This Week I Learned #2

Assumed audience: People who like reading and learning, on any of a wide array of subjects! (There’s probably something on this list for you!)

Epistemic status: Mostly just learning.

  • Attention conservation notices. I have no idea what the Viridian List was, but Viridian Note 00002 seems to have been one of the earliest sources of the kinds of notices-to-users that now gloriously festoon a number of my favorite sites (including this one): qualifying the content for readers. In this note, Bruce Sterling introduced the idea of an attention conservation notice:” a disclaimer that explains all the reasons why you shouldn’t read or pay attention to the thing described or linked in any given post on the list. (If anyone knows whether this or the rationalist community’s epistemic status” came first, I’d love to hear about it!) Particularly given claims advanced by Simone Weil midcentury and given a new hearing through Alan Jacobs’ work, I’m grateful that Sterling tried this; I wish the rest of the internet would catch on.

  • Website performance. Ethan Marcotte on a theme he’s hit on a lot over the last several years: JavaScript performance as a moral issue. In this particular piece, he picks up on a recent writeup by Tim Kadlec (who also makes similar arguments on a regular basis) to critique much of modern web development — with a fairly loaded set of assumptions. This is an interesting example of how your starting point profoundly shapes your interpretation of others’ choices. If I have a major critique of both Marcotte in this article and in general, it’s that he substantially overstates structural forces (which are very real, but are not total) and substantially understates human agency (which is also real). He also conflates all under one umbrella things about which (at least some) real users feel very differently: page takes a little while to load because it’s doing something really cool is quite different from page takes a little while to load because it’s full of an unimaginable number ads and behavioral tracking scripts. Even if we were to grant that both are bad (spoilers: I don’t!) they are bad in profoundly different ways, and this discussion would be much further along if folks like Marcotte granted as much.

  • Personal and institutional sacrifice. Kathleen Fitzpatrick argues that while the question of generosity in hard times is a difficult one, it’s not intractable. First, it requires us to work to restructure our organizations in good times such that the sacrifices that necessarily come in hard times don’t flow onto those most vulnerable, but onto those least vulnerable instead. Second, it requires us to recognize that structural work is necessary — that personal connective work is very good, but insufficient to the task of institutional reform. And at the end she gets down to the brass tacks of dealing with department-level bylaws and policies” which restructure the parts of the institution over which we do have influence: grassroots change” which can create change both outward and upward.”

  • Health insurance. Scott Alexander/Slate Star Codex on how catastrophically bad our approach to health care really is, with a frankly soul-crushing list of ways that our health care system fails people. This is just people Scott Alexander has directly encountered, and every one of these is a nightmare.

    Any other system would fix these problems. A public system like Medicare For All would fix them. A communal system like the Amish have would fix them. A free market system like our grandparents had would fix them. The prepaid doctor cooperatives Reason talks about would fix them. A half-assed compromise like Joe Biden’s Medicare For All Who Want It would fix them. But here we are, stuck with a system that somehow manages to fail everybody for different reasons.

  • AI: an interesting result in Proceedings of the National Academy of Sciences showing formal mathematical limits on what certain techniques used in current machine learning/AI approaches can actually accomplish. There’s the real takeaway here: this is an important limitation of current approaches, because some of the modeling techniques inherently cannot encode sufficiently rich information to correctly represent complex kinds of networks (such as social networks). And then there’s the hot take version of the takeaway: AI is permanently limited and is broken!” Ignore that one. And finally, there’s at least one really interesting question that comes out of this: why do these work so well given they can’t actually encode the relevant information? — because they do! HT: Alan Jacobs.

  • Politics and sexual assault. The fact that the 2020 presidential election is ending up being two men credibly accused of sexual assault is horrifying. Sarah Jones’ writeup of the serious and credible accusations that Joe Biden sexually assaulted Tara Reade is incredibly difficult to read, but it’s also worth reading. The choice before us appears to be: hold your nose and vote for the perverted, wicked man who is closer-aligned to your particular set of politics. Would that either party had the integrity to simply say no to that follow. Me, I will not be voting for either of these men; perhaps if enough of us chose the same, things might actually change.

  • Leaving conservatism behind. A memoir and conversion story, as it were: Matthew Sitman’s essay from 2016 on how he became an out-and-out leftist — not despite but, as he tells it, precisely because of his blue-collar conservative upbringing. This was sobering. I remain a conservative” of some sort. Both for the reasons suggested by the above comments on politics as well as for the kinds of problems outlined in this article, though, I am not on board with what the word conservative” has come to connote in terms of party and politics.

  • Historiography and Gnosticism and Theory.” An interest polemic by Cyril O’Regan against a particular brand of scholarship on religious views often described as Gnostic” from the first few centuries of Christianity — especially around the Nag Hammadi texts. His claim: that fashionable arguments against the use of Gnosticism” and in favor of Christianities” in its place are so much nonsense. The argument is an interesting one (I think!) even if you’re not all that engaged by the specific question, because it gets at the very human tendency (very much loudly at play in certain corners of academia!) to act as though one’s own view is unbiased and it is only those one critiques who really have problems to deal with. The historiography he goes after is, accordingly, genuinely pretty bad, even if you don’t agree with all of his polemic.

  • Type theory. Jonathan Goodwin surveys the development of type theory. He starts with Hilbert’s Program and the various developments in mathematics and formal logic of the early 20th century, then moves forward through the connections that formed between those disciplines and computing in the 1930s and 1940s, then follows the threads through the present day (with appropriate humility about the last couple decades). A lot of this was review to me, but the ways Goodwin connected the dots between the items was helpful.