Category Archives: Life as a Physicist

Academia Has Changed Less Than You’d Think

I recently read a biography of James Franck. Many of you won’t recognize the name, but physicists might remember the Franck-Hertz experiment from a lab class. Franck and Hertz performed a decisive test of Bohr’s model of the atom, ushering in the quantum age and receiving the 1925 Nobel Prize. After fleeing Germany when Hitler took power, Franck worked on the Manhattan project and co-authored the Franck Report urging the US not to use nuclear bombs on Japan. He settled at the University of Chicago, which named an institute after him.*

You can find all that on his Wikipedia page. The page also mentions his marriage later in life to Hertha Sponer. Her Wikipedia page talks about her work in spectroscopy, about how she was among the first women to receive a PhD in Germany and the first on the physics faculty at Duke University, and that she remained a professor there until 1966, when she was 70.

Neither Wikipedia page talks about two-body problems, or teaching loads, or pensions.

That’s why I was surprised when the biography covered Franck’s later life. Until Franck died, he and Sponer would travel back and forth, he visiting her at Duke and she visiting him in Chicago. According to the biography, this wasn’t exactly by choice: they both would have preferred to live together in the same city. Somehow though, despite his Nobel Prize and her scientific accomplishments, they never could. The biography talks about how the university kept her teaching class after class, so she struggled to find time for research. It talks about what happened as the couple got older, as their health made it harder and harder to travel back and forth, and they realized that without access to their German pensions they would not be able to support themselves in retirement. The biography gives the impression that Sponer taught till 70 not out of dedication but because she had no alternative.

When we think about the heroes of the past, we imagine them battling foes with historic weight: sexism, antisemitism, Nazi-ism. We don’t hear about their more everyday battles, with academic two-body problems and stingy universities. From this, we can get the impression that the dysfunctions of modern academia are new. But while the problems have grown, we aren’t the first academics with underpaid, overworked teaching faculty, nor the first to struggle to live where we want and love who we want. These are struggles academics have faced for a long, long time.

*Full disclosure: Franck was also my great-great-grandfather, hence I may find his story more interesting than most.

When to Trust the Contrarians

One of my colleagues at the NBI had an unusual experience: one of his papers took a full year to get through peer review. This happens often in math, where reviewers will diligently check proofs for errors, but it’s quite rare in physics: usually the path from writing to publication is much shorter. Then again, the delays shouldn’t have been too surprising for him, given what he was arguing.

My colleague Mohamed Rameez, along with Jacques Colin, Roya Mohayaee, and Subir Sarkar, wants to argue against one of the most famous astronomical discoveries of the last few decades: that the expansion of our universe is accelerating, and thus that an unknown “dark energy” fills the universe. They argue that one of the key pieces of evidence used to prove acceleration is mistaken: that a large region of the universe around us is in fact “flowing” in one direction, and that tricked astronomers into thinking its expansion was accelerating. You might remember a paper making a related argument back in 2016. I didn’t like the media reaction to that paper, and my post triggered a response by the authors, one of whom (Sarkar) is on this paper as well.

I’m not an astronomer or an astrophysicist. I’m not qualified to comment on their argument, and I won’t. I’d still like to know whether they’re right, though. And that means figuring out which experts to trust.

Pick anything we know in physics, and you’ll find at least one person who disagrees. I don’t mean a crackpot, though they exist too. I mean an actual expert who is convinced the rest of the field is wrong. A contrarian, if you will.

I used to be very unsympathetic to these people. I was convinced that the big results of a field are rarely wrong, because of how much is built off of them. I thought that even if a field was using dodgy methods or sloppy reasoning, the big results are used in so many different situations that if they were wrong they would have to be noticed. I’d argue that if you want to overturn one of these big claims you have to disprove not just the result itself, but every other success the field has ever made.

I still believe that, somewhat. But there are a lot of contrarians here at the Niels Bohr Institute. And I’ve started to appreciate what drives them.

The thing is, no scientific result is ever as clean as it ought to be. Everything we do is jury-rigged. We’re almost never experts in everything we’re trying to do, so we often don’t know the best method. Instead, we approximate and guess, we find rough shortcuts and don’t check if they make sense. This can take us far sometimes, sure…but it can also backfire spectacularly.

The contrarians I’ve known got their inspiration from one of those backfires. They saw a result, a respected mainstream result, and they found a glaring screw-up. Maybe it was an approximation that didn’t make any sense, or a statistical measure that was totally inappropriate. Whatever it was, it got them to dig deeper, and suddenly they saw screw-ups all over the place. When they pointed out these problems, at best the people they accused didn’t understand. At worst they got offended. Instead of cooperation, the contrarians are told they can’t possibly know what they’re talking about, and ignored. Eventually, they conclude the entire sub-field is broken.

Are they right?

Not always. They can’t be, for every claim you can find a contrarian, believing them all would be a contradiction.

But sometimes?

Often, they’re right about the screw-ups. They’re right that there’s a cleaner, more proper way to do that calculation, a statistical measure more suited to the problem. And often, doing things right raises subtleties, means that the big important result everyone believed looks a bit less impressive.

Still, that’s not the same as ruling out the result entirely. And despite all the screw-ups, the main result is still often correct. Often, it’s justified not by the original, screwed-up argument, but by newer evidence from a different direction. Often, the sub-field has grown to a point that the original screwed-up argument doesn’t really matter anymore.

Often, but again, not always.

I still don’t know whether to trust the contrarians. I still lean towards expecting fields to sort themselves out, to thinking that error alone can’t sustain long-term research. But I’m keeping a more open mind now. I’m waiting to see how far the contrarians go.

Knowing When to Hold/Fold ‘Em in Science

The things one learns from Wikipedia. For example, today I learned that the country song “The Gambler” was selected for preservation by the US Library of Congress as being “culturally, historically, or artistically significant.”

You’ve got to know when to hold ’em, know when to fold ’em,

Know when to walk away, know when to run.

Knowing when to “hold ’em” or “fold ’em” is important in life in general, but it’s particularly important in science.

And not just on poker night

As scientists, we’re often trying to do something no-one else has done before. That’s exciting, but it’s risky too: sometimes whatever we’re trying simply doesn’t work. In those situations, it’s important to recognize when we aren’t making progress, and change tactics. The trick is, we can’t give up too early either: science is genuinely hard, and sometimes when we feel stuck we’re actually close to the finish line. Knowing which is which, when to “hold” and when to “fold”, is an essential skill, and a hard one to learn.

Sometimes, we can figure this out mathematically. Computational complexity theory classifies calculations by how difficult they are, including how long they take. If you can estimate how much time you should take to do a calculation, you can decide whether you’ll finish it in a reasonable amount of time. If you just want a rough guess, you can do a simpler version of the calculation, and see how long that takes, then estimate how much longer the full one will. If you figure out you’re doomed, then it’s time to switch to a more efficient algorithm, or a different question entirely.

Sometimes, we don’t just have to consider time, but money as well. If you’re doing an experiment, you have to estimate how much the equipment will cost, and how much it will cost to run it. Experimenters get pretty good at estimating these things, but they still screw up sometimes and run over budget. Occasionally this is fine: LIGO didn’t detect anything in its first eight-year run, but they upgraded the machines and tried again, and won a Nobel prize. Other times it’s a disaster, and money keeps being funneled into a project that never works. Telling the difference is crucial, and it’s something we as a community are still not so good at.

Sometimes we just have to follow our instincts. This is dangerous, because we have a bias (the “sunk cost fallacy”) to stick with something if we’ve already spent a lot of time or money on it. To counteract that, it’s good to cultivate a bias in the opposite direction, which you might call “scientific impatience”. Getting frustrated with slow progress may not seem productive, but it keeps you motivated to search for a better way. Experienced scientists get used to how long certain types of project take. Too little progress, and they look for another option. This can fail, killing a project that was going to succeed, but it can also prevent over-investment in a failing idea. Only a mix of instincts keeps the field moving.

In the end, science is a gamble. Like the song, we have to know when to hold ’em and fold ’em, when to walk away, and when to run an idea as far as it will go. Sometimes it works, and sometimes it doesn’t. That’s science.

In Life and in Science, Test

Think of a therapist, and you might picture a pipe-smoking Freudian, interrogating you about repressed feelings. These days, you’re more likely to meet a more modern form of therapy, like cognitive behavioral therapy (or CBT for short). CBT focuses on correcting distorted thoughts and maladaptive behaviors: basically, helping you reason through your problems. It’s supposed to be one of the types of therapy that has the most actual scientific evidence behind it.

What impresses me about CBT isn’t just the scientific evidence for it, but the way it tries to teach something like a scientific worldview. If you’re depressed or anxious, a common problem is obsessive thoughts about what others think of you. Maybe you worry that everyone is just putting up with you out of pity, or that you’re hopelessly behind your peers. For many scientists, these are familiar worries.

The standard CBT advice for these worries is as obvious as it is scary: if you worry what others think of you, ask!

This is, at its heart, a very scientific thing to do. If you’re curious about something, and you can test it, just test it! Of course, there are risks to doing this, both in your personal life and in your science, but typical CBT advice applies surprisingly well to both.

If you constantly ask your friends what they think about you, you end up annoying them. Similarly, if you perform the same experiment over and over, you can keep going until you get the result you want. In both cases, the solution is to commit to trusting your initial results: just like scientists pre-registering a study, if you ask your friends what they think you need to trust them and not second-guess what they say. If they say they’re happy with you, trust that. If they criticize, take their criticism seriously and see if you can improve.

Even then, you may be tempted to come up with reasons why you can’t trust what your friends say. You’ll come up with reasons why they might be forced to be polite, while they secretly still hate you. Similarly, as a scientist you can always come up with theories that get around the evidence: no matter what you observe, a complicated enough chain of logic can make it consistent with anything you want. In both cases, the solution is a dose of Occam’s Razor: don’t fixate on an extremely complicated explanation when a simpler one already fits. If your friends say they like you, they probably do.

In Defense of the Streetlight

If you read physics blogs, you’ve probably heard this joke before:

A policeman sees a drunk man searching for something under a streetlight and asks what the drunk has lost. He says he lost his keys and they both look under the streetlight together. After a few minutes the policeman asks if he is sure he lost them here, and the drunk replies, no, and that he lost them in the park. The policeman asks why he is searching here, and the drunk replies, “this is where the light is”.

The drunk’s line of thinking has a name, the streetlight effect, and while it may seem ridiculous it’s a common error, even among experts. When it gets too tough to research something, scientists will often be tempted by an easier problem even if it has little to do with the original question. After all, it’s “where the light is”.

Physicists get accused of this all the time. Dark matter could be completely undetectable on Earth, but physicists still build experiments to search for it. Our universe appears to be curved one way, but string theory makes it much easier to study universes curved the other way, so physicists write a lot of nice proofs about a universe we don’t actually inhabit. In my own field, we spend most of our time studying a very nice theory that we know can’t describe the real world.

I’m not going to defend this behavior in general. There are real cases where scientists trick themselves into thinking they can solve an easy problem when they need to solve a hard one. But there is a crucial difference between scientists and drunkards looking for their keys, one that makes this behavior a lot more reasonable: scientists build technology.

As scientists, we can’t just grope around in the dark for our keys. The spaces we’re searching, from the space of all theories of gravity to actual outer space, are much too vast to search randomly. We need new ideas, new mathematics or new equipment, to do the search properly. If we were the drunkard of the story, we’d need to invent night-vision goggles.

Is the light better here, or is it just me?

Suppose you wanted to design new night-vision goggles, to search for your keys in the park. You could try to build them in the dark, but you wouldn’t be able to see what you were doing: you’d lose pieces, miss screws, and break lenses. Much better to build the goggles under that convenient streetlight.

Of course, if you build and test your prototype goggles under the streetlight, you risk that they aren’t good enough for the dark. You’ll have calibrated them in an unrealistic case. In all likelihood, you’ll have to go back and fix your goggles, tweaking them as you go, and you’ll run into the same problem: you can’t see what you’re doing in the dark.

At that point, though, you have an advantage: you now know how to build night-vision goggles. You’ve practiced building goggles in the light, and now even if the goggles aren’t good enough, you remember how you put them together. You can tweak the process, modify your goggles, and make something good enough to find your keys. You’re good enough at making goggles that you can modify them now, even in the dark.

Sometimes scientists really are like the drunk, searching under the most convenient streetlight. Sometimes, though, scientists are working where the light is for a reason. Instead of wasting their time lost in the dark, they’re building new technology and practicing new methods, getting better and better at searching until, when they’re ready, they can go back and find their keys. Sometimes, the streetlight is worth it.

“X Meets Y” Conferences

Most conferences focus on a specific sub-field. If you call a conference “Strings” or “Amplitudes”, people know what to expect. Likewise if you focus on something more specific, say Elliptic Integrals. But what if your conference is named after two sub-fields?

These conferences, with names like “QCD Meets Gravity” and “Scattering Amplitudes and the Conformal Bootstrap”, try to connect two different sub-fields together. I’ll call them “X Meets Y” conferences.

The most successful “X Meets Y” conferences involve two sub-fields that have been working together for quite some time. At that point, you don’t just have “X” researchers and “Y” researchers, but “X and Y” researchers, people who work on the connection between both topics. These people can glue a conference together, showing the separate “X” and “Y” researchers what “X and Y” research looks like. At a conference like that speakers have a clear idea of what to talk about: the “X” researchers know how to talk to the “Y” researchers, and vice versa, and the organizers can invite speakers who they know can talk to both groups.

If the sub-fields have less history of collaboration, “X Meets Y” conferences become trickier. You need at least a few “X and Y” researchers (or at least aspiring “X and Y” researchers) to guide the way. Even if most of the “X” researchers don’t know how to talk to “Y” researchers, the “X and Y” researchers can give suggestions, telling “X” which topics would be most interesting to “Y” and vice versa. With that kind of guidance, “X Meets Y” conferences can inspire new directions of research, opening one field up to the tools of another.

The biggest risk in an “X Meets Y” conference, that becomes more likely the fewer “X and Y” researchers there are, is that everyone just gives their usual talks. The “X” researchers talk about their “X”, and the “Y” researchers talk about their “Y”, and both groups nod politely and go home with no new ideas whatsoever. Scientists are fundamentally lazy creatures. If we already have a talk written, we’re tempted to use it, even if it doesn’t quite fit the occasion. Counteracting that is a tough job, and one that isn’t always feasible.

“X Meets Y” conferences can be very productive, the beginning of new interdisciplinary ideas. But they’re certainly hard to get right. Overall, they’re one of the trickier parts of the social side of science.

At Aspen

I’m at the Aspen Center for Physics this week, for a workshop on Scattering Amplitudes and the Conformal Bootstrap.

A place even greener than its ubiquitous compost bins

Aspen is part of a long and illustrious tradition of physics conference sites located next to ski resorts. It’s ten years younger than its closest European counterpart Les Houches School of Physics, but if anything its traditions are stricter: all blackboard talks, and a minimum two-week visit. Instead of the summer schools of Les Houches, Aspen’s goal is to inspire collaboration: to get physicists to spend time working and hiking around each other until inspiration strikes.

This workshop is a meeting between two communities: people who study the Conformal Bootstrap (nice popular description here) and my own field of Scattering Amplitudes. The Conformal Boostrap is one of our closest sister-fields, so there may be a lot of potential for collaboration. This week’s talks have been amplitudes-focused, I’m looking forward to the talks next week that will highlight connections between the two fields.