"Steelmanning" and Interpretive Charity

Two different useful forms of good argumentative hygiene.


One of the things lawyers need to do, and therefore one of the things I try to teach my students how to do, is to understand the strongest arguments against their position. Lots of people agree that this is important in principle, but can falter in practice. (Look, I understand the strongest arguments against my position, it's just that they are all bad arguments! That's why my position is correct!)

So a useful exercise to get there is what some people on the internet call "steelmanning." A steel man is the opposite, of course, of a straw man. Steelmanning "is the art of addressing the best form of the other person's argument, even if it's not the one they presented." For instance, given the inconsistent quality of judicial opinions, this can be a very useful supplement to just reading cases if you are trying to think through an area of law.

Indeed, I now sometimes test a version of this skill on my exams, asking students to write up both sides of an argument, with the rule that their grade will be based on the quality of the worse of the two arguments. (I discussed this a little bit towards the end of my recent podcast appearance on Rationally Speaking with Julia Galef.)

On Twitter last night, my friend Jacob Levy asked whether this practice is really just a re-labeled version of the older notion of "interpretive charity," in which we try to put others' claims in a sympathetic rather than an unsympathetic light.

Reflecting about it, I think these two strategies are obviously related, and both are forms of good intellectual hygiene, but there are some differences.

Interpretive charity is ultimately more interpretive than steelmanning. There is a potential limit to interpretive charity if you think: I know he should have said X, but he just plainly rejected X or didn't think of it. "You have to be able to declare, at some point: 'For crying out loud, he never would have thought of that!'" By contrast, with steelmanning it doesn't really matter what a particular person said. It just matters what the best argument is.

Relatedly, there are multiple ways one can be "charitable" to somebody else's claim. One can try to make it as true as one can. One can try to make it as consistent or coherent as one can. One can try to make it fit in to a broader philosophical world view, even if it's a world view that you don't hold. One can try to trim off particularly ugly implications. Etc. Steelmanning, by contrast, focuses on making an argument as true as possible.

Relatedly relatedly, I've sometimes seen people use interpretive charity in ways that actually make an argument easier for them to reject. Justice so-and-so says X, which seems weird to me. But maybe I'll be charitable and assume that this argument was motivated by his underlying political philosophy. And I reject his underlying political philosophy because it is wrong/uninformed/racist/whatever, so I can now safely reject his argument. This may indeed be charitable to the speaker; but steelmanning is less focused on the speaker and more on the listener. What is the version of this argument that hits closest to home for me?

Again, both of these can be useful tools. I don't think steelmanning is categorically superior to interpretive charity, even when trying to get at the truth. For instance interpretive charity might be a better tool for understanding the argument of an epistemic peer—somebody whose arguments may well be persuasive to you if you only understood them better.

By contrast, steelmanning may be a better tool when dealing with the arguments of those who start from different premises. Rather than safely reduce their argument to its premise and then reject the premise, you can be forced to search for other premises that might ground the argument closer to where you stand.

Regardless of the differences between these two, both ought to be a standard part of our critical thinking toolbox.

NEXT: A Potential Way Forward for Nondelegation Concerns

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  1. While we’re at it, why not teach students to make the best case for *their own* arguments?

    1. The disputation method of arriving at a decision, let alone, the truth is quackery from Scholasticism. It conveniently generates a lot of worthless, but billable lawyer hours. It violates the Establishment Clause. It is fraudulent. It is quackery. It results in high rates of false negatives, and false positives. Outside of making money, this quackery is worthless and a criminal, fraudulent enterprise by the lawyer profession.

      1. Your profession stinks at the deepest most fundamental level. I await your steel man reply.

        1. Charity forbids baiting the Behar

          1. Two sides to the argument.

            Argument 1: Don Nico is a good, kind hearted person who can’t bring himself to so easily abuse an individual who is clearly mentally deficient.

            Argument 2: Charity is Don Nico’s wife, and she’ll beat him with a proverbial stick if he continues to argue online.

            1. Being kind to Behar is today’s mitzvah

              1. Try sending your steel man. So far, your personal insults just violate the Fallacy of Irrelevance, and prove my point about what happens to the intellectual level after passing 1L. Every self stated goal of every law subject is in utter failure. What more can you want as evidence? The sole success is the rent, collected at the point of a gun.

                1. I have nothing to send you Behar. Today’s comments contain no insults. Try rereading.

                2. Once more: there is no such thing as violating a fallacy.

                  1. David. Violating any of these fallacies violates the Fifth Amendment procedural due process right to a fair hearing. It makes the assertion irrational. This is from high school course in critical thinking thus beyond your grasp. You can use this list as a checklist for dismissal of a case. The Rules of Evidence and of Criminal Procedure are filled with these. Here is a really big one, in your profession, appeal to authority. Another is fallacy of the single cause, ridiculous today. The list of violations is long, and each is unconstitutional.


                    1. I have suggested a course in critical thinking as a pre-law course to immunize students against the utter idiocy of 1L. It almost all garbage, and needs to be restarted from scratch.

                    2. Violating any of these fallacies

                      Sigh. There. Is. No. Such. Thing. As. Violating. A. Fallacy.

                      Let’s say I appeal to authority. Would I be violating the fallacy? No. I would be committing it.

                    3. You would be committing a fallacy and violating the Fifth Amendment. Good point.

  2. I think the biggest problem would involve accepting the legitimacy of an opposing argument.

    1. Both accepting the legitimacy of an opposing argument, AND the risk of having one’s ‘steelman’ attributed to you.

      Under current conditions, putting forward a strong argument for the opposing side might be considered by people on your side of the argument as evidence you actually WERE on the other side.

      It’s still a good technique, just that there are a lot of trends that make clear reasoning dangerous these days.

      1. Yup. See the responses of some commenters here when Robby Soave or Ilya Somin make even a slight acknowledgement of opposing arguments.

          1. I’m not going to call out anyone by name. There have been plenty of cases where Soave clearly took a strong pro-free-speech position and then got hammered merely for mentioning the counterarguments or uncertainty in how courts would rule.

            1. For the very best duck salad, try Poggio in Sausalito. Highly recommended

      2. “Under current conditions, putting forward a strong argument for the opposing side might be considered by people on your side of the argument as evidence you actually WERE on the other side.”

        Under current conditions, that’s a problem for people merely suggesting that the other side out to be permitted to make their argument. See the Harper’s Letter reaction, for example.

        1. *ought.

    2. Ed,
      If you don’t try to see the situation from the other party’s point of view, you will never be a good negotiator or even realize how wrong you may be.

      1. I don’t think he’s arguing for not acknowledging alternate points of view, but rather that doing so can, by itself, render you subject to cries of “racist”, “sexist”, “alt-right” by erstwhile allies.

      2. I fully agree — my issue is the social acceptance that there possibly could *be* a legitimate opposing argument, which doesn’t exist anymore.

        In an earlier (saner) era, I used to do things like challenge high school students by saying Hitler did do some good things — he did, e.g. Autobahn, Volkswagen, Berlin subway system. I’d never dare do that now…

  3. Indeed, I now sometimes test a version of this skill on my exams, asking students to write up both sides of an argument, with the rule that their grade will be based on the quality of the worse of the two arguments.

    This is interesting. Sort of like, “You cut the cake, and I’ll get first choice of slices.”

    1. Yeah, this part stood out to me as a really great way to test students. It also helps mitigate professors own biases in grading, by having everyone present both sides, so human’s natural instincts to question things they disagree with don’t unduly bias a grade.

    2. Isn’t interpretative charity a Rule of Conduct (3.3), and any such violation must be reported by the judge and by all lawyers to the Disciplinary Counsel? The failure by any judge of such reporting is a violation of the Canons of Judicial Conduct.

      1. Isn’t interpretative charity a Rule of Conduct (3.3),

        No. This has been yet another episode of Simple Answers to Stupid Questions.

        1. David. Your ipse dixit contradicts the Comment associated with 3.3 (c) of New York.

          Comment 4: Legal Argument
          [4] Although a lawyer is not required to make a disinterested exposition of the law, legal argument based on a knowingly false representation of law constitutes dishonesty toward the tribunal. Paragraph (a)(2) requires an advocate to disclose directly adverse and controlling legal authority that is known to the lawyer and that has not been disclosed by the opposing party. A tribunal that is fully informed on the applicable law is better able to make a fair and accurate determination of the matter before it.

          See the limits of In re Greenberg, 15 NJ 132, 137, 104, A.2d 46, 49 (1954)

          1. That’s great. But what does it have to do with your question, which was not, “Isn’t knowingly misrepresenting the law or facts a violation of Rule of Conduct 3.3?” but rather was “Isn’t interpretative charity a Rule of Conduct (3.3)”?

            1. You have to reveal the adverse point of law to the judge if not done by the other side. That is correct. Interpretive charity is required by 3.3. If you have ever failed to do so, you need to report yourself to the DC (8.3).

              1. I guess it might look that way, but the way lawyers actually apply this is as follows:

                Question 1: –Do the cases we are citing really say what we say they say?
                Yes: –Good!
                No: –Okay, either cite some other cases or put some nuance in our discussion of this case.

                Question 2: –Is there a case that is directly opposed to what we are saying?
                Yes: –Are you sure? I bet we can distinguish it. Look at all the facts in that other case and see how ours are different. Also, maybe it’s not binding authority or the bad part is dicta. You have to find something. We can’t just admit that we’re wrong.
                No: –good.

    3. bernard11, I learned it as: I cut, you choose

      1. I learned it from Taran Wanderer, where Taran decides that Lord Goryon will divide the mixed-up herds of cattle but Lord Gast will get the first pick of the halves (I may have those two backwards).

    4. Interesting until actual test cases come up.

      For example: US v Windsor.

      1. I’m not following. Can you explain what Windsor is an example of?

  4. True and helpful. Thanks.

  5. “Indeed, I now sometimes test a version of this skill on my exams, asking students to write up both sides of an argument, with the rule that their grade will be based on the quality of the worse of the two arguments.”

    Do you mean that you present a problem with one side that you think is correct and the students are graded on how well they present the opposing case?

    Or do you mean the problem is balanced in your mind, and you figure out which argument is the “worse” of the ones made by the student, and then go back and judge the quality of that?

    1. My interpretation is he presents an issue. Had students argue both sides, then he grades both arguements and the student recieves the lower of the two grades.

      1. Okay, and if everyone is getting the lower of the two grades, I assume on the ultimate grade curve it will be the same distribution in the end?

        1. It may end up the same class distribution but the individual won’t necessarily be in the same place in that distribution. Someone can argue a point the agree with very well but not have the ability to argue a point they don’t, they will end up further down in this grading system. I think there is a lot of merit to this given that as lawyers you very rarely get to decided your cases based solely on your view of the law.

        2. “Okay, and if everyone is getting the lower of the two grades, I assume on the ultimate grade curve it will be the same distribution in the end?”

          From a math POV, I don’t think so. Suppose a class is really bad at steelmanning, so all their arguments for their preferred side get an A, B, or C, but the arguments for the side they don’t prefer are total dreck, and all those get an F. The two distributions are different, even if graded on a curve.

          1. But even if the class as a whole is bad at it, won’t the levels of badness typically be in a normal distribution for a large enough class?

            1. Normal distributions are only feasible near the middle of the range. If everybody’s crowded near one end or the other, one of the tails gets clipped, and you get something that looks different.

            2. ‘typically be in a normal distribution for a large enough class’

              Maybe :-). My comment was talking about what could possibly happen, which is a different thing from what is likely to happen. I was addressing “I assume on the ultimate grade curve it will be the same distribution in the end?”, and I wouldn’t go past ‘maybe’ for that.

              As Brett points out, lots of things aren’t normal distributions. Some things are other symmetrical hump shaped distributions, but not normal[1]. And things that crowd an endpoint are rarely normal – income being the classic example. My only info on law school grades is what I read here, but my impression is that they heavily crowd into the A/B range, so perhaps not normally distributed at all.

              [1]n.b. as a statistical term of art, a ‘normal’ distribution refers to a specific symmetric hump shaped distribution. There are others – scroll to the bottom of this page for a list. If you click on ‘Logistic Distribution’ from that list it is a good example of one that looks like a normal distribution but isn’t.

              This concludes today’s edition of Overly Pedantic Discussions of Statistical Questions 🙂

  6. “The best case against the new school tax is ‘lol, who cares about the children, anyway.'”

    There, I know this technique!

    1. No, it’s that money not spent on taxes can get spent on tires…

  7. How can one truly know what the best form of opposing argument is? That is what I am having a hard time with, logically. The premise of ‘steelmanning’ is that you know with certainty the best form of the opposing argument. I am just not seeing how that is possible.

    What is a good example of steelmanning?

    1. Not the best form objectively. The best form to you. Whatever arguement that could most persuade you of the validity of the other side.

      Also, learning how to find that best arguement is the exercise.

      1. Ahhhhhhhhh…..Ok, now I see (Thank you!). That is where I was off-track. This goes to the re-grounding the argument he wrote about.

    2. I don’t think employing a technique is premised on your being perfect at it, which would seem to be the basis of your own argument.

      You use what you think is the best argument for the opposing position.

      I’ll grant you it requires a certain empathy towards positions you disagree with, which can be hard for some people.

    3. To add to what they say, a somewhat crude way is to simply ask which argument do I have the hardest time countering. There is a good chance that is going to at least be a strong opposing argument if not the strongest

      1. a somewhat crude way is to simply ask which argument do I have the hardest time countering.

        That looks like a good approach to me.

    4. So called steelmanning is what a set of red team reviewers do as one tries to craft the most formidable picture of one’s competitor in an effort to make the most convincing argument of the superiority of your team’s offer. It is done all the time in preparing major contract bids.

    5. There was a livejournal post at one point that steelmanned timecube. It was pretty good, but livejournal is apparently no longer letting you view things without logging in…

  8. Good post. I try to find the best most convincing argument for something I disagree with.

    1. If everyone did that, most internet forums would vanish.

      1. I get your point, but I find forums such as this can occasionally be useful for this very purpose, finding the best counterargument.

        1. Of course, on the Internet it might be deemed rude to ignore the actual arguments people make – and they might accuse you of strawmanning.

          And the actual arguments people make on the Internet are not necessarily the *best* arguments.

  9. I large competitive contract proposals, the capture manager is always best off trying to craft the proposal to counter the claims and offers of the strongest possible competitor, who may be an amalgam of multiple actual competitors.

  10. As someone who thinks that licensing is the n general constitutional, whether or not wise or good policy, I think this requires a bit more discussion. The line between what can and can’t be licensed doesn’t seem quite so obvious.

    A great many professions that primarily involve speech routinely require licenses. Attorneys, therapists, teachers, and more all mainly involve talk and writing. Accreditation requirements for universities to receive statw approval or federal funding impose de facto requirements on teachers and are in effect a licensing scheme. And although Reason Magazine and Conspirators routinely object to licensing things like tour guides, courts have generally permitted it.

    One possible distinction is between “pure” speech – speech that solely communicates ideas – and speech intended to achoeve a specific result. Another might be speech plus, i.e. speech plus some other specific knowledge or skill within a defined area.

    But it seems to me the distinctions are largely “know it when I see it” and it’s hard to come up with a rule with no counterexamples. For example, I think mathematicians can’t be licensed, but actuaries and engineers can be, and maybe also statisticians and other applied math people. I would think a state or Congress could require the kind of computer programmers who write software for autopilots or businesses to have licenses, but not those who write things like video games.

    I think my approach would exclude general fact-checkers with no ties to a specific profession or area of application. But I suspect a state could impose licensing requirements covering more narrowly defined, more applied professions much of whose actual work consists of fact-checking.

    1. I think you’re commenting on the wrong post.

  11. Except for the sexism of the term I really like steelmanning as described in this post, its use in the test approach that was described, and the thoughtful comparison w/ interpretive charity. The world could surely use more of this.

    The method has clear value when one is trying to get closer to truth, and also when one is trying to craft a better argument for one’s position, true or not. But it seems to me (a nonlawyer) that it would almost by definition be an endangered species in the courtroom. If I’m wrong about that, can anyone provide counterexamples?

    1. Effective lawyers certainly try to anticipate the best possible arguments that could be deployed against their positions so that they’re prepared to counter them.

      1. Lawyers in the Supreme Court broadcast have an answer for everything. That is really suspicious. It means they are just full of it, not serious or thoughtful.

  12. Interpretive charity would, if practiced, shut up many commenters here. And PolitiFact would go out of business.

    Devil’s advocate is one of my favorite rhetorical tools.

    1. The little legal training I had made me a whore. For example, I have compelling, persuasive arguments why the American legal profession is not that bad. No one has ever made them in rebutting my criticism of it. They could be new to the geniuses here. I could do it if asked. No one has been smart enough to ask. They just stupidly insult people, looking frustrated.

Please to post comments