21 January 2024

Nonconsensus Link Sausage Platter

This platter of link sausages has been thoroughly p-hacked because I'm making an ideological point.

  • We just can't seem to get away from the malign influence of eighteenth-century English privilege as expressed in not just the common law, but in particular aspects of it. Despite the US having the First Amendment — which, for all of its problems, seems superior to the conceivable alternatives, and certainly to those actually implemented anywhere — the English system still acts to suppress speech… even from Americans that would be allowable, and even encouraged, Over Here (notwithstanding ex post amelioration of adverse foreign judgments concerning that very type of speech).

    Even former colonies overtly hostile to England are in on it. The upper classes and ultra-nouveau-rich are being quite effective at maintaining parts of the colonial systems that benefit them. What that implies about how much they really have separated themselves from the colonial system is rather disturbing.

  • Then there's the eternal question faced by dissenters within a government: How many wrongs make it right? More to the point, how many of those convinced that the wrongs they are engaging in create an on-balance right are prepared to pay the price… especially when that price rises to the level of lives, fortunes, and sacred honor? (Consideration of the current unpleasantness in Gaza in this context is for another time and forum.)
  • Not this (dysfunctional) forum, however. With all due respect to Mr Wolfe, he has missed the forest of "who has a credible professional platform concerning interpreting not just the Constitution, but all law?" for the trees of "who should sit on this particular court?" This dataset bias problem is restriction of the pool of future Supreme Court justices, and most other judges, by not later than the third undergraduate year… when, prior to the 26th Amendment, they weren't even eligible to vote. This is a restriction in practice and effect through the law-school admission process: The vast majority of judges (and all potential Supreme Court justices nominated since the Depression-era "switch in time that by effect, if not necessarily by intent, changed a fundamental assumption in Supreme Court litigation) went straight from their undergraduate educations to law school — and law-school admission decisions (and financial-aid decisions), especially since draft deferment to keep away from Vietnam became a "thing," very mechanically sort opportunities on undergraduate GPA as of the fall-of-the-senior-year application period. Even the NFL does a better job of not relying solely upon seventeen-year-old five-star recruits to college teams for its future on-field leaders than does the law.
  • All of the above feed into the impulse to make all decisions seem "objective," often through reasoning based on purported algorithms. There's a fundamental problem with doing so: Distorted datasets. Consider the coffee-shop design example in that article in the context of who is being excluded — by design — from the dataset upon which the algoritm operates. Obviously, non-coffee-drinkers (or, at least, those who don't drink tea either), and specifically non-Western-context coffee drinkers. There are also economic, health, and temporal biases in play — not just the excrutiatingly obvious "poverty line," but "wheelchair and other limited-mobility access" and "productive remote worker in another time zone" subpopulations that don't get into the dataset even as outliers.

    Why does this matter right now? Dewey Defeats Truman points the way — and so does my mail. Prerestriction of membership in datasets on the basis of presumed correlations, epitomized by my inability for four decades to get the various Heffalump fundraisers to remove me from their mailing lists as a "probable donor" because I was a commissioned officer, results in gathering data that will reinforce those very presumed correlations. (They obviously don't read this blawg, which is rather the point.) That's especially so when there's an unconsidered and intentional barrier to the data: I refuse to participate in "polling" because I not only value the secret ballot, but despise any candidate who will change his/her/their (fundamentally unreliable) campaign rhetoric to appeal to me Because Numbers and not Because Persuasion.

    It's not just for politics, either. Misuse of probabilistic analysis of flawed datasets influences what music reaches me without significant, well-above-"budget" effort, what art I might purchase (or even see, since museums follow a similar path), and — returning to the other end of this sausage — espresso versus other means of brewing coffee. Intentionally, and even accidentally, excluding outliers from the dataset means not just that the outliers won't be served, but that the "analysts" will not consider them. (Go ahead: Find a romantic ballad or personal-relationship revenge song fitting any "data-driven" hit-making model on one of the top few albums of all time.)