s07e23: For the people, by the people
0.0 Sitrep
It’s Wednesday, November 27 2019 and I’m writing this while my wife is watching Noelle, the Disney+ original movie in the background. I am trying to not pay any attention to it because if I look up I might see that scene where Anna Kendrick finds out she knows ASL because she’s a Santa and the world is horrible and burst into tears.
[ACTUALLY, there was a really nice thing I noticed which was… I guess the titles equivalent of diegetic audio? It was a scene cut and the camera was craning its way down over a shopping outlet with a sign on a building that said 03 SHOPPING DAYS LEFT UNTIL CHRISTMAS and it was kind of preferable or at least interestingly nice compared to the giant FLOATING RENDERED LETTERS IN 3D SPACE that we used to get with Fringe or whatever]
Both kids are asleep, and I had a movie night with the older one watching My Neighbor Totoro.
Actually, can I just say how annoying it is to not be able to get the Ghibli movies streaming? It’s not because I’m super excited about a future where we only rent DRM content from the Content Masters of the Universe, but more because if you have young children then you are also probably familiar with the experience of hoping that you can clean the {peanut butter, jam, felt-tip pen, snot, other unidentified substance} off the surface of the Blu-Ray.
Tomorrow is Thanksgiving in America. There’s a lot going on. ICE (Immigration Customs and Enforcement) is bullshit.
1.0 The Things That Caught My Attention
One slightly longer thing, and then a few shorter things this time.
1.1 For the people, by the people
There was a report that came out a few weeks ago from Muckrock. I have to admit that every so often I get a bit confused about Muckrock because it is next to the space in my head occupied by the Drudge Report. But Muckrock is a very different thing and not something that is familiar to me as British trash tabloid journalism; Muckrock aggregates public records and public records request.
The report in question was about the effectiveness of data-driven, predictive policing. If you’re the kind of person reading this newsletter, then there’s a chance you’re already somewhat cynical about the effectiveness of such policing, but at the same time, likely in favor of it because what’s the alternative? Non-data-driven policing? Policing at random? Policing at people who just don’t look like they’re up to any good? Policing people for doing perfectly reasonable, normal things, but just because they’re black?
I mean, I leave it up to you to figure out if policing at random would actually deliver more equitable justice and safe neighborhoods and so on than policing as is currently practiced in some locations.
Anyway.
Here’s an excerpt of the story, with my emphasis added at the end:
MuckRock has submitted requests to more than 50 agencies known to have used the technology, asking for information on their predictive policing training and use. Of those that have responded, a few have been able to send their contracts, input data, scientific paperswritten by PredPol’s makers, annual mayoral presentations, and even some training materials, but none have been able to provide validation studies.
The report, released on 5 November, stuck in the back of my head, bouncing around waiting for something to associate with and gel in some way.
That same day, I sat in on a quarterly stakeholder meeting at the day job. What I’m going to say I want to be taken in the general sense, as a sort of application of the mediocrity principle: if there’s no reason for things to be completely different in other contexts, just assume that everything is more or less the same.
The quarterly stakeholder meeting was a very fulfilling meeting and that’s hard for me to type because, let’s be honest, it’s called a quarterly stakeholder meeting and it is not a) a meeting where there are stake holders; b) a meeting where such stake holders are quartered, and c) and so on. I do not like these meetings because they are unnecessarily full of jargon, and every time I hear some jargon there’s a part of me that looks around the room and is convinced at least 25% of the people are just nodding their heads and pretending they know what the jargon-saying person means. I don’t think they do know, but everyone is too afraid to stick up their hand and say EXCUSE ME, I DO NOT KNOW WHAT THIS MEANS, CAN YOU JUST USE NORMAL WORDS PLEASE.
Anyway, the problem is this: people are hard. Society is hard. We do not live in a fully automated luxury space communism. While it’s true that, like the future, we have lots of resources, just unevenly distributed, we still have to deal with the difficult issue of distributing those not-yet un-scarce resources.
So, we have to make decisions: where do we spend money? Where should we live? Should people go hungry? If people shouldn’t go hungry, how should we deal with that? If we only have so many police, where should they go? What should they do? Should police, for example, prioritize fare evasion? If so, why? What’s the opportunity cost? There’s the tell, see: the cost. What’s the theory? Is there some sort of broken window of public transit fare evasion where if someone doesn’t pay, within 60 days all hell breaks loose?
I asked this question on Twitter: is it that
a) bureaucracies are considered harmful at scale; or
b) scale is considered harmful?
There is of course the third option: why not both. A couple people pointed out that the more direct issue is how close decision-makers are to the decisions being made, and that there’s a perennial tension between the benefits of bureaucracy being able to act at scale (otherwise, how are we going to give people all the things?) and the inevitable downsides of acting at scale (impersonal bureaucracies that deliver scale through inflexibility).
And then I got to have a conversation with Janet Haven from Data & Society. We talked about, amongst other things, the work that Rebecca Wexler, their lawyer-in-residence had been doing to prevent the mechanism of trade secrets being used to shield the workings of proprietary software involved in someone’s conviction.
There are a few things going on here.
First, the Muckrock point: we don’t even know if mechanisms like data-driven or predictive policing work. Second, combined with the Data & Society point, in many cases, we’re not even allowed to understand how the mechanisms behind them work.
I had always thought, somewhat naively, that the promise of automation would be that the machines would do the drudge work leaving the edge cases to humans. That the baseline cases, the ones that didn’t need any intervention, would be dealt with and that where we needed nuance or understanding or even goddamn sympathy or compassion, there’d be a place for a human to do that.
The promise of the project I’m working on for the day job — a child welfare system — is that social workers will be freed to spend time concentrating on the people they’re supposed to be helping instead of, as the stereotype goes, spending time on paperwork and procedures. People, then, making decisions, instead of systems.
This was the part of my conversation with Janet that set me thinking: we set up these systems, and in this day and age, we may well have access to the data and information to try and figure out if we’re achieving what we set out to achieve.
At the same time, we have more public understanding of the usage of “algorithms” in the public space (and the private space, I suppose), even though those algorithms are seldom that different from opaque bureaucracies. And all of this in the service of being able to act “at scale”. If you wanted to take advantage of this rhetoric and to, say, put humans back in charge of making decisions about other humans, one thought I had was this: in America, at least, you’d make the point that government is of the people, by the people, for the people and nowhere in that does it say that impersonal, non-personal algorithms get to make decisions about people. I mean, just grab that rhetoric straight from the Gettysburg Address (or wherever it was cribbed from so that it could be part of the address).
The point being this: I suspect a politician might be able to make some hay with saying that government should be by the people, for the people and, like I said, not some algorithmic black box. People might get things wrong and people might also institute the wrong systems, but at least you can yell at them about it and, I dunno, vote them out.
2.0 Some shorter things
The Aspen Tech Policy Hub recently published the results of its first year fellowship, and one of the projects was about test-driven development for policymaking, which I see in the same space as Code for America’s paper on delivery-driven policy.
There’s an 8K immersive holographic display that uses light field tech called Looking Glass (which, ha, for anyone who’s played Prey) and supports “45 distinct and simultaneous perspectives of three-dimensional content”. If this works, maybe give it about 10 years before it hits consumer adoption?
As an aside, I always thought 3D TVs in the late early 2010s fizzled out because there just wasn’t any content for them, especially when streaming started becoming a thing at the same time. 3D just wasn’t that interesting or relevant compared to “all the movies and TV shows”. But I think we’re in a different place now with something like Looking Glass - the display comes with Unity and Unreal 3D engine integration and a javascript library for browser-based applications. The 15.6in devkit model starts at $3,000 which feels like it’s in the 15in TFT flat-panel price point back in 2001.
The Reagan Library (which, you know, is problematic) is getting its own F-117 stealth fighter, so I guess we can look forward to all manner of drone-related items in the Obama library.
The passenger name record in travel GDS systems is very old and doesn’t really handle a name ending with the letters “MR” or “MRS”; this Stack Exchange answer is pretty cool for including the quick reference guide on how to create a PNR and why names ending with MR/MRS have problems.
Cory Doctorow has written some short stories about the effects inflexible software will have on all of us.
I’m bookmarking this piece in Quartz about Microsoft having “a plan to make anyone a software developer” for my novel-in-progress; if you have any idea about how I’ve been playing with fiction on Twitter, you might see where I’m going with this.
People are terrible, people make software, therefore all software is terrible: Valentine’s Day edition, also this is not surprising to anyone who knows how SMS works and how much of a kludge it is on the network backend.
Well, that’s it! It’s Thanksgiving tomorrow and then it’s Friday which is the latest possible time at which I’m allowed to make it onto a 40-under-40 list which, having had some time to get used to, I’m no longer bothered about in any respect at all.
I hope you’re all well!
Best,
Dan