American Exceptionalism and the Curse of Dimensionality
The United States is frequently described as an international outlier on human rights. Numerous books and articles have been written bemoaning American exceptionalism. But is it true?
I have been thinking a fair bit about comparative law recently and it strikes me that the only valid way to describe the United States as a constitutional outlier is to do a truly thorough international comparison. It requires that we not only locate our country in relation to other countries, but also that we locate other countries in relationship to each other. Describing the United States as an “outlier” simply assumes that other nations are clustered together, which may or may not be the case. The United States may be different from other countries, but each country may also be different from the other.
Take free speech as an example. Harvard’s Frederick Schauer has recently argued that “the American First Amendment remains a recalcitrant outlier to a growing international understanding of what the freedom of expression entails.” Really? How does he know this? By comparing our defamation and hate speech jurisprudence with a handful of other countries?
If one were to thoroughly compare freedom of expression in different countries, how would one go about that task? First, one would need to choose the countries for comparison. Comparing only a handful of countries would be random and prone to selection bias. Choosing only peer nations, one could compare all modern capitalist democracies, say, thirty countries that are OECD members as a proxy for that distinction. If one just compared the United States to the other OECD countries, this requires only 29 country comparisons. But if you compare each country to the other, it requires 435 comparisons ((30 x 29)/2 = 435). Such cross-country comparisons offer a much deeper understanding of the spatial location of the United States relative to other countries.
But this comparison of OECD countries assumes that free speech is one dimensional. It is not. There are numerous free speech variants, including hate speech, defamation, political speech, pornography, and commercial speech, to name but a few. Is the United States an outlier with all peer nations on all speech variants? It would seem rather doubtful. So if one were to undertake the OECD comparison above using just five free speech variants, we no longer have 435 comparisons, but 2,175 comparisons. ((30 x 29)/2 x 5 = 2,175).
But even this comparison assumes that we are only looking at one contextual factor, say judicial precedent. Doesn’t text, structure, and culture impact the scope of speech guarantees? Doesn’t it matter that the Canadian Constitution expressly subordinates other civil liberties, including speech, to the principle of equality? Doesn’t it matter that Japanese culture places a tremendous premium on honor and reputation? Isn’t it relevant that some fragile democracies are less capable of tolerating dangerous political speech because of proportional representation?
If one deepens the analysis by considering common contextual factors such as constitutional text, structure, history, judicial precedent, and national experience, we are adding five contextual factors into the mix. Now if we compare five speech variants in 30 OECD countries using five contextual factors it requires not 435 comparisons, or even 2,175 comparisons, but rather 10,875 comparisons. ((30 x 29)/2 x 5 x 5= 10,875).
What I am suggesting is that comparative constitutional analysis suffers from the curse of dimensionality. A comparison of hate speech in two countries (e.g., the United States and Germany) using one contextual factor (e.g., constitutional text) is easy. A comparison of five speech variants in 30 OECD countries using five contextual factors is exceedingly difficult. But only such comparisons undertaken in high dimensional space (i.e., analyzing multiple countries using numerous speech variants based on several contextual factors) can reveal whether or not there are free speech clusters and outliers as is often asserted.
This curse of dimensionality raises serious questions about the feasibility of comparative constitutional anlaysis. A comparison of a few countries is prone to error and selection bias. A comparison of numerous speech variants in many countries using multiple contextual factors offers a far more accurate picture of free speech guarantees throughout the world. But as a practical matter, almost no one uses a large-data-set, high-dimensional methodology. As a result, to overcome these conceptual difficulties scholars and judges typically opt for narrow, shallow, and simplistic comparisons, and make hollow pronouncements about American exceptionalism.