Skip to main content
Epistemic Communities in American Public Law


Epistemic Communities in American Public Law




What do judges know, or think they know? What do judges not know, and not know that they do not know? When and why do judges sort themselves into competing “tribes”? The answer is that like everyone else, judges are part of epistemic communities. Consider some illustrations. In the last two decades, there has been an extraordinary outpouring of careful historical work on two of the most fundamental questions in constitutional law: (1) whether Congress may delegate open-ended discretionary power to the executive branch (or others) and (2) whether Congress may restrict the president’s power to remove high-level officials in the executive branch. The best reading of the new evidence is that there was no robust nondelegation doctrine during the founding period, if there was a nondelegation doctrine at all. Though the issue is closer, the best reading of the new evidence is that during the founding period, the Constitution was understood to authorize Congress to restrict the president’s power of removal, even over principal officers (with important qualifications). What is remarkable is that in both contexts, no originalist on the Court has been convinced by the relevant evidence, or even seriously grappled with it. There are three plausible explanations for the apparent impotence of historical evidence in this context (and others). The first points to optimal search, and hence to simple lack of awareness of the relevant evidence. The second is Bayesian and spotlights rational updating. The third points to motivated reasoning. All three accounts offer lessons for lawyers and others seeking to marshal historical or other evidence to disrupt engrained judicial beliefs. 


Keywords: constitutional law

How to Cite:

Sunstein, C. R., (2024) “Epistemic Communities in American Public Law”, Political Philosophy 1(1). doi:

Epistemic Communities in American Public Law

Cass R. Sunstein

Law, Harvard University, US


My focus here is on what judges know, or think they know, and on what judges do not know, and do not know that they do not know.1 I suggest that judges live in, and help constitute, epistemic communities, broadly defined as “professional networks with authoritative and policy-relevant expertise.”2 One result is epistemic overconfidence, which is a serious problem in American public law. The existence of epistemic communities, and the problem of epistemic overconfidence, create serious challenges for the project of originalism, but they also create serious challenges for public law more broadly. And while my focus is on American public law, the account of epistemic communities, and of why they exist, is meant to have general implications for politics and processes more broadly, especially insofar as I emphasize optimal search, rational updating in light of priors, and motivated reasoning.

It is difficult to defend these propositions in the abstract, so let us focus on some concrete issues. We are going to linger over those issues for a good while, but it is important to keep the broader topics of epistemic communities and epistemic overconfidence in mind; I will return to them in due course. I use the concrete problems, and American public law, not only because they are interesting and important in their own right, but also for purposes of illustration. Sometimes we can find a universe in a grain of sand.

In the United States, there has been an extraordinary, even unprecedented outpouring of historical research on the views of the founding generation and the original public meaning of the U.S. Constitution, especially in the domain of separation of powers.3 We may well know more about those views, and about that meaning, than we ever did.4 It is now clear that from the originalist standpoint,5 the idea of a “nondelegation doctrine” is a myth6 – a concoction of the twentieth century.7 The founding generation simply did not believe that the Constitution limited Congress’s power to grant broad discretion to others, including the executive branch.8 We know too that from the originalist standpoint,9 the idea of unrestricted removal power is a myth – another concoction of the twentieth century, not the eighteenth.10 Among the founding generation, it was widely understood that Congress could limit the president’s removal power over high-level employees, including (some) principal officers.11 Understood in terms of its original public meaning, the founding document did not forbid grants of broad discretion, and it did not require unrestricted removal power.

Here is the puzzle. The clear historical understandings appear to have had no impact – none at all – on originalist judges who have been committed to the nondelegation doctrine12 and unrestricted removal power.13 Their commitments seem impervious to the current evidence.14 It is true that some of the historical research is new;15 we cannot rule out the possibility that its effects will grow over time. But some of the research is not so new, and even so, its effects have been essentially nonexistent.16

The appropriate conclusion is that when preexisting convictions are firm, careful historical evidence does not matter, even or perhaps especially in constitutional law.17 If that is true, it is a serious challenge for all those concerned about constitutionalism, and perhaps especially for originalists. One way to understand the situation is that constitutional law has become, or often is, essentially tribal, even on questions of historical fact, and even when the ideological or political valence of the issue is far from clear.18 Some people are part of Team Nondelegation Doctrine, and some people are part of Team Myth of the Nondelegation Doctrine. Some people are part of Team Unlimited Presidential Removal Power, and some people are part of Team Limited Presidential Removal Power. It is exceedingly difficult to get people to switch teams.19


Let’s start over. Is there another way to think about the recent research? Would a judge or scholar from a different epistemic community offer a different understanding of what we have learned? Might we be living, in some sense, in parallel epistemic worlds?

In recent decades, we have seen an extraordinary outpouring of historical research on the views of the founding generation. This is especially true in the domain of separation of powers. We now know that from the originalist standpoint, the idea of a “nondelegation doctrine” is exceedingly well-founded.20 The founding generation clearly believed that the Constitution limits Congress’ power to grant broad discretion to others, including the executive branch.21 We know too that from the originalist standpoint, the idea of a “unitary executive” has firm and deep roots.22 In the founding generation, it was widely agreed that Congress could not limit the president’s removal power of high-level employees, including Cabinet heads.23 These historical facts speak volumes about the original public meaning of the founding document.

Here is the puzzle. The clear historical understandings appear to have had no impact – none – on judges who have been unenthusiastic about the nondelegation doctrine24 and about the proposition that Article II forbids congressional restrictions on the president’s removal power.25 Their commitments are impervious to evidence.26 It is true that some of the historical research is new,27 and we cannot rule out the possibility that its effects will grow over time. But some of the research is not so new, and its effects appear to have been nonexistent.28

The appropriate conclusion is that when preexisting convictions are firm, careful historical evidence does not matter, even or perhaps especially in American constitutional law. One way to understand the situation is that in the United States, constitutional law has become, or often is, essentially tribal, even on questions of historical fact, and even when the ideological or political valence of the issue is far from clear. Some people are part of Team Nondelegation Doctrine, and some people are part of Team Myth of the Nondelegation Doctrine. Some people are part of Team Unlimited Presidential Removal Power, and some people are part of Team Limited Presidential Removal Power. It is exceedingly difficult to get people to switch teams.


Let’s start over (one last time).29 (Cards on the table; this is what I actually think.)

In recent years, we have seen an extraordinary amount of historical research on the views of the founding generation and the original public meaning of the Constitution. This is especially true in the domain of separation of powers. We now know that from the originalist standpoint, the idea of a “nondelegation doctrine” has weak foundations – not no foundations, but weak foundations.30 At the constitutional convention and during the ratification debates, the doctrine was not mentioned.31 Early congresses granted broad discretion to others, almost always without constitutional objections.32 True, the word “almost” is important; it would be an overstatement to say that the nondelegation doctrine was entirely absent.33 We cannot rule out the possibility that the founding generation would have found a constitutional defect if Congress had said that (for example) the president may do “whatever he deems necessary or appropriate to promote the general welfare.”

But no evidence demonstrates that such a law would have been deemed unconstitutional, and the early practice speaks decisively against the idea that Congress cannot delegate broad policymaking authority to others. These points bear directly on the original public meaning. The central conclusion is clear. Understood in terms of that meaning, “the Constitution at the Founding contained no discernable, legalized prohibition on delegations of legislative power, at least so long as the exercise of that power remained subject to congressional oversight and control.”34

The question of the removal power is altogether different. With respect to the nondelegation doctrine, the problem is that the evidence is so sparse; there are plenty of dogs who did not bark, day or night.35 The sparseness of the evidence, in the face of broad grants of discretion, is powerful evidence against a robust nondelegation doctrine.36 The basic point is that members of the founding generation were not worried or even concerned about such grants of discretion. That point tells us a great deal about original public meaning of the relevant provisions of the founding document, above all Article I, section 1.

By contrast, the evidence with respect to Article II, section 1 is anything but sparse; there is voluminous material on the president’s removal power.37 Alexander Hamilton, for one, changed his mind. In The Federalist No. 77, he said that the removal power followed the appointment power and so was shared:38

It has been mentioned as one of the advantages to be expected from the co-operation of the Senate, in the business of appointments, that it would contribute to the stability of the administration. The consent of that body would be necessary to displace as well as to appoint. A change of the Chief Magistrate, therefore, would not occasion so violent or so general a revolution in the officers of the government as might be expected, if he were the sole disposer of offices.

What is remarkable about this statement is not only the clear conclusion; it is also the effort to ground the conclusion in a theory, which has to do with stability over time. Madison spoke in similar terms in The Federalist No. 39:39

The tenure by which the judges are to hold their places, is, as it unquestionably ought to be, that of good behavior. The tenure of the ministerial offices generally will be a subject of legal regulation, conformably to the reason of the case, and the example of the State Constitutions.

After the founding, Hamilton concluded that the removal power resided in the president alone. In 1789, he said, “[t]he constitution affirms, that the executive power shall be vested in the president” and that “inasmuch as the power of removal is of an executive nature … it is beyond the reach of the legislative body.”40 For his part, James Madison was all over the place.41 Early on, he said, “it is in the discretion of the legislature to say upon what terms the office shall be held, either during good behaviour, or during pleasure.”42 But he shifted to the view that the president had unrestricted removal power43 (and then shifted back again44). Insofar as he (sometimes) supported unlimited presidential authority, Madison was not alone. Several others shared his view.45 Still, Madison’s sometimes view in favor of unlimited removal authority did not command anything like a consensus,46 and the best account is that most people in Congress disagreed with him.47 For that reason, it is challenging to support the view that understood in terms of its original public meaning, any provision of the Constitution forbids Congress from restricting the president’s removal authority, at least across-the-board.48

The conclusion is simple: “Once a series of misreadings and omissions are corrected, it is unclear what evidence remains for the Taft/Roberts interpretation of the Decision of 1789.”49 Indeed, “the real Decision of 1789 was a rejection of the unitary model,” for “a significant majority opposed the interpretation that Article II implied a general removal power, and even among the minority claimed by the unitary theorists, there was explicit opposition to the claim of legislative indefeasibility.”50

What should we do with that? Does the claim of unlimited removal power stand defeated? Perhaps not. Perhaps we can say that even if most people in Congress rejected that idea, it nonetheless remains on the table: With respect to the original public meaning, Congress’ early interpretation is hardly authoritative, especially on this kind of question, because Congress has an institutional interest in aggrandizing its own power. It follows that on the question, whether Congress can restrict the president’s removal authority over high-level executive branch employees, we have to offer the Scottish verdict: Not proven. On that question, there was no clear original public meaning.51

The historical evidence, then, supports the following conclusions.52 (1) The founding generation did not believe in anything like a robust nondelegation doctrine, understood as prohibition of broad grants of discretion to the executive branch.53 This evidence bears strongly on the original public meaning of the founding document. (2) It is unclear whether some kind of nondelegation doctrine, banning not merely broad but entirely open-ended grants of discretionary authority, would have been believed to be unconstitutional; we do not have reliable evidence on that question. Nor do we have reliable evidence on the original public meaning of the founding document on that question. (3) Reasonable people can differ over whether Congress can restrict the president’s removal authority, but the weight of the evidence is that the founding generation believed that it can.54 That evidence is also strongly suggestive about the original public meaning of Article II. On the removal question, the text of the Constitution is opaque, and the historical materials make it challenging to defend the view that interpreted in accordance with its original public meaning, the Constitution requires all high-level executive officers to be at-will employees of the president.

Here is the puzzle. The historical research appears to have had no impact – none – on judges, in the sense that judicial opinions have not seriously grappled with the relevant evidence.55 On these questions, judicial commitments seem impervious to evidence. Much of the time, judges appear to be treating contrary historical evidence as a college debater would – as welcome supporting material for a position that has long been fixed in advance, or as a problem to be overcome through strategic silence, or through maneuvers and tricks.

One way to understand the situation is that constitutional law has become, or often is, essentially tribal, even on questions of historical fact, and even when the ideological or political valence of the issue is far from clear. Some people are part of Team Nondelegation Doctrine, and some people are part of Team Myth of the Nondelegation Doctrine. Some people are part of Team Untouchable Removal Power, and some people are part of Team Touchable Removal Power. It is exceedingly difficult to get people to switch teams. The reason, in short, is that team members are part of epistemic communities, and a member of one such community is unlikely to be convinced by a member of another such community.56 But this point raises puzzles of its own.


Of the three accounts just given, some version of the third is the most plausible.57 The original public meaning of Article I, section 1 does not clearly include a nondelegation doctrine, taken as a prohibition on broad grants of discretion, by Congress, to the executive branch (and others).58 Broad grants of discretion were widely deemed to be constitutional.59 To be sure, it would be too categorical and definitive to say that the founding generation did not believe in any kind of nondelegation doctrine, again understood as a restriction on open-ended grants of discretionary authority to the executive branch (or others). The evidence does not demonstrate that the founding generation believed in any such doctrine, but it does not show that it did not.

The view that the Constitution forbids restrictions on the president’s power to remove high-level executive branch officials was embraced explicitly and by prominent people in the founding era.60 Was that the original public meaning of Article II? We do not know for sure, but the best evidence suggests that it was not. To be sure, that evidence is not conclusive. But as of now, contextual materials suggest that congressional restrictions on the president’s removal power did not offend the original public meaning of Article II.

On both of these issues, the recent historical research is deeply informative, and in important respects jarring. The sheer number of broad grants of discretionary authority in the 1790s seems to speak volumes. No one could read the historical research and continue to have confidence in the existence of a robust nondelegation doctrine. No one could read the historical materials, and the relevant statutes, and remain certain that the founding generation believed that Congress lacked authority to restrict the president’s removal authority. And yet we do not appear to see any movement in the views of prominent judges on the relevant questions. We do not see any kind of updating, in the direction of either a shift in view or a reduced level of confidence.

To be sure, the research raises puzzles, and conscientious originalists will want to focus on them.61 Suppose we learn a great deal about actual practices during the founding era, or about the views of members of Congress about the permissibility of broad grants of discretion or restrictions on presidential removal power. Have we learned the original public meaning of relevant terms? Not necessarily. Many originalists now distinguish between the original expectations, or contemporaneous “intentions,” and the original public meaning.62 What is the original public meaning of the relevant constitutional terms? To answer that question, congressional practices and views expressed during legislative debates would not be authoritative.

Nonetheless, it would be most surprising, in view of the relevant research, if the original public meaning of the relevant terms supported a nondelegation doctrine that prohibited Congress from granting broad discretion to agencies, or that prohibited Congress from imposing (some) restrictions on presidential removal power. The reason is that abstract terms (“legislative,” “executive,” “take care”) are most unlikely to have had concrete specifications that would run roughshod over widespread contemporaneous practices, especially when those who engaged in those practices were keenly aware of those abstract terms and focused directly on constitutional meaning.


It is possible, of course, to invoke stare decisis. When, if ever, should historical research be sufficient to dislodge longstanding constitutional interpretations?63 That is a large and disputed question. One might urge that the nondelegation doctrine has been with us for a very long time,64 and that new historical work, inevitably disputed, is hardly sufficient to dispense with longstanding doctrine. Constitutional Burkeans might embrace that view; they do not favor French revolutions, even if those revolutions are based on compelling historical research. One might urge that the idea of unrestricted removal authority has also been longstanding,65 even if some qualifications are necessary,66 which means that recent historical work must grapple with respect for precedent. In that event, the relevant work must be placed in the context of the longstanding debate about how an originalist, or for that matter any judge who cares about the original public meaning, should deal with historical work that is inconsistent with apparently settled judicial understandings. Here again, Burkeans and originalists might be at odds.67

That is certainly a possibility. As applied to the two questions at hand, the problem with this claim is that the relevant issues are not so settled. The modern effort to reinvigorate the nondelegation doctrine68 is hardly Burkean or preservative. It is designed to alter longstanding practices, not to preserve them, and to do so in the interest of the original public meaning. So too, the effort to expand presidential removal authority has shown skepticism about longstanding precedent.69 Explicitly invoking the original understanding and the supposed Decision of 1789, the Court has asserted that the constitutional settlement severely restricts congressional power to limit the president’s power of removal and that some of its decisions, recognizing such power, must be rethought and cabined.70

Return in this light to the recent and not-so-recent historical findings. What is striking is that we do not see anyone saying: “I used to believe X, but having reviewed the historical evidence, I now believe Y.”71 We do not see anyone saying, “I used to believe X, but I now do so with less confidence; I continue to believe X, but I have read the evidence, and I now see that reasonable people can differ, and I am now aware that I might be wrong.” We do not even see anyone saying, “I am aware that my view is inconsistent with significant historical research, which reaches a very different conclusion. Nonetheless, I believe that my view is right.” At the very least, it seems fair to say that in these two domains, the failure to engage with the historical research72 is more than a bit shocking. At first glance, it seems staggering.

How can we explain that failure?


One answer might seem disappointingly mundane: Perhaps the relevant judges are simply unaware of the relevant research. Judges, including Supreme Court justices, are not Large Language Models. They have limited time and information, and they are busy. They have to think about optimal search.73 They do not have time to read the latest academic papers or the latest books from university presses. In their spare moments, they might be inclined to do some of that, but they are naturally focused on the cases that come before them, and on the materials that are cited in the briefs. It might simply be the case that the relevant historical research is not known to them, not because they are not doing their jobs, but because doing their jobs is what they are doing.

To be sure, the historical research on which I have focused here is not exactly secret, and some of the central work on nondelegation and removal is not exactly new.74 But judges are judges, and it is not their practice to read the latest academic writing, whose credibility might not be self-evident (not to put too fine a point on it). Is some jarring new paper in the Columbia Law Review or the University of Pennsylvania Law Review really worth their attention?75

We might sharpen this point. Like all of us, judges are part of epistemic communities.76 They know what they know, and they like what they like, and what they know and like is in large part a product of whom they trust.77 Something similar can be said about when they are willing to learn, and from whom. Suppose that you are in a community of people who think that the Warren and Burger Courts were, in many respects, an enduring model for judicial behavior, that the modern administrative state is legally secure, that Roe v. Wade78 was clearly right, and that originalism is foolish and hopeless. If that is what you were taught, and if that is what you believe, historical work questioning the decisions of the Warren and Burger Courts, and suggesting that Roe v. Wade is indefensible on originalist grounds, might not interest you at all, and you might not bother to try to understand it.

Even academic work that appeals to your own theoretical premises, but that suggests that your own deepest commitments are wrong, might not end up high on your reading list. Something similar might be true of originalists who are told that the nondelegation doctrine rests on weak historical foundations or that the president lacks unrestricted removal authority. If so, we are not speaking of a simple absence of information, or of materials that happen not to have happened on the judicial viewscreen. We are speaking of an absence of information that is founded on a judgment, rational whether or not correct, about what is worth reading.

Of course these points are generalizable. They apply not only to American judges, but also to politicians in all nations, and to citizens engaged or not-so-engaged with politics. They even apply to scientists. They help to explain the existence of epistemic communities of various sorts, and help to identify some of the foundations of polarization.


Let us indulge, for a moment, the assumption that some of that research is, in fact, known to relevant judges,79 and that it nonetheless has little or no effect. Why?

Any explanation would be speculative, but two accounts, drawn from broader research on when people do and do not change their minds, seem plausible.80 Both accounts are related to the idea of epistemic communities, and they are connected with the idea of optimal search. The first account is Bayesian; it involves rational updating with respect to relevant probabilities.81 The Bayesian account helps explain why an epistemic community might be difficult to move. It also helps explain why disparate epistemic communities might arise in constitutional law. Consider this phrase, often heard among judges and lawyers: “That has to be right.” What does that phrase mean? It often means that in light of one’s prior convictions, “that,” as opposed to “not-that,” has an exceedingly high probability of being true or correct.

The second involves motivated reasoning; people (including judges) believe what they want to believe, and it is difficult to convince them to change their minds if they do not want to change their minds. Here again, we can understand the apparent intransigence of epistemic communities. In sketching these points, I make no claim to originality; my hope is that identifying and separating them might be helpful in understanding why and when historical evidence will have little or no impact in law. Both accounts help explain the tribalization of constitutional law, and the bafflingly clear relationship between a given view on one or both of the issues here and one’s political or ideological orientation, broadly speaking.

Begin with rational updating. With respect to delegation:

Suppose that it seems provisionally clear that the founding document simply must contain some kind of nondelegation doctrine. Suppose that this is what you were taught by those whom you most trust. Suppose that this belief is a kind of bedrock. Suppose that the very separation of legislation from execution, made plain in Article I and Article II, produces an exceedingly strong inference: Congress cannot grant open-ended authority to the executive. The reason is that if it did so, legislative and executive power would no longer be separated, and the separation of powers would collapse.82 If one believes that, then any historical work — purporting to demonstrate that in terms of the original meaning, a grant of open-ended authority would be unobjectionable — would be presumptively unconvincing. Such work would seem daft (or motivated). It would be a little like saying that dropped objects do not fall or that the Holocaust did not happen. To say the least, it would have to meet a singularly heavy burden of justification. If one reads it, it is with a singularly skeptical eye.

Relative or even incomplete silence in the convention or ratification debates could not possibly meet the relevant burden. One might update one’s belief a bit, but without a significant shift in one’s judgment that a nondelegation doctrine was part of the original public meaning of Article I. The prohibition on delegation of open-ended authority simply went without saying; why would it be necessary to say or discuss something that it is self-evident? So too, early acts of Congress, seeming to give broad authority to the executive, do not prove much at all. In context, the grants of authority might not have been nearly as open-ended as they might now seem.83 Or they might have involved a grant of authority to the president in domains in which the Constitution recognized a kind of presidential primacy.84 In any case, some or many of the early acts might have been unconstitutional. Stranger things have happened. And recall that nondelegation doctrine was not exactly absent from founding era debates.85 James Madison believed in the doctrine, and his views may deserve pride of place.86

With respect to removal:

If anything is clear, this is: The Constitution vests executive power in a president of the United States,87 and it requires him to take care that the laws be faithfully executed. The executive power is not vested in the Secretary of State, the Secretary of the Treasury, the Attorney General, or the Consumer Product Safety Commission. In that light, it would be singularly odd to say that high-level officials can stay on the job even though the president wants them to leave. If one begins with those convictions, any historical evidence, purporting to show that Congress can immunize such officials from presidential removal, would seem implausible, even bizarre.88 At the very least, it would have to meet a heavy burden of justification.

Everyone knows that the Decision of 1789 was real,89 and it was straightforward enough: The president was given unlimited removal power.90 In that light, new evidence, showing that members of Congress had different and competing views and that in the early decades, some officers might have been given some protection from removal,91 is not exactly shocking, and cannot possibly shake the basic conclusion. Recall that Congress itself has an institutional interest in aggrandizing its own power at the expense of the president. To be sure, new historical evidence might produce a little updating in one’s subjective probability judgment with respect to presidential authority, but only a little; one would still believe, with very high probability, that the removal power cannot be limited.

To the extent that rational updating accounts for judicial rejection of historical evidence, those who provide such evidence might be inclined to despair. What can one do with an epistemic community that is essentially closed?92 Those inclined to despair might ask: Might this be a kind of cult? But let us step back from the most extreme examples. Just as rational people are unlikely to be willing to believe apparent evidence that dropped objects do not fall, or that the Holocaust did not happen, so might rational people be skeptical about evidence that purports to establish a proposition about constitutional law that they essentially know, or think they know, to be false. Still, despair would be excessive. For those who want to establish a proposition that runs up against entrenched judicial convictions about some proposition of fact or law, including the original public meaning, the task is simple: produce overwhelming evidence, sufficient to dislodge those convictions. Here too, stranger things have happened.93


Why do people change their beliefs? Why are they sometimes intransigent? Thus far my emphasis has been on rational updating; if one firmly believes that X is true, it is rational to demand a great deal of evidence before concluding that X is false. But a great deal of work explores the idea of “motivated reasoning.”94 The basic idea is that people believe what they want to believe. If we focus on epistemic communities in constitutional law, the line between rational updating and motivated reasoning might start to become thin. A community that is intensely committed to a particular belief might view a supposed counterargument not only with skepticism but also with venom and scorn. Still, the distinction is both important and real.95

For illustration (admittedly far afield from our current topics96), here is a dialogue, from the actual writing of The Return of the Jedi, involving George Lucas, the mastermind behind Star Wars, and his great collaborator Lawrence Kasden:

Kasdan: I think you should kill Luke and have Leia take over.

Lucas: You don’t want to kill Luke.

Kasdan: Okay, then kill Yoda.

Lucas: I don’t want to kill Yoda. You don’t have to kill people. You’re a product of the 1980s. You don’t go around killing people. It’s not nice.

Kasdan: No, I’m not. I’m trying to give the story some kind of an edge to it… .

Lucas: By killing somebody, I think you alienate the audience.

Kasdan: I’m saying that the movie has more emotional weight if someone you love is lost along the way; the journey has more impact.

Lucas: I don’t like that and I don’t believe that.97

Note Lucas’ closing comment: “I don’t like that and I don’t believe that.” Not liking precedes, and perhaps accounts for, not believing. The basic idea is that if people do not like believing something, they are less likely to believe it. According to a standard account of motivated reasoning, “people do not reason like impartial judges, but instead recruit evidence like attorneys, looking for evidence that supports a desired belief while trying to steer clear of evidence that refutes it.”98 What is noteworthy here is that this account, not involving law, makes an explicit distinction between “impartial judges” and “attorneys,” and suggest that people act more like the latter than the former, even when they are confronted with new evidence.99 Actual judges might sometimes be like that as well. (The same is true for participants in political processes more generally; consider debates over climate change and immigration).

A distinctive form of partiality, also with evident legal resonance, involves allocation of the burden of proof. If people are considering a claim that they would greatly like not to believe, “people tend to ask themselves something like ‘Must I believe this?’”100 It is worth pausing over that suggestion. If one asks oneself whether one “must” believe something, the usual answer is a firm no, because “some contradictory evidence can be found for almost any proposition.”101 For the two issues under discussion here, there is certainly some evidence on both sides.

A common source of motivated reasoning, of self-evident relevance to constitutional law and politics, is confirmation bias: People tend to believe things that are consistent with their antecedent beliefs.102 Confirmation bias can involve rational updating,103 in which case it need not be seen as a bias at all. I will not credit information supposedly suggesting that Richard Thaler and Bob Dylan did not deserve the Nobel Prize, because I believe that I have a great deal of information suggesting that both of them deserved the Nobel Prize, which means that I will rationally dismiss apparently disconfirming information. But when confirmation bias is motivated, it is because it is unpleasant to learn that one was wrong, and people do not want to believe that they were wrong.104 We can easily imagine that if judges begin with a firm conviction that nondelegation doctrine has solid historical roots, and then receive information suggesting that the doctrine lacks such roots, confirmation bias will immediately affect their reception of the apparent evidence – and make them inclined to ignore or dismiss it.

A more recent and highly relevant finding involves “desirability bias.”105 The basic idea is that people tend to believe what they find it desirable to believe. It is tempting to identify desirability bias with confirmation bias, but they are different. You might believe that you are mortal but not find that desirable; confirmation bias suggests that you will be inclined not to believe a claim that immortality is around the corner, whereas desirability bias suggests that you will be inclined to believe that same claim. We can easily imagine information (in constitutional law and elsewhere) that is disconfirming but desirable, even wonderful, and we could also easily imagine information that is confirming but undesirable, even horrifying. In the face of conflict between confirmation bias and desirability bias, which bias is more powerful? In general? In law? Some evidence suggests that desirability bias wins out,106 but any generalization is likely to be treacherous.107

Suppose that judges find the implication of certain historical research undesirable – for example, because it leads to the conclusion that racial segregation is acceptable, that affirmative action is impermissible, that the national government can ban blasphemy without violating the First Amendment,108 that open-ended grants of discretion are acceptable, or that Congress has the authority to restrict the president’s removal authority. If so, they will not be inclined to believe it. And indeed, forms of constitutional tribalism, showing divisions on matters of historical fact, can plausibly be attributed, at least in part, to desirability bias.109

For those who seek to convince judges of a proposition that they do not want to believe, the natural reaction might again be one of despair. It would be useful in that connection to recall the advice that the late Archibald Cox, former Solicitor General as well as law professor, used to give to those asking about how to persuade the Supreme Court of the United States: “First you have to get them to want to agree with you. Then you have to show them how.” Cox was speaking of motivated reasoning. What he mentioned first was “want”; “how” followed. With respect to the issues discussed here, there is a great deal of “how,” and perhaps less in the way of “want.” Attention to the want question – by emphasizing, for example, how the goals of the nondelegation doctrine might be achieved through other means110 – might be a good place to start.111


The frequent impotence of careful historical research is hardly limited to the area of the separation of powers.112 The modest or nonexistent impact of painstaking historical work on nondelegation and the removal power is mirrored in many other domains, attesting to the tribal nature of constitutional law, even on basic questions of historical fact.113 This point raises challenges for some of the most appealing goals of the originalist project, which is designed, in large part, to discipline the exercise of judicial discretion by turning to history rather than abstract values, and thus to promote the rule of law.114

Judges have limited time and attention;115 they must rely on briefs; they do not know, and cannot be expected to know, all relevant historical research, or all of the most recent findings. Recall that like everyone else, they are not Large Language Models. They do not and cannot know everything; they must try to engage in optimal search.116 Like everyone else, they are part of epistemic communities, with predictable patterns of trust and distrust, and with the goals of expanding the ranks of their communities and of defeating competing communities. And if judges can be sorted into relevant teams, even on questions of historical fact – Team Nondelegation vs. Team Myth of Nondelegation, Team Campaign Expenditures Are Political Speech vs. Team Money Is Not Speech At All, Team Individual Gun Rights vs. Team Militia, and so forth – some of the central aspirations of originalism might be seen to be in jeopardy. Nonetheless, we may hope that some people will, over time, be seen to switch sides, at least if the evidence is indeed overwhelming.

Indeed we can go further. Epistemic communities include those who embrace identifiable constitutional methods. Team Originalism and Team Common Good Constitutionalism,117 Team Thayerism118 and Team Moral Readings,119 Team Minimalism120 and Team Democracy Reinforcement121 – each of these operates, at least some of the time, as an epistemic community. Of course members of such communities are in contact with one another, but far less often than would be ideal. There are private law analogues.122 The public also consists of epistemic communities, even on questions of law. Of course the public includes people with different levels of focus and expertise: law professors and lawyers, informed journalists, advocacy groups of various sorts, political activists, interested observers, uninterested observers. Information about constitutional issues will flow, in way or another, from and to such groups, and it will flow from some of those groups back to judges as well. There are feedback loops, sometimes helping to entrench and divide epistemic communities.

My emphasis has been on American constitutional law, but as I have suggested, the underlying mechanisms are general, even if they manifest themselves in different ways. Participants in politics engage in optimal search, and epistemic communities are sometimes created for that reason; consider the issue of immigration. When such communities form, it is often because of rational updating, as people credit different sources of information, depending on their priors; consider the issue of climate change.123 Motivated reasoning in general, and desirability bias in particular, play a central role in fueling polarization on questions of law and politics (id.). Epistemic communities form, and become entrenched, as a result of these mechanisms.

However we might think about the largest questions, some things are clear. The relative judicial silence that has greeted relevant separation of powers research is remarkable. It needs an explanation. Any such explanation must point to the existence of epistemic communities and to the relative intransigence of their members. That intransigence is, of course, a problem in many domains of law and politics; it extends far beyond the area of the separation of powers. Solving that problem should be a singularly high priority.


I am grateful to William Baude, Stephen Breyer, Jack Goldsmith, Stephen Sachs, and Lawrence Solum for valuable discussions and comments, and to David Olin and Victoria Yu for superb suggestions and research assistance. I am also grateful to participants in a superb legal theory workshop at the University of Pennsylvania Law School.


The author declares that he has no competing interests.


  1. For an important, relevant discussion, focusing on time constraints on Supreme Court justices, see Hart 1959. In some ways, this essay might be seen as a spiritual descendant of that one. [^]
  2. Cross 2013. The term “epistemic communities” originally emerged out of and is most often used in the field of international relations, in ways that overlap with but are more specific than my usage here. For overviews of the concept as it developed in international relations, see Haas 1992; 2021. For a valuable generalization of some of the ideas I am exploring here, see Hardin 2009. Epistemic communities can be found in politics, policy, science, law, and other fields, and they have been studied in those fields as well. See, e.g.: Dunlop (2017) and Mabon et al. (2019) on epistemic communities in policy formation; Maliniak 2021 on epistemic communities in climate change policy and with respect to public opinion; Abri and Buchheim 2022 on epistemic communities in misinformation, fake news, and the internet; and von Lingen 2022 on epistemic communities among lawyers. For a vigorous plea for epistemic diversity, see McBrayer 2024. [^]
  3. For a sampling, see, e.g.: Mortenson and Bagley 2021; 2022; Arlyck 2021); Chabot 2021; Parrillo 2021a,b; Whittington and Iuliano 2017; Mashaw 2012; Shugerman 2022a,b; 2023; Vlahoplus, 2021. The substantive topic here is the separation of powers, but there has been extraordinarily illuminating work in other domains, and some of it seems to call for large-scale revisions in longstanding understandings, at least for originalists. See Barnett and Bernick 2021 and Solum and Crema 2022. It remains to be seen whether and when such work will have influence. As noted below, there are hard questions, for originalists and others, about how to handle conflicts between settled law and new historical findings. See Barrett 2015. One might think, for example, that even if historical work showed that understood in terms of its original public meaning, the First Amendment does not protect blasphemy, it would be a mistake to rule, now, that the First Amendment does not protect blasphemy. (I think that would be a mistake.) See Anon 2021. Or one might think that even if new historical work showed that understood in terms of its original public meaning, the Second Amendment does not protect an individual right to possess firearms, it would be a mistake to rule, now, that the Second Amendment does not protect an individual right to possess firearms, and thus to overrule District of Columbia v. Heller, 554 U.S. 570 (2008). (I think that would be a mistake.) [^]
  4. This is admittedly a strong statement, depending on how we specify the “we.” Do “we” know more about the original public meaning of the Constitution than the founding generation did? That is not impossible, but it would be most surprising. Let us say, then, that “we” know more than we did in the last (say) 150 years. Not incidentally, particular credit is due to the careful analysis in Mashaw (2012) which seems to have been widely ignored; let us hope that his defining treatment will receive far more attention in the future. Also not incidentally, the Court’s current interest in the original public meaning of the founding document obviously incentivizes historical research; more than ever before (I think), such research would seem to be central to the proper resolution of constitutional debates. The tribalism identified here raises some cautionary notes, on which I touch lightly here. [^]
  5. I am bracketing some questions about how to identify the original public meaning; I will get to some of them in due course. For defining discussions, see Solum 2017a,b. [^]
  6. Mortenson and Bagley 2021. Posner and Vermeule 2002. Parrillo 2021a. Chabot 2021. [^]
  7. Whittington and Iuliano 2017. Mashaw 2012, p. 5 (“From the earliest days of the Republic, Congress delegated broad authority to administrators, armed them with extrajudicial coercive powers, created systems of administrative adjudication, and specifically authorized administrative rulemaking.”). [^]
  8. Mortenson and Bagley 2021. Mashaw 2012, pp. 45–46 (“More general claims that early congressional practice establishes a narrow view of could constitutionally be delegated to administrative officials are not convincing…Congress made broad delegations of authority in a host of other statutes, often to the President.”). [^]
  9. To be sure, there is a question here about the right conception of originalism; I will turn to that question below. [^]
  10. If you reject this account, or even despise it, please be patient. We will start over, and soon. (There is a method to my madness, if that is what you consider it.) [^]
  11. Shugerman 2023; Mashaw 2012; Lessig and Sunstein 1994. [^]
  12. See, e.g., Gundy v. United States, 139 S. Ct. 2116, 2131 (2019) (Gorsuch, J., dissenting). It is worth giving careful attention to Justice Gorsuch’s dissenting opinion, which offers an elaborate account of the supposed historical origins of the nondelegation doctrine, but without attending to the massive evidence that on originalist grounds, there is no such doctrine. Justice Gorsuch’s opinion might best be understood as structural and purposivist, rather than strictly originalist. [^]
  13. See Free Enter. Fund v. Pub. Co. Acct. Oversight Bd., 561 U.S. 477 (2010); Seila L. LLC v. Consumer Fin. Prot. Bureau, 140 S. Ct. 2183 (2020). It is worth emphasizing that Justice Kagan gives careful attention to the historical materials in her opinion in Seila Law. See ibid. at 2231 (Kagan, J., concurring in the judgment with respect to severability and dissenting in part) (“Contrary to the majority’s view, then, the founding era closed without any agreement that Congress lacked the power to curb the President’s removal authority. And as it kept that question open, Congress took the first steps— which would launch a tradition—of distinguishing financial regulators from diplomatic and military officers.”). Startlingly, the Court offered nothing of substance in response. [^]
  14. For a general set of findings to this effect, see Cross 2013. I am not seeking here to answer any general questions about originalism, though some of the analysis bears on those questions. [^]
  15. See, e.g.: Mortenson and Bagley 2021; Shugerman 2023. [^]
  16. On delegation see, e.g., Posner and Vermeule 2002; on both delegation and removal see, e.g., Mashaw 2012. Note also that Justice Kagan directed the Court’s attention to the historical materials in Seila Law and consider Mashaw’s summary: “Our failure to grasp how administration worked during nearly one half of the life of the Republic causes us to misunderstand the administrative constitution of the twenty-first century as well” (Mashaw 2012, p. 286). For older sources see: Myers v. United States, 272 U.S. 52, 285 n.73, 286 n.75 (1926) (Brandeis, J., dissenting); Corwin 1927, pp. 12–23, 12 n.22; Currie 1997. Here is a prominent statement from 2011: “Madison and his allies succeeded in their motions not because a majority of the House subscribed to the Madisonian view of presidential power, but rather because their strategic sequencing of motions allowed them to build coalitions on particular points with proponents of the other constitutional positions” (Manning 2011, p. 2031). A possible view is that the relevant decisions are structural and purposivist, not originalist, and that they stem from a judgment about the Constitution’s broader structural commitments, not about the original public meaning. In my view, that is the best foundation, and the actual foundation, of Seila Law. See Sunstein and Vermeule 2020. [^]
  17. It is true, of course, that many judges are nonoriginalists, and on nonoriginalist grounds, it would be possible to embrace unrestricted presidential removal power or the nondelegation doctrine. For one example, see Lessig and Sunstein 1994. (In case it is of interest, the present author would embrace some version of the nondelegation doctrine on nonoriginalist grounds, and would also favor stronger limits on congressional power to restrict the president’s removal authority than could be readily justified on originalist grounds.) [^]
  18. The removal power is an obvious example. Currently, those on the political right seem to like the idea of a strongly unitary executive, and those on the political left do not, and the disagreement is mirrored in constitutional law, with right-of-center judges predictably differing from left-of-center judges. But why should that issue turn out to be tribal? That is a genuine puzzle. After all, some presidents are Democrats, and they too might well like to assert broad control over the administrative state. A possible answer has to do with the arc of history, with path dependency, and in a sense with accident: Who occupies the presidency in certain historical periods? Who finds it useful to make one argument at a given point in time? Who rejects that argument, and offers a different argument, partly or even mostly because a political opponent is making it? Who ends up thinking that the argument they are making outstrips a particular historical moment? If those on the political right embrace a strongly unitary executive, and if those on the political left deplore that idea, we might want to consider the possibility that with a little twist or turn, the sides might have flipped. For relevant discussion, see Sunstein 2024. At first glance, the nondelegation doctrine is similar; it could be embraced or rejected by either the left or the right (see Ely 1981). But if the court restricts Congress’ power to grant broad discretion to administrative agencies, we should expect to see less in the way of federal regulation, and that fact might help explain the political or ideological division, with right-of-center judges being drawn to the nondelegation doctrine, and left-of-center judges not being drawn to it. The struggle over the nondelegation doctrine might be seen as a struggle over the New Deal and the Great Society of the 1960s. [^]
  19. This phenomenon raises important questions about what might be seen as the sociology of law: How do such teams arise? How does such sorting occur? What is the role of politics? Why might people on the right choose one side, and people on the left choose another? What is the role of accident? I will offer a few observations on those questions. [^]
  20. See, e.g.: Wurman 2021; Lawson 2002; Hamburger 2014; 2023; Epstein 2021; Gordon 2023. [^]
  21. Wurman 2021. [^]
  22. Bamzai and Prakash 2023. Prakash 2006. [^]
  23. Bamzai and Prakash 2023. Prakash 2006. [^]
  24. See, e.g., Gundy v. United States, 139 S. Ct. 2116 (2019). [^]
  25. See Seila L. LLC v. Consumer Fin. Prot. Bureau, 140 S. Ct. 2183, 2224 (2020) (Kagan, J., dissenting). This statement is unfair to Justice Kagan, who did grapple with the historical materials in some detail. But my method here has a madness. [^]
  26. Importantly, some of these judges may not be originalists, which means that they need not by bound by the original public meaning. But they have referred to the historical evidence. See Seila L. LLC v. Consumer Fin. Prot. Bureau, 140 S. Ct. 2183, 2225 (2020) (Kagan, J., dissenting); Gundy, 139 S. Ct. at 2123. Nonoriginalists might deem the original public meaning to be relevant even if not authoritative. [^]
  27. See, e.g.: Wurman 2021; Bamzai and Prakash 2023. [^]
  28. On delegation, see, e.g., Lawson 2002; on removal, see, e.g., Prakash 2006. [^]
  29. It should be clear that I offer three starts to underline my main thesis, which is not to take an excessively firm stand on what the historical evidence actually shows, but to emphasize that it seems to be having little or no effect on any judge or justice, at least in dislodging their antecedent convictions. The third start reflects my own view of the historical evidence, and I am not shy about endorsing it. [^]
  30. Mortenson and Bagley 2021. Whittington and Iuliano 2017. [^]
  31. Mashaw 2012, p. 39 (“The delegates to the Constitutional Convention seem to have been so fixated on the question of appointment that they virtually ignored the question of removal”). Mortenson and Bagley 2021, p. 293 (“there is trifling evidence of a nondelegation doctrine even being argued for by aggressive legal innovators, let alone broadly accepted by the Founders as a group”). [^]
  32. See Posner and Vermeule (2002, pp. 1735–36) listing broad early delegations. [^]
  33. Wurman 2021, p. 1556 (“nondelegation is at least consistent with discussions in the First Congress”). [^]
  34. Mortenson and Bagley 2021, p. 280. [^]
  35. Ibid. [^]
  36. See generally Posner and Vermeule 2002. [^]
  37. See, e,g.: Shugerman 2023; Prakash 2006. [^]
  38. Hamilton 1788. [^]
  39. Madison 1788. [^]
  40. See The Congressional Register (June 16, 1789), reprinted in Bickford et al. 1992, vol. 11, p. 868. [^]
  41. Shugerman 2023, p. 776 (“Madison himself did not pick a side”). [^]
  42. The Congressional Register (May 19, 1789), reprinted in Bickford et al. 1992, vol. 10, p. 730. [^]
  43. See Shugerman (2023, p. 779) describing Madison “embrac[ing] the presidential side for a little over one full week.” [^]
  44. Shugerman 2023, pp. 826–33. See also Mashaw 2012, p. 49 (“James Madison suggested that the Comptroller should not hold office subject to presidential removal, but should be given a term of years.”). Some people have interpreted Madison’s views on the Comptroller as supportive of unrestricted removal authority. See, e.g., Wurman 2020, p. 142 n.205; PHH Corp. v. Consumer Fin. Prot. Bureau, 881 F.3d 75, 177 n.4 (D.C. Cir. 2018) (Kavanaugh, J., dissenting). It is a plausible view. But it appears to be a mistake, for reasons explained by Shugerman (which means that Mashaw got it right). See Shugarman (2023, pp. 828–33) responding to what he calls “misunderst[andings]” of Madison’s position. [^]
  45. See Shugerman (2023, p. 840) listing those who agreed with Madison’s presidentialist view. [^]
  46. For a broad overview, see generally Mashaw 2012. He writes that “if practice is any guide, early Congresses created departments and officers, charged them with administrative tasks, and subjected them to political supervision in ways that exhibit modest concern for rigid or formal conceptions of the separation of powers” (Mashaw 2012, p. 44). [^]
  47. See Shugerman (2023, p. 840) noting that the “presidentialists mostly gave up on pursuing the theory.” [^]
  48. There are complex issues here about originalist methodology, which I mostly bracket. History is not the same as the original public meaning. See generally Solum 2017b. As I have noted, there are also important questions about which officers may be subject to which kinds of restrictions on removal authority. The Secretary of State is plausibly not the same as the members of the Consumer Product Safety Commission. [^]
  49. Shugerman 2023, p. 757. [^]
  50. Ibid. The view is vigorously and adeptly contested in Bamzai and Prakash (2023), but in my view, Shugerman’s very detailed account is essentially decisive. [^]
  51. See generally Shugerman 2023. Mashaw (2012) suggests a more definitive conclusion, to the effect that the founding generation did not believe in unrestricted removal power. I am bracketing the possibility that on one or another of the issues here, judges are in the “construction zone.” See Solum 2010. [^]
  52. I will get to some complexities in the idea of original public meaning in short order. [^]
  53. See generally: Mortenson and Bagley 2021; Mashaw 2012. [^]
  54. This is the lesson that seems to emerge from Shugerman 2023. One more time: I am bracketing the question of whether we should distinguish among different kinds of principal officers. It is one thing to say that (for example) Congress can immunize members of the Federal Reserve Board or the National Labor Relations Board from unrestricted removal authority; it is another thing to say that Congress can immunize the Secretary of State or the Secretary of Defense from that authority. [^]
  55. The silence greeting Mashaw’s Creating the Administrative Constitution is particularly noteworthy, because the book was published in 2012. [^]
  56. Note that scholars in other fields have written about how “competing epistemic communities” can influence whether new knowledge is adopted. See, e.g., Cohen (2012) observing that the existence of “three different, and occasionally competing, epistemic communities” shaped the develop of a climate policy and Löblová (2018) observing similarly in the public health context. [^]
  57. I say this with humility; I know that I may be wrong. [^]
  58. Some of the discussion in Mortenson and Bagley (2021) seems to suggest that the founding generation did not believe in any nondelegation doctrine, but in my view, a more cautious formulation is consistent with their evidence (and with some of what they actually find and say). [^]
  59. Mortenson and Bagley 2021. [^]
  60. See, e.g., Shugerman (2023, p. 753) observing that “nine of fifty four participating representatives” did “explicitly endorse[] the presidentialist view that Article II implied a presidential removal power.” [^]
  61. Solum 2017a. [^]
  62. Solum 2019. [^]
  63. For relevant discussion, see Barrett 2015. [^]
  64. How long? Perhaps not all that long. See generally Posner and Vermeule 2002. But the doctrine unquestionably dates to 1935. See A.L.A. Schechter Poultry Corp. v. United States, 295 U.S. 495 (1935). [^]
  65. See Myers v. United States, 272 U.S. 52 (1926). [^]
  66. Humphrey’s Ex’r v. United States, 295 U.S. 602 (1935). [^]
  67. Sunstein 2006. [^]
  68. Gundy v. United States, 139 S. Ct. 2116, 2131 (2019) (Gorsuch, J., dissenting). [^]
  69. See Seila L. LLC v. Consumer Fin. Prot. Bureau, 140 S. Ct. 2183, 2198 (2020) (describing earlier cases); Free Enter. Fund v. Pub. Co. Acct. Oversight Bd., 561 U.S. 477, 483 (2010) (similar). [^]
  70. See Seila, 140 S. Ct. at 2198; Free Enter. Fund, 561 U.S. at 483. [^]
  71. For the record: I used to believe that on originalist grounds, the president had unrestricted removal authority, and then shifted given what I came to see as the weight of the historical evidence (see Lessig and Sunstein 1994), and then shifted back again after reading Prakash (2006), and now have shifted yet again on the strength of Shugerman (2023). I used to believe that the nondelegation doctrine was built into the founding settlement, and part of the original public meaning, but extremely challenging for courts to enforce. See Sunstein 2000. I am now far less confident that there was any such doctrine, on the strength of the evidence in Mortenson and Bagley (2021); but I retain the view that in very extreme cases, the historical evidence is not inconsistent with the view that a nondelegation doctrine exists. (Maybe I am unusually impressionable?) [^]
  72. See Seila, 140 S. Ct. at 2198; Gundy, 139 S. Ct. at 2131. [^]
  73. For relevant discussion, focusing on information search more broadly, see Hardin 2009. [^]
  74. See, e.g., Mashaw 2012. [^]
  75. I single out these reviews against interest, because they are two of my favorites. [^]
  76. The term “epistemic communities” has been used for multiple purposes in many different fields, and I mean to understand it in an ordinary language sense here. For one usage, see Zollman 2013. Some political scientists have started to use the term to describe the judiciary and how ideas are developed among groups of judges and legal thinkers. See, e.g.: Hollis-Brusky 2015; Christensen et al. 2011. [^]
  77. Zagzebski 2012. Hardin 2009. [^]
  78. 410 U.S. 113 (1973). [^]
  79. Some evidence in support of that assumption comes from Justice Kagan’s dissenting opinion in Seila Law, which deals with the historical evidence, including some recent historical evidence, in depth. In light of the dissenting opinion, those in the majority cannot possibly be unaware of the historical evidence. [^]
  80. An attempted synthesis can be found in Sharot et al. 2022. [^]
  81. I do not mean to say anything controversial about Bayesianism here. The simple idea is that people are rationally updating, given their priors. If you are inclined to believe that the earth goes around the moon, or that the Holocaust happened, apparently contrary evidence might not much move your prior beliefs; you evaluate new information in light of those beliefs. For illuminating discussion, see Corfield and Williamson 2001. [^]
  82. See generally: Lawson 2002; Wurman 2021. [^]
  83. An article-length elaboration of this claim would be worthwhile. In brief, it might be urged that seemingly broad terms need not be understood to be so broad, if they are taken in context. Apparently open-ended words, such as “reasonably necessary” or “feasible,” might not be so open-ended if so taken. In this vein, see Gundy v. United States, 139 S. Ct. 2116 (2019) (examining the word “feasible”); see generally Sunstein 2023. [^]
  84. On this point, see Gundy, 139 S. Ct at 2140 (Gorsuch, J., dissenting) (emphasizing a category of cases in which grants of discretion are permissible in light of the president’s constitutional authorities). Some of the examples in Morgenson and Bagley (2021) can be understood in this way. [^]
  85. See generally Wurman 2021. [^]
  86. Wurman 2021, p. 1495 (“Not only was Madison consistent in his view, but many others in the Founding generation appear to have shared it.”). [^]
  87. U.S. Constitution, art. II, § 1. [^]
  88. I confess that at one point, I had that very reaction. [^]
  89. See, e.g., Prakash 2006. [^]
  90. Prakash 2006. Bamzai and Prakash 2023. [^]
  91. Mashaw 2012. [^]
  92. Some scholars have written on the phenomenon of “epistemic closure” in the judiciary. See Tushnet 2019. [^]
  93. For relevant discussion, see Machamer and Miller 2021. Note as well that judicial judgments are not fixed or firm on many topics. Judges might have an inclination to believe X or Y on some topic, but they might believe that with a probability of 55%, or 60%, or 65%; if they receive contrary evidence, they might shift to believing it with a probability of 40% - and so vote the other way. Much of advocacy consists of efforts to that effect. [^]
  94. Epley and Gilovich 2016. [^]
  95. It would be valuable to know which of the three accounts offered here – optimal search, Bayesian updating, and motivated reasoning – is in fact the most important. The answer is of course likely to differ in different domains. We could learn something about optimal search by investigating briefs: If they are silent or sparse on (say) the original public meaning, we would have reason to suspect that judges do not have relevant information, and if they are full of informative detail, perhaps the real driver is Bayesianism or motivated reasoning. Of course amicus briefs could pick up some of the slack, though there is a question whether and want judges read them. Much more might be learned on this topic; we could imagine a host of empirical projects here. [^]
  96. But perhaps not that far afield. See Sunstein 2018. [^]
  97. Rinzler 2013, p. 64. [^]
  98. Epley and Gilovich 2016, p. 136. [^]
  99. There is an incipient literature in law. See Sood 2013. [^]
  100. Epley and Gilovich 2016, p. 137. [^]
  101. Ibid. [^]
  102. See, e.g.: Kappes et al. 2020; Rollwage et al. 2020; Nickerson 1998. [^]
  103. It might be simply Bayesian. I believe that Labrador Retrievers tend to be very nice dogs, and if I receive evidence suggesting that they are not so nice, I will not be swayed by such evidence, because (in my view) my antecedent belief is exceedingly well-founded. [^]
  104. Lord et al. 1979. [^]
  105. Tappin et al. 2017. [^]
  106. Ibid. [^]
  107. The reason is that context greatly matters. For evidence that confirmation bias might trump desirability bias, see Sunstein et al. 2017. [^]
  108. Anon. 2021. [^]
  109. A more extreme view, offered by several readers of an earlier version of this Article, is that some judges are acting in bad faith. They know that the historical materials do not support their position, but even so, they say, in public, that the historical materials offer that support. In some sense, they are lying. For two reasons, I do not engage that view. (1) I do not believe it to be true. It is exceedingly difficult for a judge to have this psychological state: “I know that X is false, but I will say that X is true.” (2) Even one believed that, with some probability, some justices are knowingly misstating relevant facts (including historical facts), it is best to proceed with a principle of charity, and to assume that they are acting in good faith. [^]
  110. Mashaw 2018; 1985; Stiglitz 2022. [^]
  111. One possibility is that judges are less certain about the relevant evidence than their opinions suggest, and that they are aware of the historical disputes – either in detail or more broadly. An opinion can be seen as an exercise in rhetoric, and a Supreme Court has a choice. An opinion might say, “In light of the historical materials, it is clear that on the original understanding, the President must have unrestricted removal power over his subordinates.” Or an opinion might say, “The original understanding is unclear, and reasonable people differ, but the better view, we think, is that the President must have unrestricted removal power over his subordinates.” A justice might write the former sentence even if the latter is a more accurate reflection of his or her convictions. [^]
  112. Cross 2013. [^]
  113. Possible examples include the Second Amendment, see District of Columbia v. Heller, 554 U.S. 570 (2008), and affirmative action, see Students for Fair Admissions v. President and Fellows of Harvard College, 600 U.S. 181 (2023). I am bracketing the question of which side is right on the historical materials. What is clear, in both contexts, is that the justices are reading the historical materials in a way that does not make them sad – a clear signal of motivated reasoning (rather than rational updating, though we cannot exclude the possibility that rational updating is at work). [^]
  114. For different perspectives, see: Scalia 1989; Solum 2019; Barnett and Bernick 2021. [^]
  115. See generally Hart 1959. [^]
  116. To be sure, Large Language Models should alter, and are altering, the concept of what is optimal search. But how much of what emerges can be trusted? Bayesianism and motivated reasoning are relevant there. [^]
  117. Vermeule 2022. [^]
  118. Thayer 1893. A valuable discussion of Thayer’s motivations, emphasizing what he sees as Thayer’s political conservatism and desire to activate political focus on combating ill-considered progressivism, is Tushnet 1993. [^]
  119. Dworkin 1985. [^]
  120. Sunstein 1999. [^]
  121. Ely 1981. [^]
  122. See, as foundational texts for some of these: Posner 1972; Epstein 1973; Kaplow and Shavell 2006. In private law too, I do not mean to suggest that epistemic communities in these domains are closed; there is of course a great deal of conversation across communities. Some epistemic communities are more closed than others. [^]
  123. Sunstein et al., 2017 [^]


Abri, Gilad and Johannes Buchheim. 2022. Beyond true and false: fake news and the digital epistemic divide. Michigan Telecommunications and Technology Law Review, 29: 59. DOI:

Anon. 2021. Note: blasphemy and the original meaning of the First Amendment. Harvard Law Review, 135: 689.,disputes%20between%20learned%20men%20upon

Arlyck, Kevin. 2021. Delegation, administration, and improvisation. Notre Dame Law Review, 97: 243.

Bamzai, Aditya and Saikrishna Bangalore Prakash. 2023. The executive power of removal. Harvard Law Review, 136: 1758.

Barnett, Randy and Evan Bernick. 2021. The Original Meaning of the Fourteenth Amendment. Cambridge, MA: Harvard University Press. DOI:

Barrett, Amy Coney. 2015. Originalism and stare decisis. Notre Dame Law Review, 92: 1921.

Bickford, Charlene Bangs; Kenneth R. Bowling; William C. di Giacomantonio; and Helen E. Veit, eds. 1992. Documentary History of the First Federal Congress of the United States of America, March 4, 1789-March 3, 1791. Baltimore, MD: Johns Hopkins University Press.

Chabot, Christine Kexel. 2021. The lost history of delegation at the founding. Georgia Law Review, 56: 81.

Christensen, Ericka; James Newman; and Jason McConnell. 2011. A statistical treatment of legal epistemic communities and the Department of Justice: the impact upon judicial selection and judicial decision-making. Paper presented to the 2011 Annual Meeting of the American Political Science Association.

Cohen, Alice. 2012. Rescaling environmental governance: watersheds as boundary objects at the intersection of science, neoliberalism, and participation. Environment & Planning A: Economy & Space, 44: 2207. DOI:

Corfield, D. and J. Williamson, eds. 2001. Foundations of Bayesianism. Dordrecht: Springer. DOI:

Corwin, Edward S. 1927. The President’s Removal Power Under the Constitution. New York: National Municipal League. DOI:

Cross, Frank. 2013. The Failed Promise of Originalism. Stanford, CA: Stanford University Press. DOI:

Cross, Mai’a K. Davis. 2013. Rethinking epistemic communities twenty years later. Review of International Studies, 39: 137. DOI:

Currie, David P. 1997. The Constitution in Congress: The Federalist Period 1789-1801. Chicago: University of Chicago Press.

Dunlop, Claire A. 2017. The irony of epistemic learning: epistemic communities, policy learning and the case of Europe’s hormones saga. Policy & Society, 36: 215. DOI:

Dworkin, Ronald. 1985. Law’s Empire. Cambridge, MA: Harvard University Press.

Ely, John Hart. 1981. Democracy and Distrust. Cambridge, MA: Harvard University Press. DOI:

Epley, Nicolas and Thomas Gilovich. 2016. The mechanics of motivated reasoning. Journal of Economic Perspectives, 30: 133. DOI:

Epstein, Richard A. 1973. A theory of strict liability. Journal of Legal Studies, 2: 151. DOI:

Epstein, Richard A. 2021. Delegation of powers: a historical and functional analysis. Chapman Law Review, 24: 659.

Gordon, Aaron. 2023. Nondelegation misinformation: a reply to the skeptics. Baylor Law Review, 75: 152.

Haas, Peter. 1992. Introduction: epistemic communities and international policy coordination. International Organization, 46: 1. DOI:

Haas, Peter. 2021. Epistemic communities. Pp. 698 ff. in The Oxford Handbook of International Environmental Law, 2d edition, ed. Lavanya Rajamani and Jacqueline Peel. Oxford: Oxford University Press. DOI:

Hamburger, Philip. 2014. Is Administrative Law Unlawful? Chicago: University of Chicago Press. DOI:

Hamburger, Philip. 2023. Nondelegation blues. George Washington Law Review, 91: 1083.

Hamilton, Alexander. 1788. The Federalist no. 77. Pp. 515 ff. in The Federalist, ed. Jacob E. Cooke. Middletown, CT: Wesleyan University Press, 1961.

Hardin, Russell. 2009. How Do You Know? The Economics of Ordinary Knowledge. Princeton, NJ: Princeton University Press. DOI:

Hart, Henry M. 1959. The Supreme Court, 1958 term - Foreword: the time chart of the justices. Harvard Law Review, 73: 84. DOI:

Hollis-Brusky, Amanda. 2015. Ideas with Consequences: The Federalist Society and the Conservative Counterrevolution. New York: Oxford University Press.

Kaplow, Louis and Steven Shavell. 2006. Fairness versus Welfare. Cambridge, MA: Harvard University Press.

Kappes, Andreas; Ann H. Harvey; Terry Lorenz; P. Reed Montague; and Tali Sharot. 2020. Confirmation bias in the utilization of others’ opinion strength. Nature Neuroscience, 23: 130. DOI:

Lawson, Gary. 2002. Delegation and original meaning. Virginia Law Review, 88: 327. DOI:

Lessig, Lawrence and Cass R. Sunstein. 1994. The president and the administration. Columbia Law Review, 94: 1. DOI:

Löblová, Olga. 2018. Epistemic communities and experts in health policy-making. European Journal of Public Health, 28 (supplement 3): 7. DOI:

Lord, Charles G.; L. Ross; and M.R. Lepper. 1979. Biased assimilation and attitude polarization. Journal of Personality & Social Psychology, 37: 2098. DOI:

Mabon, Leslie; Won-Yu Shih; Koyoko Kodo; Hiroyuki Kanekiyo; and Yuriko Hayabuchi. 2019. What is the role of epistemic communities in shaping local environmental policy? Managing environmental change through planning and greenspace in Fukuoka City, Japan. Geoforum, 104: 158. DOI:

Machamer, Peter and David Marshall Miller. 2021. Galileo Galilei. In Stanford Encyclopedia of Philosophy, ed. Edward N. Zalta.

Madison, James. 1788. The Federalist no. 39. Pp. 250 ff. in The Federalist, ed. Jacob E. Cooke. Middletown, CT: Wesleyan University Press, 1961.

Maliniak, Daniel; Eric Parajon; and Ryan Powers. 2021. Epistemic communities and public support for the Paris agreement on climate change. Political Research Quarterly, 74: 866. DOI:

Manning, John F. 2011. Separation of powers as ordinary interpretation. Harvard Law Review, 124: 1939.

Mashaw, Jerry. 1985. Prodelegation: why administrators should make political decisions. Journal of Law, Economics & Organization, 1: 81. DOI:

Mashaw, Jerry. 2012. Creating the Administrative Constitution. New Haven, CT: Yale University Press.

Mashaw, Jerry. 2018. Reasoned Administration and Democratic Legitimacy. Cambridge: Cambridge University Press. DOI:

McBrayer, Justin. 2024. The epistemic benefits of ideological diversity. Acta Analytica, forthcoming. DOI:

Mortenson, Julian Davis and Nicholas Bagley. 2021. Delegation at the founding. Columbia Law Review, 121: 277.

Mortenson, Julian Davis and Nicholas Bagley. 2022. Delegation at the founding: a response to critics. Columbia Law Review, 122: 2323. DOI:

Nickerson, Raymond. 1998. Confirmation bias: a ubiquitous phenomenon in many guises. Review of General Psychology, 2: 175. DOI:

Parrillo, Nicholas R. 2021a. A critical assessment of the originalist case against administrative regulatory power: new evidence from the federal tax on private real estate in the 1790s. Yale Law Journal, 130: 1288. DOI:

Parrillo, Nicholas R. 2021b. Supplemental Paper to “A critical assessment of the originalist case against administrative regulatory power: new evidence from the federal tax on private real estate in the 1790s.” DOI:

Posner, Eric and Adrian Vermeule. 2002. Interring the nondelegation doctrine. University of Chicago Law Review, 69: 1721.,executive%20power%2C%20not%20legislative%20power

Posner, Richard A. 1972. A theory of negligence. Journal of Legal Studies, 1: 29.

Prakash, Saikrishna. 2006. New light on the decision of 1789. Cornell Law Review, 91: 1021.,with%20its%20rival%2C%20the%20executive

Rinzler, J.W. 2013. The Making of Star Wars: Return of the Jedi. New York: Random House.

Rollwage, Max; Alisa Loosen; Tobias U. Hauser; Rani Moran; Raymond J. Dolan; and Stephen M. Fleming. 2020. Confidence drives a neural confirmation bias. Nature Communications, 11: article # 2634. DOI:

Scalia, Antonin. 1989. Originalism: the lesser evil. University of Cincinnati Law Review, 57: 849.

Sharot, Tali; Max Rollwage; Cass R. Sunstein; and Stephen M. Fleming. 2022. Why and when beliefs change. Perspectives on Psychological Science, 18: 142. DOI:

Shugerman, Jed H. 2022a. Removal of context: Blackstone, limited monarchy, and the limits of unitary originalism. Yale Journal of Law & the Humanities, 33: 125. DOI:

Shugerman, Jed H. 2022b. Vesting. Stanford Law Review, 74: 1479.

Shugerman, Jed H. 2023. The indecisions of 1789: inconstant originalism and strategic ambiguity. University of Pennsylvania Law Review, 171: 753.

Solum, Lawrence B. 2010. The interpretation-construction distinction. Constitutional Commentary, 27: 95.

Solum, Lawrence B. 2017a. Originalist methodology. University of Chicago Law Review, 84: 269.

Solum, Lawrence B. 2017b. Triangulating public meaning: corpus linguistics, immersion, and the constitutional record. BYU Law Review, 2017: 1621. DOI:

Solum, Lawrence B. 2019. Originalism versus living constitutionalism: the conceptual structure of the great debate. Northwestern University Law Review, 113: 1243. DOI:

Solum, Lawrence B. and Max Crema. 2022. The original meaning of “due process of law” in the Fifth Amendment. Virginia Law Review, 108: 447.

Sood, Avani Mehta. 2013. Motivated cognition in legal judgments—an analytic review. Annual Review of Law and Social Science, 9: 307. DOI:

Stiglitz, Edward. 2022. The Reasoning State. Cambridge: Cambridge University Press.

Sunstein, Cass R. 1999. One Case at a Time. Cambridge, MA: Harvard University Press.

Sunstein, Cass R. 2000. Nondelegation canons. University of Chicago Law Review, 67: 315. DOI:

Sunstein, Cass R. 2006. Burkean minimalism. Michigan Law Review, 105: 353.

Sunstein, Cass R. 2018. The World According to Star Wars, rev. ed. New York: HarperCollins.

Sunstein, Cass R. 2023. It all started with benzene. DOI:

Sunstein, Cass R. 2024. How to Become Famous. Cambridge, MA: Harvard Business Review Press.

Sunstein, Cass R.; Sebastian Bobadilla-Suarez; Stephanie C. Lazzaro; and Tali Sharot. 2017. How people update beliefs about climate change. Cornell Law Review, 102: 1431.

Sunstein, Cass R. and Adrian Vermeule. 2020. The unitary executive. Supreme Court Review, 2020 1. DOI:

Tappin, Ben M.; Leslie van der Leer; and Ryan T. McKay. 2017. The heart trumps the head: desirability bias in political belief revision. Journal of Experimental Psychology: General, 146: 1143. DOI:

Thayer, James Bradley. 1893. The origin and scope of the American doctrine of constitutional law. Harvard Law Review, 7: 129. DOI:

Tushnet, Mark. 1993. Thayer’s target: judicial review or democracy? Northwestern University Law Review, 88: 9.

Tushnet, Mark. 2019. Epistemic closure and the Schechter case. Harvard Public Working Paper No. 19-42. DOI:

Vermeule, Adrian. 2022. Common Good Constitutionalism. Cambridge: Polity Press.

Vlahoplus, John. 2021. Early delegations of federal powers. George Washington Law Review Arguendo, 89: 55. DOI:

von Lingen, Kerstin. 2022. Epistemic communities of exile lawyers at the UNWCC. Journal of the History of International Law, 24: 315. DOI:

Whittington, Keith and Jason Iuliano. 2017. The myth of the nondelegation doctrine. University of Pennsylvania Law Review, 165: 379.

Wurman, Ilan. 2020. In search of prerogative. Duke Law Journal, 70: 93.

Wurman, Ilan. 2021. Nondelegation at the founding. Yale Law Journal, 130: 1288.

Zagzebski, Linda. 2012. Epistemic Authority: A Theory of Trust, Authority, and Autonomy in Belief. Oxford: Oxford University Press. DOI:

Zollman, Kevin J.S. 2013. Network epistemology: communications in epistemic communities. Philosophy Compass, 8: 15. DOI:




  • Views: 569
  • Downloads: 131


Download RIS Download BibTeX

File Checksums

  • PDF: 471bbdecceceb3fed15211edf433cf90
  • XML: 9ef82ad011666b1e907dd6d9741b8d97

Table of Contents