sukjae Lee, Michael Lynch, Chien-kuo Mi, Matthew Mullins, Emil Møller, Orestis Palermos, Shane Ryan, Jisoo Seo, Nico Sillins, Neil Sinhababu, Diana Stewart, Lani Watson, Han van Wietmarschen, James Woodbridge, and Jeremy Wyatt. I also thank two anonymous reviewers for helpful comments.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Pedersen, N.J.L. Non-rational action in the face of disagreement: an argument against (strong) non-conformism. Synthese 195, 2935–2966 (2018).

Download citation


  • Credence
  • Action
  • Rationality
  • Peer disagreement
  • Conformism
  • Non-conformism