Kieran Healy

This reminds me of one of my favorite articles: https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin… This is a great article to share with friends who don't understand how dangerous this is.

The best tutorials for indoctrinating software engineers in ML are written accessibly & use real-world business-case data, or easily comprehended data sets. One great example of this, from graph theory, was this fantastic 2013 article by Kieran Healy: https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin… There's no getting around the complexities of fit, bias and customized models for many ML problems, so my observation above is obviously limited in its applicability.

This was explored before in an excellent blog post. https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

Sure, let’s see: https://www.blog.google/documents/57/Overview_of_COVID-19_Contact_Trac… Right away I notice the glaring omission of “the surveillance state” in the wording of certain sentences like “People who test positive are not identified to other users, Google or Apple“. “With Bob’s consent, his phone uploads the last 14 days of keys for his broadcast beacons to the cloud”, also giving away his geolocation in the process. My read of the “doesn’t collect location data” is “doesn’t explicitly collect location data”, like just because it doesn’t send a field ‘geolocation’ they can technically say that and not be lying. You should check out one of my very favorite blog posts of all time: https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

Or the classic least norwegian squares method of regression in the WSJ http://kieranhealy.org/files/misc/laffer.png

There is an excellent argument and guide to this here https://kieranhealy.org/publications/plain-person-text/

Any time I read content written with the voice and philosophy of the author of this article, I am reminded of the excellent paper "Fuck Nuance". https://kieranhealy.org/files/papers/fuck-nuance.pdf Just like there is an appeal to oversimplifying and over-abstracting, there is a similar appeal to over-complicating and "over-nuancing". By trying to account for everything, you accomplish nothing, understand nothing, and gain no insights. The hacker ethos, like any simplified abstraction, does not capture the entire story, but it also doesn't oversimplfy. It provides a useful model to work with and to think about, and trying to make it a lot more nuanced dilutes the points it is making and ruins its usefulness.

It reminds me of the post about using metadata to find Paul Revere: http://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-find… Last-second blackmail isn't really the best way to defeat a political movement. Compromising the central figures in organizing the movement -- not necessarily "taking them out", but potentially just reducing their involvement or creating some alienation -- when the movement is still young is the way to really stop it.

Playing devil's advocate a little, the value in the data from an intelligence perspective is probably in the network it uncovers. If you know a circle of people are regularly exchanging encrypted messages, then you can try to target one member and get them to compromise the group by turning them into an informant. Similarly, if you suspect someone, you can look at their records to discover the networks of potential conspirators, then fan-out for each member until you have a big picture of interactions between different networks. There's a great example of exploiting simple person-person interaction graphs to generate this information here: http://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-find…

kjh the author of this link has a greatintroduction to emacs: http://kieranhealy.org/resources/emacs-starter-kit.html He also has a lot of good reference material on latex/org-mode/pandoc. Definitely a good resource to have handy if you ever have a less-techie friend who wants to get away from MS word and its ilk.

"individual liberty and personal responsibility (e.g., leave me alone, i'll leave you alone) " Those are not conservative values, those are libertarian values, we might call them 'liberal' in the rest of the world (well, kind of, using that phraseology like 'leave me alone, i leave you alone' is kind of a American colloquial/populist or crude way of expressing it - in fact it doesn't describe classical Liberalism very well at all, but nevertheless someone with that ethos would definitely fall under this camp, and not Conservatism) Classically Conservative 'values' (though not necessarily politics) are more: family, community, responsibility towards community, faith, tradition, duty etc.. Conservative and Libertarian values clash distinctly in the are of 'community' - one is individualist, one is instinctively though often not obviously community oriented. 'Duty' is generally a conservative concept you don't hear much in the Libertarian leaning crowd. But it's moot as far as this article is concerned: The data in this article is highly problematic due to the fact that most newcomers move to the cities, and so it entirely skews the data. It's not an argument of 'disperse vs. dense' - it's really just a proxy for ethnography. From the link in the comment above [1]: "population density and percent black will do a lot to obliterate many a suggestively-patterned map of the United States. Those two variables aren’t explanations of anything in isolation, but if it turns out…

It's very reductive to claim that only addresses, bank account [numbers], and SSNs could be dangerous if leaked. What about people who discuss private matters in chat? (e.g. sexuality, medical history, drug use, etc.) Yes, we know that's a bad idea, but most people aren't HN readers. What about state actors using Facebook's metadata alone to quash democratic movements before they get off the ground?[1] To advertisers, most of the social graph is available only in a semi-abstracted form (e.g. target X, Y, and Z qualities and degrees of connection). A leak or sale could make that information available directly. What about criminals using a public leak of location data and some predictive algorithms to strategically rob homes? (and maybe also using patterns of likes and connections to predict whose home is worth burgling?) Really, to say that the valuable part of Facebook is already freely available brings to mind the words "limited imagination." Facebook keeps the valuable part to itself and meters even limited access to it. And there is extensive danger with the possibility of a leak or sale of the raw data. [1]: https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

Have you tried Pandoc[0]? It's the best, simplest solution I've been able to find for writing content in Markdown, and then almost trivially exporting that same content into PDF manuscripts, or PDF presentations, or HTML presentations, or more, without having to change the source content. It seems to be increasingly popular in academia for handling class notes/slides/papers all in one go. It can have good citation processing with cite-proc [1]. One of my favorite things about it is, if you know you're just going to be exporting to a specific filetype like HTML or LaTeX, you can just plain throw in code for that language directly into the Markdown, and Pandoc will run/include it. (shameless plug alert) I wrote a super-simple Makefile script for using Pandoc, called acadoc [2]. It lets you call "make beamer", "make manuscript", "make html_presentation", etc., in a directory with Markdown files to turn them all into whatever presentation/manuscript/paper/etc. you want. The key thing, though, is that for LaTeX presentations/etc. it uses an intermediary "style/manuscript.tex" file to customize how you want your resulting LaTeX to look -- so if you want the same content, but decide the presentation looks better with white-on-black instead of black-on-white, you only need make a few changes and create a "style/manuscript_dark.tex", copy-paste that style in the Makefile, and now you can call "make manuscript_dark". It's meant to be quick and easy to use, and theoretically easy to…

I have to say, the underlying paper scares me a bit. It seems to be pretty heavy on research politics, driven by a very limited study, and the authors do not even seem to grasp the mechanics that underlie the generation of computer documents very well. Example: "In all other cases, we think that scholarly journals should request authors to submit their documents in Word or PDF format. We believe that this would be a good policy for two reasons. First, we think that the appearance of the text is secondary to the scientific merit of an article and its impact to the field. And, second, preventing researchers from producing documents in LaTeX would save time and money to maximize the benefit of research and development for both the research team and the public." I am honestly not sure how they think accepting papers in PDF format would prevent researchers "from producing documents in LaTeX". Do they really not know that not only can you generate PDF from LaTeX, but that that is what is commonly being done? Second, do they really not understand that there are options other than Word or LaTeX? I know of quite a few people who use Pandoc (Markdown + some inline LaTeX for when you need equations/references) [1]. Others use a non-Word word processor (e.g. because they're running on Linux). Others again use LyX. This casts doubt upon their ability to make policy recommendations when their understanding of the available toolset is incomplete. Third, the study seems to be awfully…

Also, only knaves and fools would be concerned with The Crown's modest regulations for reporting "meta data" on all guilds and meeting-houses. "Using Metadata to find Paul Revere": https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

An important consideration is that just the graph of who talks to whom can be quite powerful: https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

Not OP, but here you go. OP had it slightly wrong though: it's not tavern records but membership lists of colonial Boston organizations, and the author is a sociology professor (Kieran Healy), not from Princeton or Harvard. He uses basic social network analysis on historical membership data to identify Paul Revere as the key figure among 254 colonists using nothing but "metadata." The whole thing is written as a satirical report by a British intelligence analyst in the 1770s, which makes it a pretty effective commentary on the "it's just metadata" argument from the NSA debates. Link: https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

It is not historical: https://kieranhealy.org/blog/archives/2025/08/07/blueberry-hill/ Perhaps they have a hot fix that special cases HN complaints?

See https://kieranhealy.org/files/misc/levimartin.pdf "What do animals do all day?": The division of labor, class bodies, and totemic thinking in the popular imagination (2000) > Keywords: Animals; Totemism; Class body; Busytown; Symbolic domination; Division of labor (36pp, not entirely serious, about which animals do what in Busytown; contrasted with Babar, etc.)

With only metadata, just who is mailing who, you can build pretty useful social graphs of people[0]. [0] https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin… For anyone else looking for the link.

I was not familiar with this term and found this blog post that explains it: https://kieranhealy.org/blog/archives/2002/10/31/cooling-the-mark-out/

Using Metadata to find Paul Revere. -- Keiren Healy -- June 9, 2013. https://kieranhealy.org/blog/archives/2013/06/09/using-metadata-to-fin…

In that vein, this spoof captures the experience of reading a Springer "Introductory" textbook: https://kieranhealy.org/blog/archives/2022/05/20/every-springer-math-t…

"Fuck Nuance" http://kieranhealy.org/files/papers/fuck-nuance.pdf (PDF)

This resonates surprisingly well with the “Fuck nuance” paper in sociology[1], even if I can’t exactly pin down why. [1]: https://kieranhealy.org/files/papers/fuck-nuance.pdf

Kieran Healy's Data Visualization: A Practical Introduction https://kieranhealy.org/publications/dataviz/

My usual reference, now a little old: https://crookedtimber.org/2012/07/20/america-is-a-violent-country/ : http://kieranhealy.org/files/misc/assault-deaths-oecd-ts-all.png > If you remove that single factor "If you remove the murders, the US has a low murder rate" isn't helpful. > It's obvious that the murder problem in the US is a handgun problem and the stats always bear that out year after year Well, yes, but this runs into the "this is constitutionally guaranteed and we refuse to even contemplate doing anything about it" problem. But the difference is huge. In the 1980s, Detroit was more lethal than Belfast, with regular bombs going off and troops shooting civilians in the streets.