Scott Alexander’s Noncentral Fallacy, dubbed “the worst argument in the world”, is a classic example of manipulative rhetoric. The basic structure of this fallacy, as Scott describes it, is to apply technically correct labeling in order to conjure up misleading connotations — say by labeling MLK a “criminal” in order to obliquely undermine his legacy. Scott is identifying an important and pervasive dynamic, but he’s misidentifying why it’s a problem because the concept of “centrality” is both incoherent and irrelevant. Rather, the real issue is how debates over categorizations and labels get utilized as a fallacious shortcut to avoid having to debate some Other Thing. Let’s call it the “Sticker Shortcut Fallacy” instead.
What even is centrality?
Honed down to its most discrete point in semantic space, the barest definition of criminal is “anyone who has ever broken any law”. While the word certainly shoulders a constellation of emotionally salient connotations, these are heavily susceptible to context and priming, and very often vehemently disagreed with. Which archetypical avatar of criminality gets conjured up in your mind will depend on your particular background, the specific narrative presented to you, as well as whatever sympathies you may hold.
Al Capone might be the prototypical criminal in people’s minds, but on paper, he’s actually just a tax evader. Robin Hood literally stole money from people, but he gave it to the poor, so his legacy tends to be that of a folk hero rather than a felon. Based on the severity of punishment, nothing compares to the legally sanctioned torturous spectacle of drawing and quartering those convicted of high treason, but whether such a person is a quote-unquote criminal depends on which side you’re on. The Founding Fathers were unambiguous seditious conspirators against the greatest empire of their time, yet no one — except perhaps some nostalgic British monarchists — thinks of them as criminals for this.[1] How exactly do you adjudicate which one is a “central” archetype of criminal, and who decides?
The disagreement over centrality applies to other examples cited by Scott, especially when you jettison the semantic baggage and start with a blank slate. The core essence of theft might be “taking other people’s property without their consent” but depending on the context, different examples will herald primacy. For instance, if you’re a record company, it might be illegal downloads. If you’re an overlooked office worker, it might be someone else taking credit for your achievements. If you’re a 19th century anarchist, it might be the entire concept of land property. If you’re a white college student afraid of putting on a sombrero, it might be cultural appropriation. Who decides which example should take primacy?
This isn’t a navel-gazing exercise when the answer to the question controls the boundaries of fallacious reasoning, but this endeavor dredges up the same tediously familiar and intractable disputes over whether cereal is a soup, or if liquids can themselves be “wet”. Do we look up the dictionary? Do we consult linguists? Do we conduct a general population survey? Of whom? Orchestrating a consensus on simple definitions is already difficult enough, but adjudicating agreement on which archetypes should be central is beyond hopeless.
Consider Scott’s description of the Noncentral Fallacy again:
I declare the Worst Argument In The World to be this: “X is in a category whose archetypal member gives us a certain emotional reaction. Therefore, we should apply that emotional reaction to X, even though it is not a central category member.”
No one who engages in the Worst Argument will ever ever ever concede the bolded clause, because doing so completely nullifies their attempted sophistry! Since no one will hand you the refutation key so transparently, this will inevitably invite all sorts of question begging: “Your argument is fallacious because X is not a central category member. And it’s not a central category member because…I say so?”
More central is not less fallacious
I’m going to assume the above away and just pluck central archetypes based on colloquial understanding to show that the Noncentral Fallacy remains the wrong prism. The barest, narrowest, and most widely accepted understanding of murder involves the basic ingredients of 1) the unlawful 2) killing 3) of a human being 4) by another.[2]
With capital punishment, the only missing ingredient is whether the killing is unlawful (or illegal, illicit, or somehow unauthorized) which is a very peculiar state of affairs for a state-sanctioned execution. Accordingly, anyone insisting that capital punishment is murder is transparently agitating for the unlawful clause to be swallowed whole without resistance. They’re hoping that slapping the murder sticker on the conduct provides them with a shortcut to avoid direct, substantive argumentation. Hence, the Sticker Shortcut Fallacy.
The essential conditions for this type of fallacy are:
The premise you’re ultimately after is either very difficult to substantiate or is highly contested. (e.g. capital punishment should be unlawful)
There’s another adjacent premise that is much easier to establish or is otherwise pedestrian. (e.g. murder should be unlawful)
Both premises share an overlapping semantic cluster, a set of related concepts or connotations, that can be used to blur the lines between the two.
Since everyone loves illustrations, this is one way to visualize it:
The only way to accept the murder sticker is by adopting the unlawful premise and abandoning the lawful premise. In the case of “abortion is murder”, this cheap trick performs a lot of heavy lifting because insisting on the specific vocabulary is an attempt to surreptitiously assert “fetuses should be considered human beings” and “the killing of fetuses should be unlawful”, all without having to do the very hard work of substantiating either premise, or maybe just hoping you won’t notice.[3]
This demonstrates that the Sticker Shortcut Fallacy is fundamentally a snuck premises or presuppositions fallacy, one that is unlikely to happen inadvertently given the level of intentionality required. What might initially seem like a harmless categorization game of ‘which dishes go into which cupboard?’ is actually a subtle way of shifting someone’s stance by altering the foundation of their argument.
Scott Alexander was very close to properly identifying the Sticker Shortcut Fallacy in an earlier post called Diseased thinking: dissolving questions about disease:
People commonly debate whether social and mental conditions are real diseases. This masquerades as a medical question, but its implications are mainly social and ethical. We use the concept of disease to decide who gets sympathy, who gets blame, and who gets treatment.
This is exactly right. A premise such as “individuals who are addicted to drugs deserve our sympathy and should receive medical treatment” is bound to be hotly contested and difficult to substantiate, whereas if you just slap the disease sticker on addiction then you can assume away the hard work. The reason this fallacy is so pervasive and noxious is because its distraction potency is extremely high. It is very easy to get hoodwinked and dive right into debating that “capital punishment is the good kind of murder”, a quagmire no one can cleanly escape from. If someone tacitly accepts the sticker, there’s a chance they won’t realize they’ve left the gate open for the presuppositions to slink in, hopefully not until it’s too late.
The only effective response when someone is deadset on a categorization question is to directly ask why the categorization matters. What information or insight, if any, is gleaned from answering the question? There are actually valid purposes to these types of semantic arguments, perhaps when examining the extent language influences its speakers’ worldview, or when someone wants to draw attention to similarities between categories that may otherwise get drowned out.
But if someone is using the Sticker Shortcut Fallacy to mislead, to sleepwalk others into compliance, their efforts would be very effectively neutered because they would now required to expose the implied connotations they were hoping others would make by adopting the label. If they resist answering such a simple question, that’s also its own kind of answer.
Many thanks to Trembling Mad for helping me crystalize this concept.
- ^
Ironically, they ensured that the only crime specifically mentioned in the Constitution was treason. Talk about pulling up the drawbridge behind you.
- ^
This is why when a lion eats someone, we don’t use murder even though a human was definitely killed. Do lions even respect the law? A question for another time perhaps.
- ^
Its sister slogan “meat is murder” performs similar presupposition heavy lifting on whether animals should be considered to have the same moral weight as humans, and whether their killing should be unlawful.
In my view, part of what makes the non-central argument a fallacy is the ad hoc use of the ‘overly restrictive definition’.
Whoever argues that “MLK is a criminal” with the intent of instilling the negative connotation of the term is unlikely to apply the same standard everywhere.
One in that case could reply that anyone who ever opposed any non-democratic regime and was found guilty of sedition/instigation/etc.… is/was also a criminal.
If the proponent of the argument disagrees, we fall into a contradiction. If in favour of the extended use, the original concept—criminal—is stripped off its stereotypical baggage.
If the problem is ad hoc application, then it doesn’t matter if the archetype is “central” or not, no?
My objection to “MLK is a criminal” is that it has to make too many unannounced jumps to get to its conclusion. The principle I can glean from this type of denouncement is something along the lines of “[Criminals] should not be honored by statutes” but whether or not this is a good argument depends entirely on what definition of [criminal] we’re using. If we adopt a barebones definition of the word, we’d end up with something like “[anyone who has ever broken any law] should not be honored by statues” which immediately is exposed as unconvincing, which is why they have to hide behind the connotation. That’s why I argue the problem isn’t whether it’s “central” but rather using labels as reasoning shortcuts.
Upvoted for the finalmost sentence of your post; thank you so much.
This is an indictment of the human species, if this purported “unlikelihood” is true. Maybe you should not underestimate the likelihood that your interlocutors have a serious deep resentment of unlawful behavior, however alien this might be to you. Maybe part of their fundamental self-narrative includes the unforgivable harms consistently caused to them by crimes which were superficially dismissed as mild by others. They may think “If this is a mild (read: non-central) crime, I don’t want to know what the serious (read: central) ones are.” Maybe they feel they have no choice but to become a total “I’ll end it forever if it’s the last thing I do”-level enemy with criminality in all forms, as a precaution.
If humanity is willing to coexist with anything, well, imagine the worst possible thing. Imagine something worse than that. Worse in ways you didn’t even realize things could be worse by. Recursive worseness. Explosive worseness via combination of worseness-multipliers. Worseness-multipliers that might seem like normally good things if they weren’t being used by your imagination for the explicit purpose of making things worse. (Like hope, for example.) That is a thing which counts as a member of the “anything” set which humanity would be willing to coexist with, in the world where humanity would coexist with anything.
Unconditional coexistence is not safe for humans. To refuse coexistence with something that is evil in letter and spirit, on the outside and on the inside, you must have a clear sense of that thing no matter what are the stereotypes — the consensus reality — about its symbolic representation.
The only reason I could think of that this would be the “worst argument in the world” is because it strongly indicates low-level thinkers (e.g. low decouplers).
An actual “worst argument in the world” would be whatever maximizes the gap between people’s models and accurate models.
For a more recent example, I have noticed that people are currently doing this by calling Trump a “convicted felon”, as if we knew more about Trump after the verdict than we did before. (I have no vote in that election, and I’m not sure I’d bother even if I did.)
There are several non-fallacious reasons to emphasize Trump’s status as a convicted felon:
For anybody who’s learning about it for the first time and hasn’t followed every detail of the trial, it shows that a new category of people have reviewed the evidence and arguments in detail and decided he is guilty: a jury. That was far from a foregone conclusion.
For those who have heard of the trial outcome, propagating new information through a person’s belief structure and social network requires repetition and emphasis—just hearing the words ‘Trump: convicted felon’ one time isn’t enough.
It points out a deep contradiction with the family values and law and order rhetoric Republicans have traditionally used to further their political goals.
Calling MLK a ‘criminal’ doesn’t evoke the same whiff of hypocrisy and contradiction, because civil rights activists have rarely if ever based their moral argument on ‘law and order’ rhetoric. Indeed, most Black Americans agree “the criminal justice system was designed to hold Black people back.” Pointing out MLK was a criminal, given the nature of the “crime,” just lends further support to their argument.
I think this post might be a good illustration of the sticker shortcut fallacy I’m describing. Instead of directly describing the information you want to impart, you’re instead relying upon the label dredging up enough ‘good enough’ connotations attached to it.
I think it’s non-fallacious to use language as a shorthand, the same way we say “do you want to play baseball?” rather than “do you want to play a bat-and-ball sport played between two teams of nine players each, taking turns batting and fielding?”
What information, specifically, do you believe “Trump: convicted felon” conveys except that “a jury reviewed evidence and were convinced that Trump committed a particular offense categorized under New York state law as a felony”? I mean this question very narrowly.
On this point, I concede your argument. To the extent anyone is operating at the “sticker” level (e.g. we don’t support law-breakers) then pointing out that their preferred candidate is a law-breaker is indeed a valid rebuttal. But if it’s deployed outside that narrow purpose, then it becomes fallacious.
I disagree. The label ‘dredges up’ (implies) a sound argument. One syllogism that might be implied by “Trump: convicted felon” is something like this:
This is a valid syllogism, though you may reject the premise. I don’t think it qualifies as deceptively bad. It could be false but popular, but that has to be argued.
I’m sorry, but this is exactly the fallacy I’m describing in my post. Sometimes the innocent is convicted, and sometimes the guilty is acquitted, which means the only thing that makes “convicted” true in all circumstances is “the legal system has deemed an individual guilty of the allegations”. Nothing more. Now, you may certainly make very plausible Bayesian predictions about the fact that someone has been convicted, but they will always be probabilistic rather than determinative.
Consider the hypothetical where Trump’s conviction gets overturned or vacated, maybe because of some procedural defect, what would change? For me, it wouldn’t change the fact that Trump constructed a convoluted scheme to pay hush money to a porn star he had an affair with in an attempt to hide this fact from the voting public. The only thing that would change from the conviction getting overturned is whether “the legal system has deemed an individual guilty of the allegations”.
I don’t believe that anyone actually holds the syllogism you describe, because a consistent application would mean that even folks like Nelson Mandela (convicted of sabotage and sentenced to life) would be unfit to serve as president. Instead, what I gather people are doing is some combination of the composition/division fallacies:
Indeed, it’s the same fallacy.
Edit: for those who disagree, can you explain why? I don’t believe that “Trump is a convicted felon” tells us anything new except that a court found him guilty of a felony. I’m not trying to be pedantic here, but so what? If people are actually interested in the connotation avalanche that is attached to that label (such as ‘bad person’, ‘dishonorable’, ‘malicious’, etc) then why not just use those direct descriptions instead?
Because those are unsupported claims about his character, while noting his conviction (particularly given that he was covering up an affair) is specific evidence of his bad character. Moreover, it is evidence of a particular way in which his character is bad—he is not only willing to have an affair, but he’s willing to break the law to hide it.
If I tell you X is a bad person, that tells you nothing except my opinion of them. If I say “they were recently convicted of a felony for falsifying business records covering up an affair,” you can judge for yourself whether or not you think this fact reflects on their character or is worthy of punishment (ie by denying them your vote for President).
This too is another example of the fallacy I’m describing. The fact that OJ Simpson was acquitted of a double homicide doesn’t change my mind that he did in fact kill two people, all it tells me is that the legal system did not find him guilty of the allegations. If someone started every conversation about OJ with “exonerated celebrity football player OJ Simpson”, it’s obvious what connotations they’re trying to convey without having to communicate them directly.
I have found that whenever I have a disagreement about category membership, it’s usually still pretty easy to define an ordinal scale between “clear member” and “clear nonmember” that gives clear agreement. The disagreements tend to be about how tightly to draw the boundary of when we accept the use of the category word. In that sense, centrality seems like a real thing that anyone not being actively deceptive can discuss and notice and come to agreement on.
Unfortunately society as a whole almost never does this except in the most contentious cases. Law often requires binary distinctions that assume words have usable strict definitions, even if that’s not how words work. But in case you were curious, in Fort Wayne, Indiana, burritos and tacos are definitively a kind of sandwich and the fate of a restaurant depended on deciding where to draw the “made to order sandwich” category boundary.
I disagree that there is such a thing as objective “centrality”, just as I disagree there is such a thing as objective definitions. All language is made-up, and it’s only useful to the extent others share your (arbitrarily designated) meaning or boundaries. There are scores of real-life examples that clearly illustrate this, such as the fact that the word for ‘sake’ refers to all alcoholic drinks in Japanese, or how some languages make a distinction between maternal/paternal aunt/uncle, or how Russian treats light blue and dark blue as separate colors, etc.
Even setting that aside, the only insight you glean from determining whether a member is central to a category or not is...whether a member is central to a category or not. If you use category membership itself to glean any other information about a member, this is exactly the sticker shortcut fallacy I’m describing.
Statutory and other legal interpretation is exempt from my critique here, because the meaning of a word is very often explicitly spelled out in legislation (hence why legalese is so tedious to read). When the meaning is ambiguous, judges resort to specific canons of interpretations (such as legislative intent, ordinary meaning, historical meaning, rule of lenity, etc.) that are based in legal precedent.
In the space of all possible languages and words, sure, but real languages are (or at least always have been to date on Earth) created and used by humans, with human minds/tendencies. Despite massive cultural/historical/individual variation, that still moves us quite a ways away from truly, fully arbitrary word meanings. I see two counterpoints to your observation of how different languages split and lump concepts:
1) Languages tend to cluster in the ways they do or don’t split. Number of basic color words varies but the order in which they appear as number increases is mostly constant. Number of relationship words varies but as it increases the splitting happens along the axes of gender and degree-of-relatedness and which line we’re related through. Most times a language has a hard-to-translate word, it can still be translated as a compound word, and if not, it can usually be roughly defined in a sentence or two, and work as a loanword from then on.
2) Languages tend to enable new word formation by native speakers with relatively easy agreement on their meaning. English has (had?) no single word meaning “niece and nephew” but many people hear or read the word “nibling” for the first time and figure out the intended meaning with no explanation whatsoever. There’s also things like the bouba/kiki effect, which points towards how some of the ways sound ties to meaning are related to synesthesia and metaphor. On the other hand jargon (in any field including law) - where words are actually defined by inhuman category boundaries feels much less natural to most people. From an objective perspective the jargon words are least arbitrary. However, the ease of learning natural language words suggests they’re not so much arbitrary as they are running along grooves natural to human thought instead of natural to the physical, non-human world.