Are there examples of posts with factual errors you think would be caught by LLMs?
One thing you could do is fact check a few likely posts and see if it's adding substantial value. That would be more persuasive than abstract arguments.
Thanks for the suggestion, I added the "Edit 1" section to the post to showcase a small study on 3 posts known to contain factual mistakes. The LLM is able to spot and correct the mistake in 2 of the 3 cases, and provides valuable (though verbose) context. Overall this seems promising to me.
Hello,
I think having some LLM-based fact checking on LessWrong posts would be a valuable addition.
To some extent comments already serve this purpose, but LLMs can do this in a more automated and systematic way, side-stepping human blind spots and tendency to conserve energy.
1. Some pros and cons
Cons:
Pros:
2. Some thoughts on implementation details
Should fact checking be done on all posts or only some of them?
I would like fact checking to be enabled on the most read posts. This includes: The Sequences, The Codex, and the Best of Less Wrong for each year. Enabling them for other popular posts also makes sense.
I am not sure about less popular posts as it might cost a lot of money. Another concern is that it could intimidate new authors, though this could be mitigated by allowing the author to run the fact-checking before publishing.
Another consideration is whether the fact-checking should happen automatically when the post reaches some degree of popularity, or should be performed on demand by (some) users.
Where should the LLM's report go?
I imagine it going in a comment, perhaps pinned or otherwise highlighted, but subject to user feedback.
What do you think are the pros & cons of LLM-based fact checking for LW posts?
If this is to be implemented, how should it be implemented?
Edit 1: 3 Case Studies
At the suggestion of Dave Orr here is a fact checking experiment for a sample of 3 posts known to contain factual mistakes:
Case 1: Conservation of Expected Evidence
As R:A-Z Errata points out
I copied the original post at Conservation of Expected Evidence and reintroduced the mistake in the input I fed to Gemini 2.5:
Gemini 2.5 Chat - My Prompt
Perform a thorough fact checking of this post and report factual mistakes """Conservation of Expected Evidence Written by Vladimir_Nesov, Ruby, Yoav Ravid, Zack_M_Davis, Lukas Finnveden, et al. last updated 19th Mar 2023 Conservation of Expected Evidence is a consequence of probability theory which states that for every expectation of evidence, there is an equal and opposite expectation of counter-evidence [1]. Conservation of Expected Evidence is about both the direction of the update and its magnitude: a low probability of seeing strong evidence in one direction must be balanced by a high probability of observing weak counter-evidence in the other direction [2]. The mere expectation of encountering evidence–before you've actually seen it–should not shift your prior beliefs. It also goes by other names, including the law of total expectation and the law of iterated expectations.
A consequence of this principle is that absence of evidence is evidence of absence.
Consider a hypothesis H and evidence (observation) E. Prior probability of the hypothesis is P(H); posterior probability is either P(H|E) or P(H|¬E), depending on whether you observe E or not-E (evidence or counter-evidence). The probability of observing E is P(E), and probability of observing not-E is P(¬E). Thus, expected value of the posterior probability of the hypothesis is:
P ( H | E ) ⋅ P ( E ) + P ( H , ¬ E ) ⋅ P ( ¬ E )
But the prior probability of the hypothesis itself can be trivially broken up the same way:
P ( H ) = P ( H , E ) + P ( H , ¬ E ) = P ( H | E ) ⋅ P ( E ) + P ( H , ¬ E ) ⋅ P ( ¬ E ) Thus, expectation of posterior probability is equal to the prior probability.
In other way, if you expect the probability of a hypothesis to change as a result of observing some evidence, the amount of this change if the evidence is positive is
D 1 = P ( H | E ) − P ( H )
If the evidence is negative, the change is
D 2 = P ( H | ¬ E ) − P ( H )
Expectation of the change given positive evidence is equal to negated expectation of the change given counter-evidence:
D 1 ⋅ P ( E ) = − D 2 ⋅ P ( ¬ E )
If you can anticipate in advance updating your belief in a particular direction, then you should just go ahead and update now. Once you know your destination, you are already there. """
Gemini 2.5 Chat - Gemini's Answer ("Grounding with Google Search" option disabled)
Okay, here is a thorough fact-checking of the provided text on Conservation of Expected Evidence (CoEE):
Overall Assessment:
The text generally captures the core idea and implications of the Conservation of Expected Evidence principle correctly. It rightly identifies it as a consequence of probability theory, explains its meaning intuitively (expected updates balance out, anticipating evidence doesn't change priors before seeing the evidence), and correctly states the relationship
D1 * P(E) = -D2 * P(¬E)
. However, there is a significant factual mistake in the mathematical derivation presented forE[Posterior] = Prior
, and some minor points could be clarified.Specific Factual Mistakes and Points of Clarification:
Mistake in Mathematical Derivation (E[Posterior] = Prior):
Expected value = P ( H | E ) ⋅ P ( E ) + P ( H , ¬ E ) ⋅ P ( ¬ E )
P ( H ) = P ( H , E ) + P ( H , ¬ E ) = P ( H | E ) ⋅ P ( E ) + P ( H , ¬ E ) ⋅ P ( ¬ E )
E[P(H|Evidence)] = P(H|E) * P(E) + P(H|¬E) * P(¬E)
.P(H) = P(H, E) + P(H, ¬E)
. Using the definition of conditional probability (P(A,B) = P(A|B)P(B)
), this becomesP(H) = P(H|E) * P(E) + P(H|¬E) * P(¬E)
.P(H, ¬E) * P(¬E)
instead ofP(H|¬E) * P(¬E)
in both its definition of the expected value and its expansion of the prior probability.P(H, ¬E) * P(¬E)
does not simplify correctly and is not equivalent toP(H|¬E) * P(¬E)
unlessP(¬E)
happens to be 1 (which would mean E is impossible).Relationship with Law of Total Expectation (LoTE) / Law of Iterated Expectations (LIE):
E[P(H|Evidence)] = P(H)
is the LoTE applied to the random variable P(H|Evidence), where the expectation is taken over the possible pieces of evidence. So, they are fundamentally linked, but calling them "other names" might suggest they are interchangeable terms in all contexts, which isn't quite true. CoEE refers specifically to this context of Bayesian belief updating.Wording of the Initial Definition:
D1 * P(E) = -D2 * P(¬E)
clarifies this: the expected increase in probability (D1 * P(E)) equals the negative of the expected decrease in probability (D2 * P(¬E)). The initial phrasing could be misinterpreted."Absence of evidence is evidence of absence":
D1 * P(E) = -D2 * P(¬E)
, if observing E would increase your belief in H (D1 > 0), then not observing E (observing ¬E) must decrease your belief in H (D2 < 0), assuming P(E) and P(¬E) are non-zero. This mathematically underpins the idea that if you expected evidence for something, its absence logically counts against it.Summary of Findings:
E[Posterior] = Prior
contains a significant error in its formulas, although the conclusion itself is correct.D1 * P(E) = -D2 * P(¬E)
, and the connection to "absence of evidence is evidence of absence" are all factually sound.The most critical correction needed is fixing the mathematical derivation steps for
E[Posterior] = Prior
.Here Gemini spots the error in the equation and corrects it:
It also notes that (among others):
I am doing these checks in Google AI Studio, one bad point for Gemini is that I first tried the query with the "Grounding with Google Search" option enabled, and it gave roughly the same output except for the math mistake which it did not spot!!
Case 2: The Quantum Physics Sequence
There is a new sequence dedicated to fixing the math in the first few posts of the original one.
The main mistake we are looking for here is that according to the original post when a photon encounters a half-silvered mirror it follows the rule:
Whereas the correct rule as stated in the new post is:
Let's see if Gemini can spot that:
Gemini 2.5 Chat - My Prompt
Perform a thorough fact checking of this post and report factual mistakes """Configurations and Amplitude by Eliezer Yudkowsky 11th Apr 2008 So the universe isn’t made of little billiard balls, and it isn’t made of crests and troughs in a pool of aether… Then what is the stuff that stuff is made of?
In Figure 1, we see, at A, a half-silvered mirror, and two photon detectors, Detector 1 and Detector 2. """ Early scientists, when they ran experiments like this, became confused about what the results meant. They would send a photon toward the half-silvered mirror, and half the time they would see Detector 1 click, and the other half of the time they would see Detector 2 click.
The early scientists—you’re going to laugh at this—thought that the silver mirror deflected the photon half the time, and let it through half the time.
Ha, ha! As if the half-silvered mirror did different things on different occasions! I want you to let go of this idea, because if you cling to what early scientists thought, you will become extremely confused. The half-silvered mirror obeys the same rule every time.
If you were going to write a computer program that was this experiment— not a computer program that predicted the result of the experiment, but a computer program that resembled the underlying reality—it might look sort of like this:
At the start of the program (the start of the experiment, the start of time) there’s a certain mathematical entity, called a configuration. You can think of this configuration as corresponding to “there is one photon heading from the photon source toward the half-silvered mirror,” or just “a photon heading toward A.”
A configuration can store a single complex value—“complex” as in the complex numbers ( a + b i ) , with i defined as √ − 1 . At the start of the program, there’s already a complex number stored in the configuration “a photon heading toward A.” The exact value doesn’t matter so long as it’s not zero. We’ll let the configuration “a photon heading toward A” have a value of ( − 1 + 0 i ) .
All this is a fact within the territory, not a description of anyone’s knowledge. A configuration isn’t a proposition or a possible way the world could be. A configuration is a variable in the program—you can think of it as a kind of memory location whose index is “a photon heading toward A”—and it’s out there in the territory.
As the complex numbers that get assigned to configurations are not positive real numbers between 0 and 1, there is no danger of confusing them with probabilities. “A photon heading toward A” has complex value −1, which is hard to see as a degree of belief. The complex numbers are values within the program, again out there in the territory. We’ll call the complex numbers amplitudes.
There are two other configurations, which we’ll call “a photon going from A to Detector 1” and “a photon going from A to Detector 2.” These configurations don’t have a complex value yet; it gets assigned as the program runs.
We are going to calculate the amplitudes of “a photon going from A toward 1” and “a photon going from A toward 2” using the value of “a photon going toward A,” and the rule that describes the half-silvered mirror at A.
Roughly speaking, the half-silvered mirror rule is “multiply by 1 when the photon goes straight, and multiply by i when the photon turns at a right angle.” This is the universal rule that relates the amplitude of the configuration of “a photon going in,” to the amplitude that goes to the configurations of “a photon coming out straight” or “a photon being deflected.”[1]
So we pipe the amplitude of the configuration “a photon going toward A,” which is ( − 1 + 0 i ) , into the half-silvered mirror at A, and this transmits an amplitude of ( − 1 + 0 i ) × i ( 0 − i ) to “a photon going from A toward 1,” and also transmits an amplitude of ( − 1 + 0 i ) × 1
( − 1 + 0 i ) to “a photon going from A toward 2.”
In the Figure 1 experiment, these are all the configurations and all the transmitted amplitude we need to worry about, so we’re done. Or, if you want to think of “Detector 1 gets a photon” and “Detector 2 gets a photon” as separate configurations, they’d just inherit their values from “A to 1” and “A to 2” respectively. (Actually, the values inherited should be multiplied by another complex factor, corresponding to the distance from A to the detector; but we will ignore that for now, and suppose that all distances traveled in our experiments happen to correspond to a complex factor of 1.)
So the final program state is:
Configuration “a photon going toward A”: (−1+0i) Configuration “a photon going from A toward 1”: (0−i) Configuration “a photon going from A toward 2”: (−1+0i) and optionally
Configuration “Detector 1 gets a photon”: (0−i) Configuration “Detector 2 gets a photon”: (−1+0i). This same result occurs—the same amplitudes stored in the same configurations—every time you run the program (every time you do the experiment).
Now, for complicated reasons that we aren’t going to go into here— considerations that belong on a higher level of organization than fundamental quantum mechanics, the same way that atoms are more complicated than quarks—there’s no simplemeasuring instrument that can directly tell us the exact amplitudes of each configuration. We can’t directly see the program state.
So how do physicists know what the amplitudes are?
We do have a magical measuring tool that can tell us the squared modulus of a configuration’s amplitude. If the original complex amplitude is ( a + b i ) , we can get the positive real number ( a 2 + b 2 ) . Think of the Pythagorean theorem: if you imagine the complex number as a little arrow stretching out from the origin on a two-dimensional plane, then the magic tool tells us the squared length of the little arrow, but it doesn’t tell us the direction the arrow is pointing.
To be more precise, the magic tool actually just tells us the ratios of the squared lengths of the amplitudes in some configurations. We don’t know how long the arrows are in an absolute sense, just how long they are relative to each other. But this turns out to be enough information to let us reconstruct the laws of physics—the rules of the program. And so I can talk about amplitudes, not just ratios of squared moduli.
When we wave the magic tool over “Detector 1 gets a photon” and “Detector 2 gets a photon,” we discover that these configurations have the same squared modulus—the lengths of the arrows are the same. Thus speaks the magic tool. By doing more complicated experiments (to be seen shortly), we can tell that the original complex numbers had a ratio of i to 1.
And what is this magical measuring tool?
Well, from the perspective of everyday life—way, way, way above the quantum level and a lot more complicated—the magical measuring tool is that we send some photons toward the half-silvered mirror, one at a time, and count up how many photons arrive at Detector 1 versus Detector 2 over a few thousand trials. The ratio of these values is the ratio of the squared moduli of the amplitudes. But the reason for this is not something we are going to consider yet. Walk before you run. It is not possible to understand what happens all the way up at the level of everyday life, before you understand what goes on in much simpler cases.
For today’s purposes, we have a magical squared-modulus-ratio reader. And the magic tool tells us that the little two-dimensional arrow for the configuration “Detector 1 gets a photon” has the same squared length as for “Detector 2 gets a photon.” That’s all.
You may wonder, “Given that the magic tool works this way, what motivates us to use quantum theory, instead of thinking that the half-silvered mirror reflects the photon around half the time?”
Well, that’s just begging to be confused—putting yourself into a historically realistic frame of mind like that and using everyday intuitions. Did I say anything about a little billiard ball going one way or the other and possibly bouncing off a mirror? That’s not how reality works. Reality is about complex amplitudes flowing between configurations, and the laws of the flow are stable.
But if you insist on seeing a more complicated situation that billiard-ball ways of thinking can’t handle, here’s a more complicated experiment.
In Figure 2, B and C are full mirrors, and A and D are half-mirrors. The line from D to E is dashed for reasons that will become apparent, but amplitude is flowing from D to E under exactly the same laws.
Now let’s apply the rules we learned before:
At the beginning of time “a photon heading toward A” has amplitude ( − 1 + 0 i ) .
We proceed to compute the amplitude for the configurations “a photon going from A to B” and “a photon going from A to C”:
“a photon going from A to B” = i
× a photon heading toward A” = ( 0 − i ) Similarly,
“a photon going from A to C” = 1 × a photon heading toward A” = ( − 1 + 0 i ) The full mirrors behave (as one would expect) like half of a half-silvered mirror—a full mirror just bends things by right angles and multiplies them by i. (To state this slightly more precisely: For a full mirror, the amplitude that flows, from the configuration of a photon heading in, to the configuration of a photon heading out at a right angle, is multiplied by a factor of i.)
So:
“a photon going from B to D” = i
× “a photon going from A to B” = ( 1 + 0 i ) , “a photon going from C to D” = i
× “a photon going from A to C” = ( 0 − i ) “B to D” and “C to D” are two different configurations—we don’t simply write “a photon at D”—because the photons are arriving at two different angles in these two different configurations. And what D does to a photon depends on the angle at which the photon arrives.
Again, the rule (speaking loosely) is that when a half-silvered mirror bends light at a right angle, the amplitude that flows from the photon-going-in configuration to the photon-going-out configuration, is the amplitude of the photon-going-in configuration multiplied by i. And when two configurations are related by a half-silvered mirror letting light straight through, the amplitude that flows from the photon-going-in configuration is multiplied by 1.
So:
From the configuration “a photon going from B to D,” with original amplitude(1+0i)
Amplitude of ( 1 + 0 i ) × i ( 0 + i ) flows to “a photon going from D to E.” Amplitude of ( 1 + 0 i ) × 1
( 1 + 0 i ) flows to “a photon going from D to F. ” From the configuration “a photon going from C to D,” with original amplitude(0−i)
Amplitude of ( 0 − i ) × i ( 1 + 0 i ) flows to “a photon going from D to F. ” Amplitude of ( 0 − i ) × 1
( 0 − i ) flows to “a photon going from D to E.” Therefore:
The total amplitude flowing to configuration “a photon going from D to E” is ( 0 + i ) + ( 0 − i ) ( 0 + 0 i ) 0 . The total amplitude flowing to configuration “a photon going from D to F” is ( 1 + 0 i ) + ( 1 + 0 i )
( 2 + 0 i ) . (You may want to try working this out yourself on pen and paper if you lost track at any point.)
But the upshot, from that super-high-level “experimental” perspective that we think of as normal life, is that we see no photons detected at E. Every photon seems to end up at F. The ratio of squared moduli between “D to E” and “D to F” is 0 to 4. That’s why the line from D to E is dashed, in this figure.
This is not something it is possible to explain by thinking of half-silvered mirrors deflecting little incoming billiard balls half the time. You’ve got to think in terms of amplitude flows.
If half-silvered mirrors deflected a little billiard ball half the time, in this setup, the little ball would end up at Detector 1 around half the time and Detector 2 around half the time. Which it doesn’t. So don’t think that.
You may say, “But wait a minute! I can think of another hypothesis that accounts for this result. What if, when a half-silvered mirror reflects a photon, it does something to the photon that ensures it doesn’t get reflected next time? And when it lets a photon go through straight, it does something to the photon so it gets reflected next time.”
Now really, there’s no need to go making the rules so complicated. Occam’s Razor, remember. Just stick with simple, normal amplitude flows between configurations.
But if you want another experiment that disproves your new alternative hypothesis, it’s Figure 3.
Here, we’ve left the whole experimental setup the same, and just put a little blocking object between B and D. This ensures that the amplitude of “a photon going from B to D” is 0.
Once you eliminate the amplitude contributions from that configuration, you end up with totals of ( 1 + 0 i ) in “a photon going from D to F, ” and ( 0 − i ) in “a photon going from D to E.”
The squared moduli of ( 1 + 0 i ) and ( 0 − i ) are both 1, so the magic measuring tool should tell us that the ratio of squared moduli is 1. Way back up at the level where physicists exist, we should find that Detector 1 goes off half the time, and Detector 2 half the time.
The same thing happens if we put the block between C and D. The amplitudes are different, but the ratio of the squared moduli is still 1, so Detector 1 goes off half the time and Detector 2 goes off half the time.
This cannot possibly happen with a little billiard ball that either does or doesn’t get reflected by the half-silvered mirrors.
Because complex numbers can have opposite directions, like 1 and −1, or i and −i, amplitude flows can cancel each other out. Amplitude flowing from configuration X into configuration Y can be canceled out by an equal and opposite amplitude flowing from configuration Z into configuration Y. In fact, that’s exactly what happens in this experiment.
In probability theory, when something can either happen one way or another, X or ¬X, then P ( Z )
P ( Z | X ) P ( X ) + P ( Z | ¬ X ) P ( ¬ X ) . And all probabilities are positive. So if you establish that the probability of Z happening given X is 1 2 , and the probability of X happening is 1 3 , then the total probability of Z happening is at least 1 6 no matter what goes on in the case of ¬X. There’s no such thing as negative probability, less-than-impossible credence, or ( 0 + i ) credibility, so degrees of belief can’t cancel each other out like amplitudes do.
Not to mention that probability is in the mind to begin with; and we are talking about the territory, the program-that-is-reality, not talking about human cognition or states of partial knowledge.
By the same token, configurations are not propositions, not statements, not ways the world could conceivably be. Configurations are not semantic constructs. Adjectives like probable do not apply to them; they are not beliefs or sentences or possible worlds. They are not true or false but simply real.
In the experiment of Figure 2, do not be tempted to think anything like: “The photon goes to either B or C, but it could have gone the other way, and this possibility interferes with its ability to go to E…”
It makes no sense to think of something that “could have happened but didn’t” exerting an effect on the world. We can imagine things that could have happened but didn’t—like thinking, “Gosh, that car almost hit me”—and our imagination can have an effect on our future behavior. But the event of imagination is a real event, that actually happens, and that is what has the effect. It’s your imagination of the unreal event—your very real imagination, implemented within a quite physical brain—that affects your behavior.
To think that the actual event of a car hitting you—this event which could have happened to you, but in fact didn’t—is directly exerting a causal effect on your behavior, is mixing up the map with the territory.
What affects the world is real. (If things can affect the world without being “real,” it’s hard to see what the word “real” means.) Configurations and amplitude flows are causes, and they have visible effects; they are real. Configurations are not possible worlds and amplitudes are not degrees of belief, any more than your chair is a possible world or the sky is a degree of belief.
So what is a configuration, then?
Well, you’ll be getting a clearer idea of that in later essays.
But to give you a quick idea of how the real picture differs from the simplified version we saw in this essay…
Our experimental setup only dealt with one moving particle, a single photon. Real configurations are about multiple particles. The next essay will deal with the case of more than one particle, and that should give you a much clearer idea of what a configuration is.
Each configuration we talked about should have described a joint position of all the particles in the mirrors and detectors, not just the position of one photon bopping around.
In fact, the really real configurations are over joint positions of all the particles in the universe, including the particles making up the experimenters. You can see why I’m saving the notion of experimental results for later essays.
In the real world, amplitude is a continuous distribution over a continuous space of configurations. This essay’s “configurations” were blocky and digital, and so were our “amplitude flows.” It was as if we were talking about a photon teleporting from one place to another.
If none of that made sense, don’t worry. It will be cleared up in later essays. Just wanted to give you some idea of where this was heading.
[Editor’s Note: Strictly speaking, a standard half-silvered mirror would yield a rule “multiply by −1 when the photon turns at a right angle,” not “multiply by i.” The basic scenario described by the author is not physically impossible, and its use does not affect the substantive argument. However, physics students may come away confused if they compare the discussion here to textbook discussions of Mach–Zehnder interferometers. We’ve left this idiosyncrasy in the text because it eliminates any need to specify which side of the mirror is half-silvered, simplifying the experiment.]
Gemini 2.5 - Gemini's Answer ("Grounding with Google Search" option disabled)
Okay, here is a thorough fact-checking of the provided text "Configurations and Amplitude" by Eliezer Yudkowsky:
Overall Assessment:
The post aims to provide an intuitive introduction to quantum mechanics concepts like superposition and interference using complex amplitudes and configurations, exemplified by beam splitters and mirrors (forming a Mach-Zehnder interferometer). While it successfully conveys the qualitative strangeness of QM and the failure of classical "billiard ball" intuition, it contains a significant factual error in the specific physical rules it presents for the beam splitter, which leads to mathematically incorrect results (though the relative probabilities sometimes coincidentally match standard results). The philosophical framing reflects a specific interpretation (likely Many-Worlds) presented strongly.
Specific Factual Mistakes and Inaccuracies:
Incorrect Beam Splitter (Half-Silvered Mirror) Rule:
1/√2
for transmission andi/√2
for reflection (or1/√2
and-1/√2
, or other variations depending on phase conventions, but always ensuring the sum of squared moduli is 1). For example, |A/√2|^2 + |iA/√2|^2 = |A|^2/2 + |A|^2/2 = |A|^2, conserving probability.Incorrect Amplitude Values in Calculations:
Potentially Misleading Analogy for "Configuration":
Presentation of Interpretation as Fact:
Points That Are Factually Correct (within the context of QM principles):
Conclusion:
The post is a valuable attempt to build intuition for quantum interference and the role of complex amplitudes. However, its pedagogical simplification of the beam splitter physics crosses into factual inaccuracy by using a rule that violates probability conservation (unitarity), a cornerstone of QM. This leads to quantitatively incorrect amplitude values in the calculations, even though the qualitative interference effects described (like cancellation at one detector) are characteristic of the actual physics. Furthermore, it presents a specific philosophical interpretation of QM's ontology rather strongly. A reader should be aware of the incorrect physics rule used for the beam splitter and understand that the ontological claims about the reality of amplitudes/configurations represent one specific viewpoint (MWI) within a broader interpretive landscape.
Again, Gemini finds the mistake:
Case 3: Book Review: The Origins Of Woke
Scott Alexander maintains a mistakes page. The newest entry is for his book review The Origins Of Woke:
Can Gemini 2.5 spot that?
Gemini 2.5 Chat - My Prompt
Perform a thorough fact-checking of this post and report factual mistakes """The Origins Of . . . Civil Rights Law The Origins Of Woke, by Richard Hanania, has an ambitious thesis. And it argues for an ambitious thesis. But the thesis it has isn’t the one it argues for.
The claimed thesis is “the cultural package of wokeness is downstream of civil rights law”. It goes pretty hard on this. For example, there’s the title, The Origins Of Woke. Or the Amazon blurb: “The roots of the culture lie not in the culture itself, but laws and regulations enacted decades ago”. Or the banner ad:
The other thesis, the one it actually argues for, is “US civil rights law is bad”. On its own, this is a fine thesis. A book called Civil Rights Law Is Bad would - okay, I admit that despite being a professional Internet writer I have no idea how the culture works anymore, or whether being outrageous is good or bad for sales these days. We’ll never know, because Richard chose to wrap his argument in a few pages on how maybe this is the origin of woke or something. Still, the book is on why civil rights law is bad.
Modern civil rights law is bad (he begins) for reasons baked into its history. The original Civil Rights Act of 1964 was supposed to be an ad hoc response to the outrageous level of anti-black racism going on in the South, which protests and TV news had finally brought to the attention of the white majority. There was broad support for a bill which was basically “don’t be the KKK”.
Sex discrimination got tacked on half as a joke, half as a poison pill by its enemies to make the bill unpalatable (fact check: true - but there’s a deeper story, see this Slate article for more details). Ideas about “affirmative action” and “disparate impact” weren’t tacked on at all; the bill’s proponents denied that it could be used to justify anything of the sort, and even agreed to include language in the bill saying it was against that. Still, after the bill was passed, a series of executive orders, judicial decisions, and bureaucratic power grabs put all those things in place.
The key point here is that “quotas”, or any kind of “positive discrimination” where minorities got favored over more-qualified whites, were anathema to lawmakers and the American people. But civil rights activists, the courts, and the bureaucracy really wanted those things. So civil rights law became a giant kludge that effectively created quotas and positive discrimination while maintaining plausible deniability. This ended up as the worst of both worlds. Hanania specifically complains about1:
Affirmative Action
Hanania’s take on affirmative action involves the government sending companies a message like this:
We notice your workforce has fewer minorities than the applicant pool.
If this remains true, we’ll sue you for millions of dollars and destroy your company. So by the next time we check, your workforce had better have exactly many minorities as the applicant pool.
But you’re not allowed to explicitly favor minority applicants over whites. You certainly can’t do anything flagrant, like set a quota of minority employees equal to their level in the applicant pool.
Have fun!
This satisfied the not-really-paying attention white electorate, because politicians could tell them that “quotas are illegal, we’re sure not doing anything like that”. And it satisfied civil rights activists, because inevitably businesses/departments came up with secret ways to favor minorities until representation reached the level where they wouldn’t get sued.
A recent case illustrates the results of this double-bind. The FAA hires air traffic controllers. They used to judge applicants based on a test which measured their skills at air traffic control. This resulted in comparatively few black air traffic controllers. Various civil rights groups put pressure on them, and they replaced the test with a “biographical questionnaire”. The questionnaire asked weird unrelated questions about your life, and you got points if you gave the answer that the FAA thought black people might give (for example, if you said your worst subject was science). This still didn’t get them enough black employees, so they secretly told black communities exactly what answers to put on the questionnaire to go through.
It’s easy to blame the FAA here, but (Hanania says) civil rights law almost forces you to do something like this. People tried simpler things, like keeping a test but giving minority applicants extra points. The courts and civil rights bureaucracy struck these down as illegal. The almost-explicit policy was that you had to get more minority employees, but you had to hide it carefully enough that the American people (who were still against racial preferences) wouldn’t catch on.
Disparate Impact
Not only can you not explicitly discriminate, you can’t use hiring criteria that “accidentally” discriminate by favoring one race over another. To give a stupid example, if someone refused to hire anyone from Detroit, this would have “disparate impact” since Detroit is a majority black city. If you allowed stuff like this, racists could covertly discriminate by using these sorts of rules.
But Hanania challenges us to think of any criterion that isn’t potentially racially biased. For example, we know universities discriminate against Asians, so only hiring people with college degrees is a “disparate impact”. We know that more men than women have experience as miners, so a mining company only hiring employees with experience is a “disparate impact”. Since whites typically do better on IQ tests than blacks, and all cognitive skills are correlated with IQ, the Supreme Court decided in Duke vs. Griggs that all tests of any ability were potentially disparate impact, and you opened yourself to lawsuits if you used any of them.
(in theory, companies are allowed to use tests and similar criteria if they prove them nondiscriminatory. But the standards for this - they have to prove it for each race and each job site individually - are so high that, in practice, few companies take this route.)
Since this technically banned all possible criteria, companies couldn’t follow the letter of the law. Instead they hired fancy lawyers to tell them which way the winds were blowing. The lawyers told them that college degrees were okay, resumes with biographies and experience were maybe okay, and interviews were okay. Tests were out. Anything more creative was out.
A disparate impact case made the news recently. The Biden EEOC sued convenience store chain Sheetz for running criminal background checks on their employees. They didn’t allege any intentional discrimination. They just said that more minorities fail criminal background checks than whites, therefore it’s disparate impact, therefore Sheetz has to drop the criminal background check.
(the article links to another case where the Obama EEOC sued a corporate events planner, demanding they give monetary compensation to an employee who they had refused to hire simply because he had committed attempted murder and lied about it on their job application)
Is Sheetz the only company that does criminal background checks on its employees? Do they do the background checks differently than any other company? My understanding is that the technical answer is that to do background checks without being sued, you have to prove in some very formal way that the specific crimes you’re looking for would be bad for your specific industry, and maybe Sheetz didn’t prove that a general history of violence was bad for convenience stores. But if this sounds kind of fake to you, and you’re wondering whether the real rule is “the government has wide discretion to prosecute whoever it feels like”, Hanania’s answer is “definitely yes”.
His position is that all of these rules are so broad that every company is always violating them in some sense. No company has exactly the same distribution of minorities as “the applicant pool”, whatever that is. No company has some magical hiring rule that has literally zero correlation with race, especially since black people are on average poorer, less educated, and less likely to have any given achievement (so any attempt to choose better employees over worse will necessarily disadvantage them). In real life, the bureaucracy’s rules are something like “don’t do anything different from other companies in your industry, and especially don’t be caught seeming less woke”. Hanania argues this creates an arms race / ratchet. Every company wants to be at least 50th percentile wokeness or above. But not every company can be above average. So everybody gets more and more woke, with no end in sight.
Continuing with Sheetz: according to the article on the lawsuit, in 2020 they “introduced the IDEA initiative”, ie Inclusion Diversity Equality & Accessibility. Their website has a big picture of a black woman saying “We’re Building A Great Place To Work For All”, and boasts that they’ve created a special forum for black employees. They’ve made 60% of managers women, started a Woman’s Leadership Program, offered generous maternity leave, and written letters to the Chamber of Commerce on how the George Floyd murder made them realize “we quickly needed to learn, listen, be vulnerable and humbly approach . . . culture-shifting work”. Companies hope that if they do enough of this stuff, the EEOC will agree they’re an ally in the civil rights project and not sue them under their wide discretion to sue basically anyone. Too bad they’re getting sued anyway; some other convenience store must have done more of this stuff. Surely some executive is wishing they had just tried having one more mandatory diversity training…
Harassment Law
Harassment law might win the award for most complicated chain of reasoning from real legislation to enforcement:
Legislation says you can’t discriminate against minorities
If you bully minorities out of your company, that would be a way to discriminate against them.
So you can’t have an environment that’s so hostile to minorities that they inevitably leave.
In some sense, anything that offends a minority is part of this environment.
Any joke, political comment, flirtation, etc, could potentially offend a minority.
Therefore, every business owner needs to monitor their employees for jokes, political comments, flirtatiousness, and take action against any offenses.
Hanania has several complaints here. First and most legibly, it (say it with me) gets taken too far.
Volokh lists a large number of [examples of things that have been found to be] evidence of a hostile work environment: signs with the phrase “men working”; “draftsman” and “foreman” as job titles, pictures of Ayatollah Khomeini and a burning American flag in a cubicle; an ad campaign using samurai, kabuki, and sumo wrestling to refer to Japanese competition; jokes of a sexual nature not targeted at any particular person; misogynistic rap music […] even terms like “great view” and “walk-up” have been cited as potentially trying to exclude blind people and those in wheelchairs.
And
In a 2015 and 2016, a black father and son named Owen Diaz and Demetric Di-az2 [sic] worked at a Tesla plant. They sued the company for racial discrimination, with the father’s claims alone making it to trial….racial slurs were used in the presence of Diaz, and he saw racist graffiti on a bathroom wall. It appears that the workers allegedly responsible were mostly or all minorities themselves, and each time an allegation could be verified, the employee was punished. Tesla claimed that they had taken enough steps to address the concerns of Diaz […] a jury disagreed, and awarded the plaintiff $137 million, an amount that the judge reduced to $15 million. In response to the verdict, Tesla released a statement pointing out that witnesses confirmed that the slurs were used in a friendly manner, usually by African-American employees, and without hostile intent.
(fact check: this article says the racism also included demands to “go back to Africa” people leaving drawings of caricatured black cavemen at the employee’s desk, threats, and claims that black employees were "given the most menial and physically demanding work" - and that these claims were backed up by testimony from two dozen former workers and a cellphone video showing people telling a black employee that they are going to “cut you up, n—-r”. This seems like a sufficiently different story that I’d like to know whether Hanania still stands by his version)
Other parts of harassment law lead to more unfair double-binds. For example, you can’t be seen to “retaliate” against someone who accuses another worker of harassment. So suppose that a minority employee is bullying a white employee, the white employee resists, and the minority accuses them as “harassment”. Maybe there’s even a full trial, everyone agrees this is what happened, and the white employee is found totally innocent. Still, you can’t fire the bully, because that would be retaliation for a harassment complaint. And since you probably don’t want the bully and their victim in the same department, you need to move one of them. And you can’t move the bully, because that would be viewed as “retaliation” for the harassment complaint and they could sue you for millions of dollars. So you have to punish the victim.
But Hanania doesn’t just say this kind of thing goes too far. He has some broader point that I have trouble interpreting - basically that corporations used to be cozy, chummy places full of banter and flirtation that everyone enjoyed, and now this has been universally replaced with the bland soul-draining bureaucratic corporate aesthetic satirized in works like Office Space.
Is this true? People talk about Mad Men (I’ve never seen it) as reflecting some kind of corporate golden age where at least high-ranking men enjoyed their jobs. If so, did it change because of harassment law? Or because neoliberalism replaced the work-for-thirty-years-and-get-a-golden-watch corporation with the work-for-three-years-and-then-seek-a-better-job-elsewhere corporation?
Still, Hanania really hammers in this point that we should apparently all be angry about the loss of corporate flirtation - he calls the current regime, “a sexless, androgynous, and sanitized workplace” which is “contrary to human nature [and] miserable”. Without civil rights law, we could have “organizations that combined the aspects of a church, a social club, a matchmaking service, and a traditional business.”
In such a world:
Some corporations start encouraging dating and forming close personal bonds among their employees. This can take many forms, from Christian matchmaking to promoting a party-like atmosphere. These pro-relationship corporations will come in conservative or liberal forms. Other firms explicitly market themselves as providing a more “professional” or “classic” work experience . . . we will see a period of wild experimentation, with some forms of corporate organization drawing a great deal of media coverage. People will criticize many of these experiments, and they will become the subject of public outrage. After civil rights law has been defanged, however, government no longer has the ability to easily shut such efforts down. Eventually, public anger subsides, and the idea of the media attacking a firm because it dislikes its internal culture will seem as intolerant as attacking a religious community for its doctrines, or homosexuals for what they do together as consenting adults.
I appreciate my anti-civil-rights books doubling as interesting settings for pornographic stories, but I’m otherwise unable to fathom the level of Hanania’s enthusiasm here.
…And More
Richard Hanania hates all this stuff.
Partly he hates it because he thinks it’s unfair and anti-business and anti-merit. But also, Vaclav Havel talks about the indignity of life under communism. You weren’t allowed to just do your job and pay your taxes and follow the laws of the communist state. You had to be actively complicit. You had to act enthusiastic about the communism, force it upon others, inform on your colleagues and punish deviation - at least if anybody was going to check later. This kind of communism didn’t just hurt your pocketbook. It damaged your soul. It molded you into a worse and uglier type of person who would eventually abandon their better impulses in order to justify their actions to themselves.
This is how Hanania thinks of civil rights law. Business owners can’t just give blacks ten extra points on the screening test and call it a day. They have to favor blacks while insisting to everyone that they don’t do this and it’s perfectly fair and they love civil rights law. They have to twist their employment criteria into some kind of illegible monstrosity so nobody can notice all the favoritism they’re doing, then tell everybody that they believe the monstrosity is “fairer”. They have to hire a bunch of diversity coordinators - not because they’re required to hire diversity coordinators, it’s not a requirement - but because they love equality so so much (and if they don’t do this, they’ll get sued for seemingly unrelated reasons). Everyone faces a constant threat of lawsuits which can only be warded against by seeming maximally woke and maximally enthusiastic and maximally happy about all the idiotic fake laws you are being forced to comply with.
Like in communism, you have to become your own mini-police state. You have to make employees snitch on each other if they tell the wrong joke. You have to turn your company into a tyranny of HR ladies. If you do any of this even a little less than other companies, you’ll get sued for seemingly unrelated reasons, with penalties running potentially into the hundreds of millions of dollars.
Because there’s no legible law except “be the same as everyone else so you don’t stand out as sue-able”, every corporation homogenizes into the same bland HR-ocracy. Everyone agrees on the same hiring process, which is to prioritize college degree, resume, and interview, and definitely not any test or measure of ability. This leads inevitably to our current society, where everyone has to waste their childhood doing meaningless extracurriculars so they can get into the best college so they can take the best internships so they get the best jobs.
(unless they do something stupid like let themselves get the dreaded “resume gap”).
But also:
During the early 1800s, government positions were given out by the “spoils system”, basically “does the party in power like you personally?” In the 1880s, after President Garfield was assassinated by a guy who didn’t get a good enough position, they switched to a formal civil service, based on test performance and merit. The US civil service became the envy of the world, attracted some of the smartest people in the country, and obviously worked better than the old system wherever it was possible to compare. Still, this gradually (and somewhat deniably) ended in the 1970s, because the merit-based hiring system seemed like disparate impact. Hanania calls the current era “the racial spoils system”, where positions in the bureaucracy are based on the same kind of illegible morass as everything else (eg the FAA’s “biographical questionnaire”). He says every branch of government has become less effective as a result.
Hanania doesn’t mention this, but I’ve heard an additional argument elsewhere. It’s legally dangerous for companies to hire based on anything like merit. Still, if you have great lawyers and are willing to pay a lot to settle lawsuits, you can get away with legally dangerous things. This is only worth it if you really really want high-merit employees, ie if the best employee is much more financially valuable to you than the second-best. This is mostly true in Wall Street (where you want your trader to outsmart the other guy’s trader by half a millisecond or whatever) and Silicon Valley (where ten employees can write a program used by millions of people). So the government, the civil service, the schools, etc, all abandoned merit-based hiring, while Wall Street and Silicon Valley lawyered up. But that means that if you’re a smart non-minority college graduate, you know that joining the civil service will be a mess - you’ll have a tough time even getting in, and you’ll always be passed over for promotions for less-qualified minorities. Meantime, Wall Street and Silicon Valley would love to have you. So all the smart people got concentrated in a few industries that might not have been their most economically productive use, and the old American tradition where elite families would send some of their kids into public service died out.
What To Do?
Hanania stresses that most Americans hate affirmative action (and probably by extension most other civil rights law, though they’ve probably never heard of disparate impact). Affirmative action has been on the ballot nine times, and failed eight of those. Most recently, it failed in California, a deep-blue, 66% minority state where the pro-AA side outspent opponents 17-to-1.
Also, Republicans have controlled all the branches of government many times in the past fifty years, and now they control the Supreme Court. Most civil rights law is based on executive orders and judicial decisions, so you wouldn’t even need a Congressional vote to overturn it. Just an executive order, from any president who felt like it. Reagan could have overturned half of this with the stroke of a pen, if he’d wanted. So how has it survived this long?
His answer: because until about 2010, Republicans were too scared of getting called racist. Reagan wanted to overturn affirmative action, but other Republicans (like Bob Dole) begged him not to, because racism, and eventually he caved. But since 2010, everyone has already been calling Republicans racist all the time, to the point where probably this threat has lost its power. And the sort of moderate Republicans who reined in Reagan are gone. So why haven’t Republicans (eg Trump) acted? Hanania thinks everyone is so obsessed with “woke” culture war stuff that the low-hanging fruit of actual woke laws that presidents can change has slipped under the radar.
And so, this book. I would have summarized the case as “Hey, Republicans! Do you hate wokeness? Well, too bad, it’s a vast cultural movement with bastions in a bunch of places where we have no power. But some of this civil rights law stuff seems pretty related to wokeness, and we do potentially have power there. So instead of fighting the unwinnable cultural battle, how about we fight the very winnable policy one?”
But maybe this didn’t seem optimistic enough for Hanania, so he framed it as “the legal wokeness is the source of the cultural wokeness” instead. More on this later.
The Origins Of . . . Inequality A progressive, reading this book, might counter: “Sure, civil rights law - like all law - is poorly written and kludgy in parts. Like all law, it sometimes gets abused or taken too far. Those are the costs. But the benefits are that it fights discrimination and inequality. That’s very important! Don’t you think those benefits are worth the cost?”
Unless I missed it, Hanania doesn’t touch this obvious counterargument. He briefly says that in a free market, companies couldn’t consistently maintain discrimination, because that would be leaving money on the ground.
“Cool theoretical result,” objects the hypothetical opponent. “But white households earn an average of $80K and black households an average of $50K, and so on with other minority groups. So it sure seems like something inequality-related is going on.”
My tongue-in-cheek reframing of Hanania’s summary of civil rights law went:
We notice your workforce is less black than the applicant pool.
If this remains true, we’ll sue you for millions of dollars and destroy your company. So by the next time we check, your workforce had better be exactly as black as the general population.
But you’re not allowed to explicitly favor black applicants over whites. You certainly can’t do anything flagrant, like set a quota of black employees equal to their level in the applicant pool.
Have fun!
Our hypothetical opponent could argue there’s nothing necessarily contradictory or Orwellian about this. If your company is whiter than its applicant pool (eg the general population), then you must be discriminating. If you stop discriminating, you can get racial balance without any of that nasty quota stuff. So what’s the problem?
Everyone is so circumspect when talking about race that I can never figure out what anyone actually knows or believes. Still, I think most people would at least be aware of the following counterargument: suppose you’re the math department at a college. You might like to have the same percent black as the general population (13%). But far fewer than 13% (let’s say 2%) of good math PhDs are black. So it’s impossible for every math department to hire 13% black math professors unless they lower their standards or take some other drastic measure.
Okay, says our hypothetical opponent. Then that means math grad programs are discriminating against blacks. Fine, they’re the ones we should be investigating for civil rights violations.
No, say the math grad programs, fewer than 13% of our applicants are black too.
Fine, then the undergrad programs are the racists. Or if they can prove they’re not, then the high schools are racist and we should do busing. The point is, somebody somewhere along the line has to be racist, right?
I know of four common, non-exclusive answers to this question.
Yes, the high schools (or whatever) are racist. And if you can present a study proving that high schools aren’t racist, then it’s the elementary schools. And if you have a study there too, it’s the obstetricians, giving black mothers worse pregnancy care. If you have a study disproving that too, why are you collecting all these studies? Hey, maybe you’re the racist!
Maybe institutions aren’t too racist today, but there’s a lot of legacy of past racism, and that means black people are poor. And poor people have fewer opportunities and do worse in school. If you have a study showing that black people do worse even when controlled for income, then maybe it’s some other kind of capital, like educational capital or social capital. If you have studies about those too, see above.
Black people have a bad culture. Something something shoes and rap music, trying hard at school gets condemned as “acting white”. They should hold out for a better culture. I hear nobody’s using ancient Sumerian culture these days, maybe they can use that one.
White people have average IQ 100, black people have average IQ 85, this IQ difference accurately predicts the different proportions of whites and blacks in most areas, most IQ differences within race are genetic, maybe across-race ones are genetic too. I love Hitler and want to marry him.
None of these are great options, and I think most people work off some vague cloud of all of these and squirm if you try to make them get too specific. I don’t exactly blame Hanania for not taking a strong stand here. It’s just strange to assume civil rights law is bad and unnecessary without having any opinion on whether any of this is true, whether civil rights law is supposed to counterbalance it, and whether it counterbalances it a fair amount.
A cynic might notice that in February of this year, Hanania wrote Shut Up About Race And IQ. He says that the people who talk about option 4 are “wrong about fundamental questions regarding things like how people form their political opinions, what makes for successful movements, and even their own motivations.” A careful reader might notice what he doesn’t describe them as being wrong about. The rest of the piece almost-but-not-quite-explicitly clarifies his position: I read him as saying that race realism is most likely true, but you shouldn’t talk about it, because it scares people.
(I’m generally against “calling people out” for believing in race realism. I think people should be allowed to hide beliefs that they’d get punished for not hiding. I sympathize with some of these positions and place medium probability on some weak forms of them. I think Hanania is open enough about where he’s coming from that this review doesn’t count as a callout.)
His foil here is race realist Nathan Cofnas, who says you have to discuss these things. Otherwise progressives can win every argument by using the line of reasoning above - “Just look how much inequality there still is, this shows there’s still lots of racism or at least the lingering effects of past racism, obviously our job isn’t done yet and we need lots more civil rights law to combat it.”
Hanania’s answer to Cofnas is that this isn’t a debate club. “Ah, but Glaucon, your claim that affirmative action is unnecessary must imply the corollary that there must be no inequality, thus proving a contradiction.” LOL no. Realistically this will get fought on the level of “You oppose affirmative action, which makes you a gross Nazi” vs. “You support affirmative action, which makes you an annoying wokescold.” Just say the wokescold thing louder than your enemies say the Nazi thing, and you win. Talking about racial differences scares people off and doesn’t help.
I find it hard not to feel contempt for this level of contempt for reason, but Hanania is no doubt right about the strategic considerations. And in his book, he follows his own principle. There’s no discussion of why civil rights law might be necessary, or why it might be impossible for companies to hire enough minorities without reverse discrimination. As he predicts on his blog, it’s not fatal. You wouldn’t notice unless you were looking for it.
I’m not really sure what to do here. How do you review a book that has a glaring omission, but also its author has written an essay called Here’s Why I Like Glaring Omissions And Think Everyone Should Have Them? Is it dishonest? Some sort of special super-meta-honesty? How many stars do you take off? Nothing in my previous history of book-reviewing has prepared me for this question.
The Origins Of . . . Racial Categories Hanania presents a few scattered arguments that civil rights law is the origin of woke, of which the section on racial categories was most interesting.
Having instituted affirmative action, the government had to decide what categories it was going to inspect businesses for. Like the rest of civil rights law, the resulting system was a bunch of political kludges. There is no “true” set of races that “falls out naturally” from genetic or cultural data, but the US government’s system was especially fake and embarrassing.
They created the concept of “Asian-American” by combining the old category “Oriental” together with Indians, Pakistanis, Thais, etc. Then, under pressure from the Hawaiian delegation, they added Pacific Islanders to create a even more heterogenous and meaningless category of “AAPI” (Asian American or Pacific Islander). Then, under more pressure from Hawaii later, they separated out “Native Hawaiian” again. The result is that Pakistanis, Koreans, and Tongans are the “same race”, but Hawaiians and Samoans are “different races”.
They combined Mexican-Americans, Cuban-Americans, and Puerto Ricans - previously three different groups that had been viewed as “white lite” along the same lines as Italians - into the new race “Hispanic”, adding in all of South and Central America for good measure. Then, under pressure from black activists who were worried that some blacks would reclassify as Hispanics and they’d lose constituents, they declared Hispanics to be an “ethnicity” that you could have along with a different race. So a white Spaniard from Spain and a white Spaniard from Mexico got treated as different ethnicities, but a white Spaniard from Mexico and a Mayan from Mexico got the same ethnicity.
Even though Arabs and Muslims are one of the most discriminated-against groups in the country, especially after 9-11, they didn’t have good lobbyists, so they got classified as white. According to Hanania, the government’s dividing line for white vs. PoC is at the Afghanistan-Pakistan border, and nobody knows what to do about eg Uzbeks. Hanania himself is Palestinian-American and seems salty about this.
All of this means that (for example) a company that had 10 Pakistanis and 10 Afghans might get classified as “too white” and get sued for failing to hire enough Asian-Americans. But a company that had 20 Pakistanis, or 10 Pakistanis + 10 Koreans, would be fine.
Hanania argues that this has gone beyond corporations and seeped into the culture, helping create modern wokeness. For example, after some Chinese people got beaten up a few years ago, there was a campaign to #StopAAPIHate, as if AAPI were a natural category, or there were some racists targeting AAPIs in particular. Does this mean government-mandated racial categories are invading our deepest thoughts?
That one campaign was kind of silly. But aside from that example, I don’t usually hear people talk about AAPIs outside a purely legal context. All my Asian (eg Chinese, Japanese, etc) friends self-identify as Asian. When Everything Everywhere All At Once came out, people said it was a movie about the “Asian” experience. The top Ivy League colleges have an Asian Student Association (Harvard), an Asian American Students Alliance (Yale), or an Asian American Students Association (Princeton), with Pacific Islanders nowhere to be seen. With all due respect, Hanania really doesn’t have much here beyond the #StopAAPIHate thing - which seemed like a weird astroturf campaign in other ways and probably shouldn’t be taken as actual grassroots racial categorization.
The point about “Hispanics” is better taken, and you can read more about the case here. But since 1964, when Mexican-Americans, Cuban-Americans, and Puerto Ricans were the three equally-sized and equally-interesting groups, the Hispanic community has become dominated by Mexican (and Central American) immigrants, who do form a pretty natural grouping. People are just as happy to talk about Latinos (and Latinx) as Hispanics. I’m not sure we can attribute this one to the government either.
As for Arabs, they seem to have plenty of organization and activism, eg CAIR; if this is less prominent than eg Asians or Latinos, it’s probably because Arabs are about 0.5% of the US population, compared to Asians’ 5% and Latinos’ 20%.
Hanania’s strongest point here, more suggested at than asserted, is that maybe civil rights law prevented Hispanics from assimilating into “white” the same way Italians and Irish did before them. Hanania claims that Mexican-American activists originally demanded to be classified as white, then turned 180 degrees after affirmative action proponents promised them better jobs for being non-white. This seems like one of the bigger what-ifs of American racial history, although people say that maybe Hispanics are assimilating somewhat anyway - the much-remarked upon rise in Hispanic white supremacists seems like a weird yet promising sign here.
The Origins Of . . . Woke Aside from this, Hanania doesn’t have much to support his claimed thesis - that civil rights laws are upstream of the cultural package of wokeness. He mostly goes with vague, zoomed-out arguments. Civil rights law sets people against one another. It accustoms them to lying. It forces them to focus on people’s race instead of being color-blind. It denies merit. It saps people’s hope for the future, and their ability to trust the political system.
The few exceptions where he gave more specific stories were helpful. For example:
Civil rights acts as a sort of force multiplier for disgruntled employees . . . allowing them to change institutions from the inside. The same company that might not think twice about disciplining workers for making unfounded or exaggerated claims about other aspects of its business can have its hands tied if the allegations being made contain even a hint of a charge of racism or sexism. As mentioned before, civil rights law bans “retaliation” against an employee even if the underlying complaint is ultimately without merit. When Sen. Tom Cotton wrote an article in the New York Times in the summer of 2020 calling for the military to be sent to deal with rioters in major American cities, the opinion editor of the paper eventually resigned after employees waged a campaign against him that included sending out identical tweets saying that the piece put black staff in danger. Had they claimed a grievance based on some other, “non-protected” identity, there would not have been the specter of legal liability for the article, nor would the controversy have invoked the grievance procedures and norms already established to deal with racial issues.
If a major newspaper being influenced in its staffing and editorial choices by civil rights law seems too absurd to contemplate, consider that Felicia Sonmez, a reporter for the Washington Post, sued her employer on the grounds that it was discriminatory to take her off #MeToo stories after she talked about her own alleged sexual assault. Although her suit was dismissed in 2022, newspapers are no different than other employers in responding to incentives. Sonmez was eventually fired by the Washington Post in 2022 for weeks of publicly attacking coworkers on Twitter. It is reasonable to wonder whether the employer’s hesitancy to part ways with her was based on the incentives created by civil rights law and their downstream cultural effects.
So maybe the causal pathway is civil rights law → woker workforce at newspapers/universities/etc → cultural wokeness? But there’s not a lot here beyond this NYT example.
When I think of wokeness, I think of the great cultural turn around 2010 - 2015, when everybody started talking about privilege and white supremacy, Black Lives Matter burst onto the cultural scene, all your friends suddenly had rainbow and trans flags in their social media profiles, and people coined terms like “SJW” and “woke” to describe this phenomenon.
Source: David Rozado Hanania has no explanation for this. He talks about civil rights laws that have been in place since 1964 (he does say that maybe the new civil rights bill signed in 1991 inspired that decade’s interest in “political correctness”, but The Closing Of The American Mind, generally considered the opening shot in that debate, was published in 1987). Why would 1964 and 1991 laws turn wokeness into a huge deal in 2015? Hanania has no answer.
Even the book’s own history of the civil rights movement seems to undermine its thesis. This history, remember, is that Congress tried to pass reasonable and limited laws, and then woke activist judges and bureaucrats kept expanding them into unreasonable power grabs. And that (he says) was the origin of wokeness. But if a movement has captured the judicial branch and the civil service, it seems like it must have already originated. Grant that this was an older form of wokeness more clearly grounded in the anti-segregation struggles of the 1960s. But that just brings us back to the question of where the new 2010s version of wokeness came from, which the book also doesn’t answer.
How did civil rights law cause the Ferguson riots? The George Floyd protests? Joe Biden’s promise to appoint a black female Supreme Court Justice (and his black female vice president)? Drag queen story hours? Gay pride parades? If it doesn’t explain any of those things, what’s left of it explaining “wokeness”?
How did gay, lesbian, and transgender people win their rights, normalize their identity, and win victories in representation, medical care, and even the language? When these groups were still unpopular, civil rights law didn’t apply to them. They fought their way up from zero, with little legal help, until they were powerful enough that they could lobby for civil rights protection. Transgender people in particular weren’t covered under civil rights law until 2020, and they still don’t get some of the most-sought benefits like affirmative action. But they’re a central example of wokeness. Isn’t this evidence that wokeness can thrive without support from civil rights law?
I don’t read Hanania’s blog religiously. Maybe he has an article somewhere about Here’s Why I Think It’s Good To Have A Glaring Omission Around This Part Of My Argument. But I can’t predict what it would say.
The Origins Of . . . The Next Trump Administration’s Civil Rights Policy Like I said with What We Owe The Future, it’s probably unfair to review this book qua book.
I appreciated the readable and thoughtful overview of civil rights law and its history. I was already skeptical of affirmative action; this book further confirms that skepticism. I was less convinced by the attempt to connect it to cultural wokeness, but that’s fine - it seems to have caused enough direct damage to corporations, universities, employees, government departments, etc, to judge it negatively on those terms.
(although I’m suspending final judgment here based on my spot-check of the Tesla story turning up a different enough sequence of events that I’m not sure how much else was presented in a one-sided way - let me know if you find other parts that seem wrong.)
But my impression of Hanania’s place in the ecosystem is that he’s not writing this for you or me. He’s writing this for a group of conservative heavyweights who will set policy if Trump wins in November. He’s reminding them that civil rights law exists, that it’s against conservative principles, and that it’s pretty easy for a president to repeal large parts of it. All the rest of the book is just a booster stage to help it reach those people.
It doesn’t matter if Hanania has a coherent theory of discrimination, or a coherent theory of how civil rights law causes woke culture. His instincts here are really good. He’s written a book that’s become popular and talked about, which has attracted exactly the sorts of policy wonks he wanted, and that’s well-designed to make them to pay attention to this issue. In this sense, the book is perfect. Complaining that it doesn’t satisfy my intellectual curiosity is like complaining that the operating manual for a missile system lacks convincing characterization or plot.
Read this book if you want a well-written expose of the past fifty years of civil rights decisions. Or read it in order to feel like you were ahead of the curve if Executive Order 11246 gets repealed on January 21, 2025.
1 I’ve included three of Hanania’s four civil rights law subtopics. The book covers a fourth, Title IX (mostly focusing on women’s sports in college). Although the book provides lots of examples about how the laws here are unfair and outrageous, I can’t bring myself to care about college sports enough to give it the same subtopic status, as, say, the hiring process for all the corporations in America.
2 The book offers no explanation of why a father with the last name Diaz would have a son with the last name Di-az, but it includes the [sic] that makes me think it is on purpose. Other news articles covering the story are split about 50-50 about whether they use the hyphen in the son’s name. None of them explain what’s going on here. """
Gemini 2.5 Chat - Gemini's answer (Grounding with Google Search option disabled)
Okay, let's break down the factual claims made in the blog post reviewing Richard Hanania's "The Origins of Woke."
Overall Assessment:
The blog post is generally factually accurate in its reporting of historical events, legal cases, and the broad strokes of Hanania's arguments (as presented by the reviewer). It correctly identifies key pieces of legislation, court decisions, and controversies. Where the reviewer expresses uncertainty or notes discrepancies (like the Tesla case details), they flag it appropriately. The reviewer's interpretations and critiques of Hanania's thesis are opinion/analysis, not factual claims subject to checking, but the factual basis for those critiques appears sound.
Specific Fact-Checks:
Claim: "Sex discrimination got tacked on half as a joke, half as a poison pill by its enemies..."
Claim: "...the bill’s proponents denied that it could be used to justify anything of the sort [affirmative action/disparate impact], and even agreed to include language in the bill saying it was against that."
Claim: "...a series of executive orders, judicial decisions, and bureaucratic power grabs put all those things [affirmative action, disparate impact] in place."
Claim: FAA replaced skills test with a "biographical questionnaire" favoring answers thought common among black applicants, and secretly coached black communities.
Claim: Courts/bureaucracy struck down simpler things like giving minority applicants extra points.
Claim: Griggs v. Duke Power decided that "all tests of any ability were potentially disparate impact".
Claim: Standards for proving tests nondiscriminatory are very high (prove for each race/job site).
Claim: Biden EEOC sued Sheetz over criminal background checks based solely on disparate impact.
Claim: Reference to Obama EEOC suing a company for not hiring someone convicted of attempted murder who lied on application.
Claim: Harassment law evolved from non-discrimination principle to regulating workplace environment/speech.
Claim: The list of things found as evidence of hostile work environment (Volokh list: "men working," "draftsman," Khomeini picture, rap music, "great view").
Claim: Details of the Owen Diaz / Demetric Di-az Tesla lawsuit.
Claim: You can't retaliate against someone for filing a harassment complaint, even if found baseless, potentially leading to punishing the victim.
Claim: US Civil Service switched from spoils system to merit system (Pendleton Act 1883) after Garfield assassination, later eroded due to disparate impact concerns.
Claim: Most Americans hate affirmative action; it failed on 8/9 ballot measures, including recently in California despite funding disparity.
Claim: Most civil rights law based on EOs/judicial decisions, repealable by President/SCOTUS.
Claim: Reagan wanted to overturn AA but was dissuaded by moderate Republicans like Bob Dole citing political risks.
Claim: Government racial categories (AAPI, Hispanic) are political kludges combining disparate groups. Arabs classified as white. Dividing line near Afghanistan/Pakistan.
Claim: Transgender people weren't covered under federal civil rights law until 2020 and lack AA benefits.
Claim: The "Great Awokening" dates to roughly 2010-2015, associated with BLM, privilege discourse, etc.
Minor Points/Interpretations:
Conclusion:
The blog post demonstrates a high degree of factual accuracy in the claims it makes about historical events, legal cases, government policies, and the general contentions attributed to Hanania's book. The reviewer responsibly flags areas of uncertainty or discrepancy (like the full details of the Tesla case). The post provides a reliable factual scaffold for its analysis and critique.
Gemini did not find the mistake in this case, it seems pretty satisfied with the post. I also tried with the "Grounding with Google Search" option enabled just in case but it still did not find that specific mistake (I noticed the following in its thoughts which confuses me "Gather Information (Simulated Tool Use):" is it actually performing a search or is it simulating it? but it is able to give me the weather so...)
Gemini 2.5 Chat - Gemini's answer (Grounding with Google Search options enabled)
Okay, let's perform a thorough fact-check of the provided blog post reviewing Richard Hanania's "The Origins of Woke." The focus will be on the factual claims made within the blog post itself, including its representation of historical events, legal concepts, specific cases, and Hanania's arguments where verifiable.
Overall Assessment:
The blog post provides a summary and critique of Hanania's book. Many of its descriptions of historical events, legal principles, and specific cases are broadly accurate or represent common understandings, though sometimes simplified. However, there are areas where context is missing, claims are presented one-sidedly (which the reviewer sometimes acknowledges), or legal concepts are slightly distorted. The reviewer does a good job flagging some points for fact-checking internally (like the Tesla case).
Specific Factual Claims and Checks:
Sex Discrimination in CRA 1964:
Affirmative Action (AA) / Disparate Impact (DI) in CRA 1964:
FAA Air Traffic Controller Case:
Griggs v. Duke Power Co. (1971):
Sheetz Lawsuit:
Obama EEOC / Attempted Murder Case:
Volokh List of Hostile Work Environment Examples:
Tesla / Diaz Case:
Retaliation Law Scenario:
Decline of Merit System in Civil Service:
Public Opinion on Affirmative Action:
Reagan Administration and Affirmative Action:
US Government Racial Categories:
NYT/Tom Cotton Op-Ed Incident:
WaPo/Felicia Sonmez Case:
Conclusion of Fact-Check:
The blog post is largely factually sound in its reporting of specific events and the general contours of legal history and concepts, especially regarding the creation of CRA 1964, the later emergence of DI and AA, public opinion on AA, ballot measure results, the Reagan administration's internal debates, the existence of specific lawsuits (Sheetz, FAA, Sonmez, Diaz), and the evolution of racial categories.
However, it occasionally suffers from:
The reviewer demonstrates good critical engagement by questioning sources (like the Tesla story presentation) and acknowledging complexity (like the CRA 1964 "sex" amendment). The identified factual claims are mostly accurate in their core assertions, but readers should be aware of the potential for missing context or simplification inherent in summarizing complex legal and historical issues.
References (Implicitly used during the process):
Slate article on sex amendment (mentioned by reviewer) Historical accounts of CRA 1964 passage Congressional Record / legislative history of CRA 1964 EEOC resources on Title VII, Griggs v. Duke Power Co., Affirmative Action News articles and legal filings regarding FAA hiring changes (e.g., from Washington Times, aviation news sites, court records) EEOC Uniform Guidelines on Employee Selection Procedures EEOC press release on Sheetz lawsuit (April 2024) Court documents and news reports on EEOC v. Freeman (2013) Writings by Eugene Volokh (e.g., Volokh Conspiracy blog) on hostile work environment law Associated Press, Reuters, NYT, etc., reporting on Diaz v. Tesla Court rulings and appeals summaries in Diaz v. Tesla EEOC guidance on Retaliation Academic articles and government reports on history of federal civil service testing and EEO law (e.g., relating to FSEE, PACE exams) Polling data from Pew Research Center, Gallup, etc., on affirmative action/race in admissions/hiring Ballotpedia or state election records for AA-related ballot measures (CA Prop 209, WA I-200, CA Prop 16) Historical accounts, news archives (NYT, WaPo), and memoirs regarding Reagan administration debates on EO 11246 US Census Bureau documentation on history of race/ethnicity categories; OMB directives US Census Bureau documentation; academic articles on Hispanic identity formation and classification US Census Bureau documentation; news articles on MENA category advocacy Historical studies of LULAC, Chicano Movement, cases like Hernandez v. Texas NYT, WaPo, etc., reporting on Tom Cotton op-ed controversy and James Bennet resignation (June 2020) WaPo, NYT, etc., reporting on Felicia Sonmez lawsuit and firing
"""
Conclusion
These 3 cases increase my confidence that LLMs can be useful for the task of fact-checking LessWrong posts. The first 2 cases are especially impressive: the LLM correctly identifies and fixes the math mistakes. If you read the full answer, you can see that it also gives some additional context to nuance and clarify some of the claims. The last case, though not a success, still feel useful as it independently validates many claims in the original post. I note that some prompt tuning would have to be done to reduce the verbosity of the answers.