Contents:
Consider the following: A state education committee, like so many others before it, has decided to monitor teachers by having their students take an annual test. Teachers will be evaluated based on how well their students do on this test, receiving bonuses or penalties accordingly. We know what happens next: students are drilled on the exact topics that will be on the test, taught to write in the same way the rubric expects, made to perform to specifications. Instead of measuring the quality of teachers as teachers, the test now measures their quality as drill instructors. This gaming of social measurement processes by those with vested interests in the decisions those measurements inform — in this case, teachers who understandably want to be rewarded rather than penalized — is a ubiquitous failure mode commonly known as Campbell's Law.
Here, I’m going to discuss what seems to be an analogous phenomenon, which I understand as a linguistic version of Campbell's Law; employing the classic rationalist trick of giving obnoxious names to made-up things, I call this phenomenon “Campbell’s Hammer”. Before I can talk about it, I’ll introduce a couple of ideas pertaining to the analysis and application of traits. Let’s start with a question: How do you recognize the trait of being fascist?
Ideally, one would ascertain the T-ness of a given entity by starting off with a proper understanding of what T is. Since fascism (of a person, government, policy, etc.) is defined by analogy to a few primary exemplars, this understanding is built as follows:
Identify the primary exemplars of fascism. As the nature, consequences, and more generally the identity of fascism was discovered through practice in the 30s and 40s, rather than following purely from theory, analysis by way of comparison to known exemplars is our only path forward. Hence, we might single out Hitler's Germany, Mussolini's Italy, and Franco's Spain for dissection.
Identify the key properties which these exemplars share in common, which uniquely distinguish them from non-exemplars, and which lie in the general linguistic domain we consider the word as having. Namely, we consider fascism to be an ideology, so its properties must be ideological; if these three countries were also uniquely characterized by having a certain combination of inflation rate and GDP, we wouldn’t count that as being fascist because these aren’t properties of ideologies — though it would definitely be curious, and worth looking into as a possible empirical correlate of fascism. Our question, then, is this: what ideological properties did these states share that distinguished them from other states throughout history?
As these properties may generally include responses to situations indicative of a certain ideology, which responses can arise by sheer coincidence in other systems which are decidedly non-fascist (as we will see), we can’t simply list a bunch of properties and unambiguously find fascism somewhere in this list. As such, we have to actually think, determining what it is about these properties makes them fascistic. Is it their context? motivation? combination? An analysis that does not consider these factors is less than useless. Lincoln, for instance, flagrantly violated the Constitution to expand his power and defeat his enemies; it is context which separates this case from the exemplars of fascism.
A note: If you want to say that fascism is immoral, or connect something to fascism to show its immorality, it’s useful at this point to identify those particular properties by virtue of which fascism is immoral. This isn’t a matter of epistemic supererogation, but one of correctness — if you want to go on to say that some thing is bad because it’s fascistic, you’ll need to show some relation between that thing and these particular immoral traits; otherwise, you’ll be able to make no connection to actual morality.
Now, if you want to ascertain the extent to which some thing, call it X, is fascistic, you must analyze the extent to which X displays the above identified traits of fascism, and whether the desiderata of correct context, motivation, etc. are satisfied. The moral analysis of fascism noted above can then be used to figure out the extent to which X is bad due to this fascist-ness.
This kind of analysis doesn’t just apply in politics: if you’re asked to justify your calling some food “fast food”, some philosophical work “post-structuralist”, or mathematical result “analytic”, you might describe some general traits of fast food, of the post-structuralist school of philosophy, or of the mathematical field of analysis which appear in the thing you’re talking about.
Suppose you and I are driving, and we get burgers at some restaurant neither of us have ever heard of; ten minutes later, when we get into an argument about whether it’s a “fast food” restaurant or not, I might end up describing the restaurant as being a fast food restaurant because its food is greasy, calorie-dense, and we got it in 90 seconds at the drive-through. If asked to justify the relevance of these properties, I might pick out a few exemplars which are distinguished by these traits: most of the major chains which are commonly accepted as “fast food” serve food with exactly these properties, and most non-fast food restaurants don’t serve food with these properties. The strategy is the same — we appeal to the archetypal properties of fast food that these burgers demonstrate, these properties being chosen so as to characterize most restaurants we can safely agree to call “fast food”.
In general, any trait which is defined by analogy to a set of commonly accepted exemplars can be treated in this way. I call such traits analogical traits. It’s often useful to consider the size of the set of exemplars of any given trait, which I think of as a sort of “analogical bandwidth”. For instance, fascism has a low analogical bandwidth — a handful of examples in the middle of the 20th century — whereas fast food has a high analogical bandwidth, since there are so many examples of restaurant chains that we can safely call fast food.
Analogical traits are almost never categorical in nature — unless you’re talking about one of their exemplars, you can’t say that some object either does or doesn’t possess an analogical trait. The possession of said trait, instead of being binary yes-or-no, lies along a continuous spectrum of intensity. Some thing may be far more fascistic than other, similar things, but it will not be “fascist, period” unless it actually is an exemplar of fascism. In other words, analogical traits do not naturally merit categorical application. Usually, when they are applied in such a manner, it’s to denote a particularly high intensity, as with my calling the restaurant above a “fast food” restaurant.
Note: This is a subtle point, dependent on the observation that there are many different ways to use a word. We’re using the word “fascist” here as a trait — in its separate use as a descriptor of identity, some millennial can be “fascist” if they self-identify as such, but this fascism-as-identity has no necessary relation to fascism-as-trait. This is an important distinction, as is the distinction between usage of the word as a trait versus its usage as a pejorative, mental shortcut, rhetorical label, and so on, each of which again has no necessary relation to the actual trait.
However, if you wish to call something fascistic in a manner that relates to the commonly accepted notion (i.e., you’re not doing something stupid like defining “fascist” to mean “fluffy and cute thing”, such that kittens are now fascists), it is epistemically incumbent upon you to make sure that your usage of it is proper, as explained above.
There are many ways in which the proper analysis of some trait can deviate from the analogical analysis outlined above, i.e. can differ from a comparison to those contextually relevant properties distinguishing the exemplars of the trait.
The first way is through an empirically observed clustering of the trait’s properties. For instance, when you want to say that some music belongs to the genre of death metal, the analogical perspective would have you figure out how it shares those distinguishing features of “known” death metal bands. Death metal, however, is pretty much instantly identifiable even if you can’t name a single death metal band — it’s impossible to ignore the combination of dark, grim album art with $\mathfrak{Illegible}$ $\mathfrak{Blackletter}$, power chords so distorted they may as well be white noise (e.g.), and shrill screaming designed to maximize your listening displeasure.
In other words, the properties that distinguish death metal are tightly clustered, such that anything which belongs to that genre clearly belongs to that genre. Of course, there will always be edge cases, but tight clustering means these edge cases will be very rare; in the vast majority of cases, analogical reasoning will not be required. Analogical reasoning still works in such cases, but it’s more tedious and less reliable than simply taking a look at the thing and identifying its relevance to the trait.