Two of my students graduated in the span of a week. Essentially everyone but me who was on their committees was an experimentalist, with the exception of one pen-and-paper theorist.
Let me tell you this: developing computer simulations of the physical world which have real predictive power is very complicated. It requires understanding experiments, being able to understand and/or develop an appropriately detailed mathematical model, and then being able to develop and computationally implement an algorithm to numerically solve said mathematical model. It requires the skills of an old-school pen-and-paper theorist along with the skills in the computational sciences. The work is challenging and requires careful attention to detail; if you don’t know what you are doing, it will be garbage in, garbage out.
When I read experimental dissertations and attend defenses, I hear about all the details of growth, fabrication, and various characterization and measurement techniques. I consider it important to understand what those are.
Yet, whenever I present my group’s work, I have to always wave away the details of what I do because people don’t care. I have to care about various chemicals used for etching this or that, but it’s totally okay that my colleagues don’t care about the mathematical model that actually resulted in some graphs. I guess “math is hard and boring” applies to the mindsets of middle-schoolers and professors alike.
It pisses me off that my experimental colleagues cannot be bothered to try and understand the details of what my student did — at least at the student’s defense. The defense should be a time when students get to talk, at least a little bit, about the cool things they did, because that’s why they are getting a PhD.
Basically, no matter how long I collaborate with an experimentalist, they only ever care if the simulation is completely done with all the bells and whistles so that it matches perfectly with this or that, and they don’t care at all that it takes time to get to such high levels of quantitative agreement, that there are natural stages in the development of a detailed simulation, and that I cannot keep a student around for 15 years. What they want is only the information about what is yet not in the simulation and the reasons why it’s not yet perfect, rather than talking about how exactly we already got it to capture 85-90% of the physics.
Why the hell do I do this job, again? Seriously?
The other day I met a theorist who does old-school pen-and-paper theory. There was not a graph to save a life in his presentation. The dude publishes in Prestigious Society Letters far more often than I do (pedigree matters; also, PSL clings to this old-school theory and it is very hard to get much of the more modern or more applied topics past their reviewer base. I have spoken with some editors and they are all surprised why their impact factor is suffering — it’s suffering because they have not adapted to the reality of where the action is in the area that caters to the largest physics field today; journals published by a couple of other publishers, including a large society, have not had that problem and are now sporting impact factors twice or three times that of PSL. But I digress). Anyway, nobody among my experimental colleagues would have any patience whatsoever with me if I were to throw equations around with quantities in arbitrary units, only in the limits of what I can compute analytically (hint: very little), and only rarely show any actual data.
DH says that I shouldn’t care what anyone else thinks and that I do what I do because I love it. I am not so sure any more. I am kind of sick of it all.
Sick of always being second fiddle to experimentalists — more of an unpaid intern or a serf really; I sure as hell have to find my own money to do the work. Sick of working really hard and always only being asked why it isn’t all absolutely perfect yet, with anything that anyone could possibly imagine computable in a heartbeat. My former postdoc has this great saying, that he feels people think we theorists receive a giant magic box with a million buttons (presumably when we are admitted to the secret society of theorists), so that whatever anyone imagines being calculated about any physical system just requires pressing a button on the magic box, and then people get cross when they want some specific data from us and we sit on our lazy a$$es and take our sweet time pressing the button.
And then some student from the audience asks why we do this complicated theory, can’t people just buy software to do this? I almost blew my top off. No, there is nothing commercial that is even remotely like what we are developing.
But this conveys another aspect of why everyone has such a low opinion of theory and simulation — they think what I do is the same as what they do when their boss buys them some canned software and they play around with it. That’s what 99.99% people have in mind when they say “I am so cool, I can do both experiment and theory.” No, you cannot. Developing new simulation techniques and running canned software are not the same. It’s night and day.
This is all such bull$hit. Why do I even bother?
I cannot wait to go to a conference in my subfield in June.
As a person who also does computation theory that looks like it might be related to something you can do with a package, I totally related to this.
I totally empathize with your post – I have also felt from certain wet-lab colleagues that computational biology should be fast and easy. After all, “all I have to do” is analyze the data! :p
There was a theorist I knew who had a saying (paraphrasing): “Nobody believes the results obtained/presented by a theorist/modeler except the theorist/modeler themself. Everybody believes the data obtained/presented by an experimentalist except the experimentalist themself.
Maybe this is too generous, but people might just not feel like they have a path to understand what your work involves. It might just be a matter of further explaining these details. (How are they supposed to automatically know the ins and outs; how would they be expected to know the details of how what you do differs from what is commercially available?)
Of course we all (experimentalists included!) wish we could get the perfect questions for our students at PhD exams etc but the reality is that people are busy and they are going to ask questions based on their own knowledge base … in turn you can ask more theoretical questions to experimental students, right? And for your own students, bring in external readers with expertise in your area?
Dafs, there is nothing commercially available in this field. We are one of the 3-5 groups in the world really pushing the state-of-the-art of what can be predictively simulated in this field. What’s offensive/annoying is the question from some random student in the audience, “This is so complicated! Why do you bother, there has to be some commercial tool with which one can get to do this easily.” There isn’t. And this stuff is indeed hard, and it’s not just us making things unnecessarily complicated.
I guarantee I know far, far more about the details of fabrication and processing that my collaborators do than any of my experimental collaborators know about how I do what I do. It’s somehow acceptable to tune out when we present the theory work because it’s math and algorithms so it’s skippable and everyone just wants to fast forward to the results (which come from a magic box), while it’s assumed all the recipes for growth and patterning and etching and whatnot are universally riveting and should of course be shared in painstaking detail.
If I asked theory questions in other students’ defenses in as much detail as my students get asked about fabrication and processing questions, it would be a slaughter.
And for your own students, bring in external readers with expertise in your area?
That’s part of the issue, I don’t have people closely aligned with my expertise here. While some interdepartmental breadth is required, people from outside of the university are quite uncommon on defense committees.
It’s pretty amazing to me how much anti-math/anti-theory bias there is even in physics.
I remember when I asked our new “theory” hire what algorithm he uses in his simulations and his answer was “I don’t know, I think it’s [name], I just use the same code as everyone else.” I think this individual has some specific issues, and I wouldn’t assume everyone in the field is like him, but the fact that one could get away with such an attitude in academia is pretty sad.
I suspect that the reason for this is the same reason that the general public is proud to boast of being innumerate, while the same folks would be humiliated if they were illiterate. People (researchers too!) view math as this horribly complicated foreign language they have no desire to learn. I see this in my classes too (which is beyond depressing!).
It’s too bad, because as an experimentalist who wants to use calculations to help guide future research, I really want to know what the underlying mathematical machinery is doing. How else can I understand the assumptions built into the models to know where the models are particularly effective or particularly problematic? I don’t think this is all that common, though, since many of my colleagues (as you say) just try to get some calculation somewhere to match what they observed in lab rather than using modeling as additional data that can guide future work. Given the “second class” status of theory, most of those details are left out of talks and even papers unless they are targeting a theoretical audience.
I also do theory & computation, but my collaborators are in biomedicine. I am constantly struggling to keep up with the latest bench methods (and in the CRISPR world, they move fast). Based on my collaborators’ reactions to my questions, I’m not a complete fool. However, I am constantly struggling to convince them how modeling and statistics can be used both to test hypothesis and build predictive models. They seem to think that because they have experimental evidence of x and y, they understand the whole system, so what’s the need for modeling? And they don’t seem to believe that hidden states can be inferred. It is a challenge to convince them to collect the data I need, but I’m trying my hardest to show what is possible with synthetic data sets, and by drawing new conclusions from what’s already out there.
Ironically, the top biomedical journals (and I’d include Nature and Science here) are biased the other direction. You can demonstrate a large, magnificent statistical effect of something in humans, but it’s not real unless there’s some semblance of the mechanism in five to six mice. Drives me bonkers.
hmm, I see the issue now. Well, you are a top group in the world … then who cares what your local colleagues (or their students for that matter) think or ask your students! Getting through a PhD defense should just be a to-do list item for you and your student. Make a binder of every silly etching question ever asked, some boilerplate that explains what you do to generalists, etc. etc. … and then your students will have the tools they need to survive the defense … and you can spend your energy on something else.
Well, to play devil’s advocate: I think it is simply easier for theorists to understand experiments than vice versa. Anecdotally, the two best postdocs I’ve ever had (I’m an experimentalist): one was PhD in theoretical physicist, the other in applied math. They picked up the experimental stuff in months plus added their brilliant ordered thinking to produce really outstanding work.
On the other hand, I struggle to understand the theory for anything outside my sub-field. I’m still trying to understand the Physics behind the recent Nobel and it is not very far from my own field. Outside of taking a few basic field theory classes in grad school, I don’t even bother trying to understand particle theory.
So I regret that experimentalists at your place are acting like that, but it might be a defense mechanism…chances are they just aren’t as smart as you and your students.
I have to admit I am an experimentalist who struggles to understand in any detail a lot of the techniques that my theory collaborators are using. (I’m sure you’ve heard this before, but some – most? – of us feel like ignorant dopes when faced with the fact that we haven’t got nearly the mathematical background necessary to follow a given theoretical talk. If that makes you feel any better. It certainly doesn’t excuse dismissing your students’ cool innovations though!)
Do you have any thoughts on how individuals can improve in this regard? I have no real concrete ideas how to get better at understanding theoretical and computational techniques other than cracking some textbooks and review papers in my nonexistent spare time. And I have a vague understanding when X technique has a specific advantage over Y technique (sometimes) but to a large extent they tend to blur together in my mind as “yet another numeric technique” and “complicated theory stuff I don’t really understand”, so I find it hard to get a good handle.
Anyway, I ask partly because as you say, this seems like a very analogous situation — e.g. I would think that, given the social latitude, many theorists would be plenty tempted to let experimental details blur together in their minds as “yet another characterization technique” and “fiddly fabrication stuff I don’t really understand”. So I suspect that, out of necessity, you’ve probably come up with better strategies than I have for picking up a working knowledge of the specifics of what your counterparts do.
(I do recognize the irony that you’re already expected to put in more effort learning about experiments than we are with theory work, and here I am asking for more effort/education; not expecting an answer if you don’t find it an interesting topic.)
As another computational scientist specialising in developing new methods for spatio-temporal data I totally recognise this (no, commercial softwares can not do what I am doing!). So, physics, biomedicine, and in my case geographic information science, it seems the story is everywhere the same…
@another quantum mechanic: you have to put in the time — there is no substitute. There are no “strategies” or shortcuts. It helps if you are brilliant; then less time is required. But it all boils down to a question of time and effort.
Another quantum mechanic: Unlike an experimental physical system, my simulated system does not a priori exist and has no properties other than those assigned to it through the development of the simulation. Indeed, if I want to develop a realistic simulation, I have to understand as well as I possibly can the experimental reality so I would be able to properly model it, and that does involve processing and fabrication details, because they have very specific and important effects on the resulting structures. I send my theory students to take some processing classes we offer — it’s not the same as being an experimentalist, but it really helps build an understanding of the process and the appreciation of the importance and difficulty of the common steps. In contrast, my experimental colleagues do not send their students to take the higher-level, heavily theoretical/computational courses I offer. I would be delighted if my experimental colleagues asked about the nitty-gritty of the simulation; being that I am a giant nerd, I could talk your ear off about the math, the validity range of different models, the techniques in which the spatio-temporal partial differential equations can be solved (e.g., there are broad classes of approaches) and why we choose the techniques we do, all the way to how certain pieces of code are structured… And some colleagues do ask on occasion, at which point I am happy to talk, but it’s not often.
The thing is, my students take processing courses, and theory/numerics courses in adjacent areas and in different departments, and I also encourage them to take courses in whatever strikes their fancy for as long as they are in grad school; many faculty don’t want their students taking any more courses than absolutely necessary and want them spending all their time in the lab instead. I think this is unfortunate, because it’s really great if you can take a course on a topic, especially if the course is well structured, with homework and projects; that’s the fastest and most focused way to learn something that you need to know but do not plan on being an expert in.
We as faculty would probably be happier overall if we allowed ourselves to be students again, taking a class every so often, really immersing ourselves in new material and learning from another sage for a little while.
When I was a young faculty member, I took about one course a year. That dropped off a little after a couple of decades, but I’m once again taking a course a year—there is still so much to learn!
Sometimes I dream about asking my collaborators if I could shadow their grad students. I don’t understand exactly what labor goes into their plots sometimes, exactly what the data look like on the bench, and that sort of thing. Like xyka said, I wouldn’t mind if a collaborator asked me to explain in 10/30/360/… min how my models are built.
If I get tenure, one of my dreams is to resume independent math study (maybe through coursework too) and brush up my programming skills. I’m a theorist, but I still feel like I don’t have all the tools I need. I don’t feel like I have time to learn these things on the tenure track, and I’m not sure it would be good for me to sit in on classes that I’m sure some of my colleagues think I could, in theory, be teaching.
In almost the same boat as you. Some of my experimental biology colleagues think that it is just a program that you run and will get all the wonderful answers from their clumsy data out of thin air.
There are other though who know about the complexities of such algorithms and simulations.
I find this very interesting and eye opening. As a postdoc in experimental physics, my experience sort of mirrors what you describe from the other direction. It often seems to me that theorists giving talks don’t even bother trying to make them understandable to experimentalists. I hesitate going to department seminars given by theorists, since I know that most likely I won’t understand anything beyond the first slide. And to me it always seems that people are getting deep into the nitty-gritty of their calculations (though maybe more the equations, and not so much the numerics part). Whereas my impression of experimental talks was that they do make more effort to be more broadly understood, and that getting into the very details of the experiments is actually not that common. And indeed I think it’s much more common to hear theorists ask questions and be involved in the experimentalists talks than vice versa. Maybe it’s just that theorists are smarter (or at least better educated), as someone above me suggested.
Reading your blog made me wonder if I’m just seeing everything from a very distorted view – maybe experimental talks are not generally easier to understand, but they’re just easier for me, since I’m an experimentalist? Should I try to understand more deeply the theoretical work done by my collaborators? (I usually try to understand the physics behind it, but treat the actual calculation/numerics part as irrelevant-for-me details. I don’t tell you how I tune my laser, you don’t need to tell me how you tune your fitting parameters). I’m not sure what’s the answer, but it was definitely thought-provoking.
P.S. Wow, that comment by the student is just a$$holish, no matter how you turn it. Did he seriously ask that at a defense?
Just so you don’t think that only theorists receive all the crap, here’s a comment me and my phd-advisor got on several occasions in regards to my phd work: “Why do you even need to measure this (physical process that was never quantified before since it was previously experimentally impossible)? The Russian theorists have already calculated this in the 60s”