Let’s get this out of the way: In 2012, two independent teams of particle physicists discovered the Higgs boson.
Why bother with the obvious? Earlier this week, Ben Recht, a computer scientist at the University of California, Berkeley wrote an article “The Higgs Discovery Did Not Take Place,” a self-consciously contrarian yet serious polemic. In it, he basically argues that the discovery as we think of it, a revelation about the way our universe works, did not happen. After some initial praise, there’s been a fair amount of pushback. Kyle Cranmer (an experimental physicist directly involved with the discovery) has a very nice rebuttal worth reading in full. I had a few thoughts as well, so I took a stab at fleshing them out.
Recht’s argument, after a breezy introduction to quantum field theory, goes something like this:
- there are lots of uncertainties in a very complicated theory
- there are lots of uncertainties in very complicated collider data
(A brief interlude to say that so far, this is fine. From here, things go off the rails.)
- physicists searching for the Higgs engaged in questionable research practices and dubious statistics
- no one understands the entire experiment
- this means physicists cannot objectively probe reality
- therefore the discovery of Higgs was a matter of social convention
Recht writes:
The CERN collaboration establishes parliamentary rules to decide upon scientific truth. Reality is validated by majority vote. Physicists love to talk a lot about how they are probing the very nature of the universe, but they do this by a lot of boring committee meetings.
Where to begin. First, there is no ‘CERN collaboration’ nor are there ‘parliamentary rules’ by which to establish scientific truth. CERN is the European Organization for Nuclear Research. There are multiple experiments which have detectors at CERN’s Large Hadron Collider, two of which, ATLAS and CMS, discovered the Higgs boson. There are thousands of physicists on these experiments (most of whom are based at labs and universities elsewhere around the world). Accordingly, there are decisions to be made—about funding, about logistics, about whether to publish an analysis.
Boring committee meetings are part and parcel of any large-scale scientific effort, but this detail seems to come as a surprise to Recht, whose stance is that bureaucracy is in contradiction with the ability to actually discover anything fundamental about the nature of the universe.
In the late ‘70s, as postmodernism began to hit its stride, sociologists of scientists asked: “How socially constructed is science?” Works like Laboratory Life (which I have not read), and Constructing Quarks (some of which I have read) did more than just cast doubt on the purported objectivity of science—in some cases, sociologists rejected scientists’ ability to ascertain facts about reality.
This was a bit an overreach and made a lot of people quite upset.
It was, however, a useful corrective to the myth that sciences are purely objective. There is no doubt that even the most abstract and fundamental searches in physics are affected by these human factors. Kent Staley has a very nice historical account of this in The Evidence for the Top Quark. He traces the discussions that led CDF to a particular detector design, the implementation of novel silicon vertex tech, internal debates about which tagging algorithm to use, and the raucous disagreements about the interpretation of the data. About the “struggle within CDF as physicists developed strategies for finding the top quark”, Staley writes:
Political intrigues and social forces featured prominently, but not simply by making some approaches look convincing and others not convincing, as a strictly “social constructivist” reading might suggest
Staley’s careful history shows the incompleteness of a simple narrative about the discovery for the top quark. The path to say “We establish the existence of the top quark” was fraught with human subjectivity and strongly shaped by social forces. And yet it was a discovery all the same: at the end of the day, physicists had wrested forth another elementary particle from reality.
Some speculation on my part: When simple narratives about physics are complicated in this way, people can feel understandably disillusioned. They may crow about this feeling. ‘Gotcha! Not so pure after all, huh?’ Fine, no harm done. But sometimes the reaction goes too far, becomes a recalcitrant rejection of the results—of, well, reality.
-§-
On the subject of reality, Recht writes:
Whether or not the Higgs exists has no bearing outside the insular world of particle physics. If you don’t have a four teraelectronvolt supercollider, you can’t make a Higgs. The Higgs field has no bearing on any physics at any scale anyone would ever care about. So I don’t care either way if physicists think they found a Higgs. It has zero bearing on my existence.
A brief review of the history and physics of the Higgs boson may be in order.
The year is 1963 and quarks are mathematical figments; it will be another five years before the prediction of the W and Z bosons and a decade before the Standard Model exists. Some theorists are pondering an abstruse question: can massive vector bosons exist? In the span of several months, three independent groups totalling 6 physicists arrive at roughly the same conclusion. Symmetry demands vector bosons be massless, but by clever means, symmetry can be suppressed, and the bosons made massive.
All that is needed is a symmetrical system that allows a transition to a stable (apparently) asymmetrical state. In the canonical example, a marble rests on the center of the dome of a sombrero. The marble falls randomly down the sombrero’s slope and comes to rest stably in the trough of the hat. The hat doesn’t change shape, so it remains symmetrical, but the marble has created an apparent asymmetry. (This is only apparent; there is still symmetry across all possible marble configurations in the trough.)
The mathematical equivalent of the sombrero is the addition of a scalar field to existing equations. With this field in place, vector bosons can acquire mass. This is, in fact, quite important for Recht’s continued existence. Without the Higgs, elementary particles would not have mass, massless electrons would never be confined to orbitals, and atoms could not form. (n.b. Non-vector bosons also get mass from the Higgs field, but through a slightly different mechanism. Additionally, most—99%—of the mass in, say, a proton, results from the gluons, not quark masses from the Higgs mechanism).
One of those theorists, the recently departed Peter Higgs, realized the mechanism implied the existence of an observable:
It is worth noting that an essential feature of the type of theory which has been described in this note is the prediction of incomplete multiplets of scalar and vector bosons.
It was his paper that became synonymous with the mass-giving mechanism. For the first few years there was little reason for anyone to care because gauge theories were not renormalizable. But as physicists worked out the kinks in gauge theories, the importance of Higgs’ work became undeniable. It’s easy to see this progress through citations. From 1964 to 1971, Higgs’ paper received no more than 4 citations a year. Then, in 1972—following ‘t Hooft’s work on renormalizability—there were 12. The paper received 36 in 1973, and the number hovered in the few dozens for the next thirty years. In 2006, as LHC efforts were ramping up, citations spiked to 97. The paper now receives several hundred citations each year (and there are far more studies written about or depending on the Higgs than cite the original paper).
Part of the trouble was that no one knew what mass such a scalar boson would be. A 1976 profile of the then-hypothetical boson even considered a mass under .3 GeV. As the pieces of the Standard Model fell into place, the Higgs boson became an increasingly attractive answer to the question of mass generation.
When the Superconducting Super Collider went belly up in the ‘90s, a lonely nation (of physicists) turned its eyes to Europe. CERN committed to the LHC, which had support in part because of a ‘no-lose’ theorem; something had to contribute to WW scattering, so there was something to be discovered at that energy range below about 1000 GeV. By the time 2000 rolled around, there were even bumps in the data that seemed to hint at the Higgs. LEP, the precursor to the LHC ended up with a tantalizing excess around 115 GeV.
But that something was not clear. Many physicists working at the LHC doubted the existence of the Higgs and Higgsless models remained popular alternatives until July 4, 2012. (Due to the time delay in academic publishing, some papers with Higgsless models continued to be published throughout the following year).
All this is to say that when physicists eventually gathered enough evidence to say that a Higgs-like particle was lurking at 125 GeV in LHC data, it was part of a half-century long hunt. Unless you are some sort of hardcore frequentist, this context matters.
-§-
The sticking point here seems to be that Recht is unconvinced by the statistical rigor of physicists, and accuses them of “accepting what the open science community would deride as Highly Questionable Research Practices (lots of p-hacking, HARKing, and multiple hypothesis testing).” He asserts that 5 sigma is “another mindless convention needed for a well-functioning collaboration.” It’s true 5 sigma is an arbitrary line in the sand, but it is not a random or capricious one. I’ve written about some of the origins of the 5 sigma standard and the bumpy (pun intended) ride to standardization in particle physics, if you’re curious.
Recht is particularly concerned about the fact that CERN has a statistical committee, and galled that Science would write that 5 sigma does not literally mean there is a 1 in 3 million chance the result is wrong: “Should not be interpreted literally? Are you serious?“ Here’s how I put it in an article about Muon g-2 (emphasis retroactive)
The finding has a statistical significance of 3.1 sigma, which meets the standard baseline for evidence in particle physics. Precisely speaking, 3.1 sigma means that in the absence of new physics, statistical fluctuations would still lead the researchers to see a discrepancy between electrons and muons of 15 percent or more once every 740 times they performed the experiment. Although this would seem to suggest the observed muon-electron discrepancy is almost certainly more than a mirage, the three-sigma effect, in fact, falls well short of the gold standard of discovery in particle physics: five sigma, which works out to running the experiment 3.4 million times before seeing a statistical fluke that large. (These figures are subtly but importantly different from a one-in-740 or one-in-3.4-million chance of being wrong.)
Tommaso Dorigo has a very nice blog from 2022 about how to think about these kinds of statistics in light of disappearing LHCb anomalies. The gist is this though: Where strong priors like the Standard Model are concerned, you must compare the odds of a 3 sigma result (not uncommon) with the odds of actually seeing new physics (very rare!). What should be clear is this: particle physicists have been wrestling with questions of data and statistics for decades and there is a lot of thought and rigor that goes into these analyses. That doesn’t mean you can’t find people abusing statistics, getting excited over tiny anomalies. But there are very serious deliberations about the right way to treat the data, especially over anything as important as a claimed discovery.
Recht derisively writes that the Higgs is “an ‘object’ that has been ‘observed’ at a single location on Earth,” and he issues the following tragic dismissal:
See the bump that goes outside of their green error bars? That’s the Higgs Boson. (insert shrug emoji)
Here are the ATLAS and CMS plots with their shrug-worthy bumps.
Viewed from another perspective to better visualize the statistical significance, the bumps are anything but humble:
Recht writes that, “It requires well over 6 years of graduate study to fully understand what was supposed to be seen in the most ideal experimental situation.” Perhaps a full understanding is a lot to ask, but we can try to make sense of the bumps all the same. Cribbing from Cranmer’s rebuttal:
This bump or resonance is intimately tied to what physicists mean when they say ‘particle’. … under the null hypothesis, a certain distribution should be smoothly falling, while the alternate hypothesis would have an additoinal [sic] bump corresponding to a new particle.
What happens is that proton collisions at the LHC produce a certain number of events as a function of energy. Both ATLAS and CMS found an excess—both saw about 200 more events than they should have seen, right around 125 GeV. Even more compellingly, a good portion of this excess occurred as something decaying into two photons, for which the background is very low. (n.b. There was also 3.2 sigma evidence from the combined result of two collaborations at an entirely different collider, the Tevatron using a completely different kind of collision (proton-antiproton).)
In the years since, the Higgs boson has been discovered over and over and over again as ATLAS and CMS have accumulated orders of magnitude more data. We know, for example, the boson’s mass to a precision of about .1%.
If we are visited by aliens and they demand evidence that we are a sentient species which understands the cosmos (with the traditional ultimatum of destruction should we fall short of their expectations) I might put my hopes on plots of the Higgs mass.
-§-
Recht concludes his blog by asserting that, “The Higgs Discovery is a celebration of modern bureaucracy, not a revelation about material reality.”
There is absolutely bureaucracy and democracy, as any complete sociological account of ATLAS and CMS would show. But the analysis itself (which Recht does not engage with) is deeply connected to a physical reality. Real particles hit those detectors. Real analog signals were captured and turned into electronic data. That data was filtered, processed, analyzed, and interpreted to discover a Higgs-like particle.
The discovery of the Higgs boson was a revelation about material reality enabled by modern bureaucracy. Below a certain energy scale, much of the universe is discoverable with only modest tools and techniques. Before the dominance of colliders, physicists relied on cosmic rays from the skies. From these particles—individual events—they pieced together incredible details about the nature of reality: antimatter (positrons), a second generation of particles (muons), nuclear forces (pions), symmetry violations (kaons), even evidence of a second generation of quarks.
The Earth receives a small number of ultra high energy cosmic rays—ones which, conceivably, contain evidence of a Higgs boson decaying. The backgrounds are too large; the signals too small. One could, however, imagine an alien world, floating at the distal end of a quasar’s jet, with access to a collimated stream of highly relativistic particles. Perhaps there a lone experimenter armed with little more than a bubble chamber could discover the entire Standard Model.
But on this world, on our Earth and its comparatively meager supply of high energy cosmic rays, we must resort to colliders and the bureaucracy inherent to collaborations to probe this stage of reality, to discover the Higgs boson.
Dan Garisto 