Concepts named after Nazis etc

Stream: mathlib4

[Yaël Dillies]

Andrew Yang said:

For the record, as the original author of KaehlerDifferential in mathlib, I have no preference towards Kaehler or Kahler (or maybe Kahler over Kaehler) and this was kind of a random choice.
In fact I called it derivation_module initially and some reviewer suggested the current name.

Dhruv Ranganathan (who taught me algebraic geometry) prefers to drop Kähler's name because of other things he's done

[Johan Commelin]

Yaël Dillies said:

Dhruv Ranganathan (who taught me algebraic geometry) prefers to drop Kähler's name because of other things he's done

For the record, I am very uncomfortable with this $(1+\varepsilon)$-dimensional treatment of historical figures.

[Johan Commelin]

If we are able to recognize that someone did horrible things, let's also recognize that nevertheless they did really brilliant maths. By erasing the complexity, we are setting ourselves up to make similar mistakes in the future (or worse).

[Johan Commelin]

If someone got a statue for the fact that they were a slave trader or a Nazi, then by all means go and take down that statue. But if they proved a great theorem, then they proved a great theorem.

[Notification Bot]

5 messages were moved here from #mathlib4 > Holder vs Hoelder by Yury G. Kudryashov.

[Notification Bot]

A message was moved from this topic to #mathlib4 > Holder vs Hoelder by Yury G. Kudryashov.

[Eric Wieser]

There was a stained glass window removed at the college @Yaël Dillies and I went to, for similar reasons; though here the mathematic innovation was in some sense motivated by doing the horrible things

[Henrik Böving]

Yaël Dillies said:

Andrew Yang said:

For the record, as the original author of KaehlerDifferential in mathlib, I have no preference towards Kaehler or Kahler (or maybe Kahler over Kaehler) and this was kind of a random choice.
In fact I called it derivation_module initially and some reviewer suggested the current name.

Dhruv Ranganathan (who taught me algebraic geometry) prefers to drop Kähler's name because of other things he's done

Ah nevermind the english wikipedia article has more info than the german one

[Yury G. Kudryashov]

On the one hand, I understand people who don't want to have concepts named after Nazis. In fact, I think that relatively mild consequences for Nazi scientists in 1940s contribute to willingness of today's scientists to, e.g., develop bombs for Russia. On the other hand, it's a very slippery path. Where would you draw the line?

[Johan Commelin]

I'm very hesitant to draw lines. I don't say that's a good thing... but speaking of lines, I'm reminded of

The line separating good and evil passes not through states, nor between classes, nor between political parties either -- but right through every human heart -- and through all human hearts. This line shifts. Inside us, it oscillates with the years. And even within hearts overwhelmed by evil, one small bridgehead of good is retained.
— Aleksandr Solzhenitsyn

I don't think this quote is a triviality. And Solzhenitsyn knew that all too well.
There are dark chambers in my soul that I don't understand. That I might not even know exist. That I would be very scared to look into. Occasionally I'm confronted with this darkness a little bit. But I don't see the full force.

This doesn't mean that I think we should ignore the evil things that people do. Or just apologize them away. We should be very much aware of them. And be reminded of them.

Ok, I'm rambling... I just think this matter is very complex.

[Frédéric Dupuis]

On a more practical note, I don't like the idea of having to debate the morality of mathematicians' lives on this Zulip whenever we need to name a def.

[Edward van de Meent]

Frédéric Dupuis said:

On a more practical note, I don't like the idea of having to debate the morality of mathematicians' lives on this Zulip whenever we need to name a def.

arguably, this means we should never name concepts after people

[Johan Commelin]

I think it means that we should feel free to name concepts after people.

[Eric Wieser]

More of docs#EuclideanGeometry.dist_sq_eq_dist_sq_add_dist_sq_iff_angle_eq_pi_div_two it is then!

[Johan Commelin]

Of course, if thinks are misattributed, that's a different story. And of course this happens a lot in maths. It's a folklore result by P. Random that every theorem in maths is attributed to the wrong person.
So I'm not advocating freely attributing concepts to anybody at all.

[Johan Commelin]

It would be very strange to stop attributing WWII to Hitler, because he was a Nazi. Of course you can claim that I'm putting up a strawman.
But my point really is: if person X proved a great theorem, then attribute it to person X. And if X is/was also a terrible person outside (or inside!) the maths department, then that is/was a terrible thing, and it makes person X a bad person. But they still proved a great theorem.
So add a historical note if you want. To illustrate that people are complex.

[Arthur Paulino]

I once lied to my father (sorry dad) and I've got a few things merged into mathlib4. The repo is now cursed because I did a terrible thing

[Antoine Chambert-Loir]

Johan Commelin said:

Of course, if thinks are misattributed, that's a different story. And of course this happens a lot in maths. It's a folklore result by P. Random that every theorem in maths is attributed to the wrong person.

I think it's a theorem of André Weil.

[Antoine Chambert-Loir]

(ref : page 7 of https://www.ams.org/notices/199511/forum.pdf)

[Antoine Chambert-Loir]

and the example given by Eric is certainly a kind of over-attribution (but while I believe there is some interesting history in Serge Lang's paper, I do not advocate for erasing Pythagoras). (To derail the discussion even farther, Michael Harris has an interesting take on the question raised by Lang, https://arxiv.org/abs/2003.08242)

[Antoine Chambert-Loir]

and about drawing lines, I believe that everybody draws lines, at least internally, even if those lines move along time, as reminded us in Solzhenitsyn's quote, but I would not like to spend a minute with people whose line could approach the nazis's one.

[Yury G. Kudryashov]

What I want to say is that we have a large community with very different opinions on many topics, including this one. The approaches "if a theorem carries a name in a math paper or a textbook published by a major publisher, then this name can go to Mathlib" and "don't use names for defs/theorems" are easy to formulate and enforce. Anything in between will be much harder to formulate and/or enforce. E.g., what about people who collaborated with NKVD/KGB? Most of the archives are still sealed; quite a few USSR mathematicians were arrested&killed after a denunciation (anonymous or not) from a colleague. Also, I guess that this community is large enough to have members supporting opposite sides of ongoing wars (please don't start discussing them now - this is a very good way to spike the number of pairs of people who refuse to talk and review each other's PRs).

[Violeta Hernández]

From a moral standpoint I'm more than glad to cleanse math from the names of awful people. I agree with the sentiment of others here though - I think that's a more divisive move than to simply acknowledge that yeah, these great mathematicians were not in fact enviable human beings.

[Violeta Hernández]

(deleted)

[Violeta Hernández]

I'm failing to remember the exact names, but I'm sure there's way more concepts not yet in Mathlib named after bad Germans, and I'd rather not have to see a bunch of drama unfold each time one of those gets added. Naming is not endorsement.

[David Michael Roberts]

Otoh I think renaming the Nevanlinna prize was a good idea, because the choice of famous person to grace an award name is wholly arbitrary.

[Johan Commelin]

Yep, I can definitely get behind that. But I would be uncomfy about renaming Nevanlinna theory. (Tbh, I haven't looked far into the details of this historical example.)

[Michael Rothgang]

Violeta Hernández said:

I'm failing to remember the exact names, but I'm sure there's way more concepts not yet in Mathlib named after bad Germans, and I'd rather not have to see a bunch of drama unfold each time one of those gets added. Naming is not endorsement.

Ludwig Bieberbach and Oswald Teichmüller come to mind. Searching for Deutsche Mathematik (wikipedia for the lazy ones, such as me) finds more.

And I think the Teichmüller space of Riemann surfaces (however it is named) ought to be added to mathlib.

[Dominic Steinitz]

Michael Rothgang said:

Violeta Hernández said:

I'm failing to remember the exact names, but I'm sure there's way more concepts not yet in Mathlib named after bad Germans, and I'd rather not have to see a bunch of drama unfold each time one of those gets added. Naming is not endorsement.

Ludwig Bieberbach and Oswald Teichmüller come to mind. Searching for Deutsche Mathematik (wikipedia for the lazy ones, such as me) finds more.

And I think the Teichmüller space of Riemann surfaces (however it is named) ought to be added to mathlib.

There's much more detail on this in Mathematik in Österreich und die NS-Zeit 176 Kurzbiographien Peter Michor (nur zufällig entdeckt) and Mathematicians under the Nazis Sanford L. Segal.

Teichmüller was indeed a committed Nazi and was a leader of the student protests at Göttingen against Landau and Courant (that German students should not be taught by jewish professors). He voluntarily signed up to fight on the Eastern Front and was involved in the Kursk offensive. He died on 11.09.1943 somewhere to the west of Kharkiv. I have a theory that Ulla Hahn based a character on him in her novel Fuzzy Pictures but that's really by the by.

I have wandered way off topic. My apologies.

[Jon Eugster]

After a discussion with a flatmate of mine - not mathematician but rather working with german literature - I'd like to input a different perspective.

I think I'm all in favour for reducing the apperance of "Kaehler" to a single sentense in the file's module docstring (or def docstring).

I'm not suggesting that there should be explicit effort in mathlib by anybody specifically, but personally I think it would be desirable if PRs in that direction would be welcomed and accepted.

There seems to be no harm in choosing some more neutral name (provided search functionality still works) but it feels weird to insist on keeping a name just because "that's what people before us have been doing".

[Ruben Van de Velde]

Well, mutual intelligibility with the rest of the mathematical literature is useful, though we also lean towards descriptive names in general. (Though then the docstring should also have the variant spellings for searching)

[Yury G. Kudryashov]

@Jon Eugster What rule/procedures do you suggest for other names?

[mars0i]

My opinion here shouldn't carry any weight--I'm just a beginner with Lean--but I'll offer a couple of points that others can consider, if they wish. I want to start with non-mathematical scholars, because if my idea has merit at all, I think it works better with names of mathematical entities.

I'm one of those who has sympathy to the idea of restricting honors to people who've done bad things, while also thinking that there are wonderful things worth celebrating by some of those people. Lately I've wondered whether it would be possible to celebrate a person's beneficial research, and name it as theirs, without celebrating the person as a whole. The idea would be that rather than naming something "David Hume Tower", name it "Hume's Epistemology Tower". I know, that doesn't roll off the tongue. And some arguments for renaming might still apply.

In math I feel that it's easier to think of a theorem as something that goes beyond any person who proved it (especially for those who have Platonistic tendencies to view theorems as discovered rather than created). A theorem is a very specific, relatively discrete piece of work, even if proving it or understanding it requires knowing and thinking through a lot of other things.

So whether or not we should start naming things "Hume's Epistemology Tower" or "Ronald A. Fisher's Genetical Theory of Natural Selection Library", perhaps it's reasonable to think of a named theorem or other mathematical object as celebrating some specific work that contributed to others' understanding of the mathematical entity, rather than celebrating the person as a whole. In which case maybe there's less reason to remove well-known names for theorems, etc. (Putting aside Stigler's law of eponymy, alluded to above).

[David Michael Roberts]

The font of all wisdom, ChatGPT 4o, suggested "Hermitian-symplectic manifold" as an alternative descriptive name for "Kähler manifold" that doesn't refer to the mathematician (I didn't prompt it with info about the Nazi association). The irony of course with it using Hermite's name....
Such descriptive names are ideally clear to people who know what the object actually is, and for people who don't know what a Kähler manifold is, that's instantly suggestive. Having the name in the docstring for searchability is I think a bit of a non-negotiability. Saying that mathlib uses "Hermitian-symplectic" (unless there is anything existing that clashes with that) as a convention doesn't strike me as much worse than a number of other neologisms that are more opaque to the average mathematician/new user.

[Yury G. Kudryashov]

More generally, do you suggest to make decisions on a case by case basis, or have some criteria for "people whose names we don't want to see used many times in Mathlib"?

[Yury G. Kudryashov]

E.g., lots of Soviet mathematicians collaborated with NKVD (later KGB) one way or another, including public and anonymous denonciations that in many cases led to arrests and/or deaths. See https://en.wikipedia.org/wiki/Nikolai_Luzin#Luzin_affair_of_1936 for one of many examples, most of them less public, the KGB archives in Russia are still sealed. Also, quite a few mathematicians actively participated in denying jewish students admission to the math department of the Moscow State University (one of the entrance exams was an oral exam, and they gave jewish students super hard problems with short solutions, with difficulty close to IMO problems of that time).

[Yury G. Kudryashov]

Note: I'm trying to say that the question is hard, and IMHO we shouldn't make changes to the library before we agree on some more-or-less actionable policy (which may include voting or some other procedure).

[David Michael Roberts]

Yes, the slope is indeed slippery the further one goes. I just think the above example is instructive as to how one might style replacement names that are useful and informative, divorced from the historical baggage of the existing name (here's another one: Kähler differentials -> universal differentials, again with the caveat that this was suggested by an LLM)

[Johan Commelin]

I think what these people did is too terrible too have their names purged. What is the good of "cleansing maths"? If we don't want to credit Kähler for his work, then we should also stop using it.

[Johan Commelin]

I do not want operate under the assumption that I am on the "good side" of history.
(And once again, I am not saying that I think maybe the Nazi's were right. I think they were doing terrible things. What they did in the real world is very wrong. But apparently that didn't mean their maths was wrong.)
I really think that humanity is incredibly complex.

So I think I would much rather be confronted with the fact that some beautiful maths was produced by very evil people.
Scientist ≠ saint. That's a fact I want all scientists and the general public to be very aware of.

[Joël Riou]

As some names are very much used in the mathematical literature, it seems difficult to think seriously about erasing the names of the authors of some contributions to mathematics (e.g. "Kähler manifold" seems very standard terminology). However, I do not feel that we absolutely need to make their names more visible than necessary (e.g. in Hartshorne's book, "Kähler differentials" and "relative differentials" both appear, while it seems there is only "1-différentielles" in the EGA). The addition of a scholarly reference to biographical data and a note about the naming choice in the docstring of the file should be welcome.

[Jon Eugster]

Yury G. Kudryashov said:

Jon Eugster What rule/procedures do you suggest for other names?

I think my suggestion above was a bit too vague and too strongly formulated (especially the "except a single occurence" sentense)

Concretely, I'm suggesting to allow such naming choices to occur if there is a reasonable alternative that's used in literature (not necessarily by the majority) or if there is a nice descriptive naming which "is reasonable". And they should be accompanied by a note in the module docstring (I think for example ## Implementation Details is a header that's used often and would be a good place) about explicit naming choices made, as well as some reasoning behind it.

I do think this is very similar to what @Joël Riou wrote above.

As was pointed out very validly, a "Kähler manifold" would not fall into this category, but "Kähler differentials" might as well.

[Shreyas Srinivas]

How about a "historical notes" section in the docs with citation-like links and an explanation of why the name is kept in mathlib.

[Yury G. Kudryashov]

@Jon Eugster This answers the question "what to do if we decide to reduce the number of times a name is mentioned in Mathlib", not "how do you suggest to decide which names do we want to ban, if any?"

[Yury G. Kudryashov]

E.g., @Johan Commelin suggests a consistent answer: "we don't want to ban any names".

[Kevin Buzzard]

I strongly believe that we want to stick to the notation in the human literature because we want to make mathlib as comprehensible as we can to humans.