social hacking computer vision

our final assignment is to use computer vision techniques. i have two ideas:

  1.  an IRL game where we use our “meat eyes” (trevor paglen’s evocative words) to see like a computer
  2.  something about the history of photography, which has to do with the history of the mugshot, phrenology, and the practice of reading criminality on the body. some of the face tracking diagrams remind me of cesare lombroso’s face measurements:


unfinished thoughts:

i’ve been thinking about trevor paglen on meat eyes. in a recent talk, he said: “most of the images made in the world are made by machines, for other machines, and human eyes are not in the loop.” what is an image if we can’t see it? is there a way for our meat eyes to see the output of computer vision processes in a way that doesn’t interfere?

are we externalizing our brain when we try to make a computer see like we see? or, rather, to make a computer translate what it sees into something we can see? if our visual systems, in our brains, evolved from needs in our environments, needs that computers don’t share, and we are now building systems to replicate those visual systems with inorganic parts, isn’t that interesting?

like construction machines that mirror body parts. the way diggers look like arms with scooping hands at the ends. what does it do to have a bunch of body parts with no brain, only their own affordances, roaming around the city? to have these disembodied arms everywhere, digging stuff up and making holes for condos to go in? what about these disembodied computer eyes? to whom do they belong? where is their brain?

last but not least, john berger in ways of seeing: “we only see what we look at… every image embodies a way of seeing.”

need to come back to these:
the brain’s visual processing system
john berger’s ways of seeing
kyle mcdonald’s notes on computer vision

two things

This is a sketch of what could be a larger project about data and databases, knowledge production, and the electronic turn in social services. It’s inspired by Craig Willse’s 2016 book, “The Value of Homelessness: Managing Surplus Life in the United States”—specifically, the “Governing through Numbers” chapter.

At this point, there is only one modal. When you click the “First Name” input, you see part of an interview from the Signs for the Homeless project (link below). I’d like to create modals for every input box.

In disrupting the data field this way, the idea is to show how databases constrain the kinds of information we can know about a person. The resulting “statistical portraits,” to use Willse’s phrase, stand in for a population that doesn’t actually exist until the database brings it into being. This reality has important applications for how we use databases (and algorithms that rely on them) to create and inform policy.

a chrome extension for bill bratton & pat lynch

remember that one time when arrests for minor offenses dropped 96% as a result of the nypd’s virtual work stoppage? when cops decided they’d only make arrests ‘when they have to’?

“Summonses for low-level offenses like public drinking and urination also plunged 94 percent — from 4,831 to 300.”

these kinds of offenses, as matt taibbi describes, are quality-of-life offenses—part of a broken Broken Windows policing philosophy that doesn’t actually solve any problems, but does sweep poor people off the streets so we can pretend they don’t exist 🙈🙉

what stars aligned for bill bratton, pat lynch, and thousands of cops to decide on a virtual work stoppage? and how do we make it happen again?

all signs point to a dangerous conspiratorial trifecta: mayor bill deblasio, labor arbitrator howard edelman, and quentin tarantino.

i’m working on a chrome extension for nypd leadership that will collect evidence of the trinity’s unholy work. in light of such anti-cop instigating, the only reasonable step for the nypd is another work stoppage.

Screen Shot 2016-02-25 at 10.36.45 AM

closed data

my initial idea for the social hacking API assignment forked into a few different projects.

@nyclandlord twitter bot: the idea for this bot was that it would tweet things like

  • “just kicked out a tenant at [address]” — this would come from eviction data
  • “just bumped the rent at [address]” — this data set doesn’t exist
  • “successfully gentrifying on [block] in [neighborhood] today” — this would be a combo of historical rent changes? maybe?
  • “still ignoring tenants at [address]” — this would come from housing code violations, AN ACTUAL COMPREHENSIVE DATA SET WITH LOTS OF DATA
  • “#winning at #gentrification today”

i got a version of the bot up and running, and i still have the twitter handle. the problem is that most of these data sets don’t exist in a ready-to-use way. they are their own research projects in themselves. which i would love to do, but not in a week. that problem inspired what i ultimately ended up making for this assignment (called “closed data” here).

if 596 acres + had a baby: the idea behind this long-term project is adding organizing capability to website, or doing the 596 acres website around rent stabilized apartments. it’s ridiculous that there is no publicly available database where tenants can check whether they are rent stabilized. they have to call the city, they have to follow up, they have to check the historical rent of their own apartment. the onus falls completely on the tenant. and then! there’s no way for future tenants of that apartment or in that building to benefit from the research that has come before them. so why not let tenants publish their findings on an apartment or building’s specific page? like a social media profile, but for a space?

this is obviously beyond the scope of a week-long assignment. i also don’t think it qualifies as “creative misuse of the API” because it’s exactly what APIs are for: making new databases from existing databases.

closed data: the thing i ultimately made, here. this grew out of my frustration trying to find a single database that could tell me whether or not (not whether maybe possibly) my apartment was rent stabilized. after many propublica rabbit holes and hours spent on both the city and the state’s open data platforms, i am coming to the conclusion that the city has open data platforms so they can

  1. bleat about their transparency, in a positive sense (e.g., “we are so open! make apps with our openness!”)
  2. bleat about their transparency, to avoid negative feedback (e.g., “police brutality? rent hikes? show us in the data. it’s all open!”)

except it’s not all open and it’s not all usable. “closed data” uses the names of all the existing nyc open data sets i could find. and that’s all i have to say about that.

vertical carousel with JSON data

i’m so frustrated with the process of looking for rent stabilization data that i’m making this thing:

Screen Shot 2016-02-14 at 7.58.50 PM

here’s my initial sketch:


i want to load data set names from a json file. then, i want the names to scroll vertically every few seconds. i’ve spent some time with this stack overflow post and this jsfiddle, but i’m having trouble both getting the json to load and getting the animation.

hacking APIs, pt 2

many things in my brain with this assignment!

i keep thinking about catherine d’ignazio’s piece “what would feminist data visualization look like?” and especially the part where she asks “Can we ask of our data that it point to its own outsides?” she goes on:

“What if we visually problematized the provenance of the data? The interests behind the data? The stakeholders in the data? A single CSV file or streaming feed often has no reference to any of these more human, material elements that are nevertheless essential to understanding the where, why and how of data.”

i think yes! but how How HOW?!

she suggests: “Invent new ways to reference the material economy behind the data.”

in the process of researching all this rent stabilization stuff, it’s interesting to me that new york city beats its chest about open data, but does not make readily accessible a list of rent stabilized apartments. even though this data set certainly exists. same with the new york state department of finance, which definitely knows which landlords it gives tax abatements to (rent stabilization is partly a function of whether landlords receive tax abatements for making repairs to old units).

some data sets i did find:

Screen Shot 2016-02-12 at 10.58.06 PM

what i really needed was the back end of the form below, a form that apparently exists in many disparate corners of the internet, pulling, in every corner, from the same secret database:

Screen Shot 2016-02-13 at 2.00.53 PM

^^new york city department of finance

Screen Shot 2016-02-13 at 2.01.05 PM

^^third-party, ny rent stabilized buildings

Screen Shot 2016-02-13 at 2.00.48 PM

^^ new york city department of finance again

oy. the data set i will likely use for this assignment comes from chris henrick’s Am I Rent Stabilized? project. he put a lot of work into cleaning up data from multiple places and sharing it here.

i also have a dream of creating my own little database/JSON file with names and photos of everyone who attended this atrocious real estate summit in brooklyn at the end of last year.

oh, what’s that? you need a rent stabilization fact sheet? i got one fer ya. and another. and another.

hacking APIs

our assignment for social hacking is to “creatively misuse an existing API in order to reveal something about the service.”

among other readings, we were assigned a great article by taina bucher about the politics of APIs. it felt a lot like tarleton gillespie’s article on the politics of algorithms (which i outline here).

i’m thinking about the workshop i gave last week on mapping and data. i’m wondering what it would look like if 596 acres and (from now on, AIRS) had a baby. that is, if AIRS had the organizing functionality that the 596 acres map has. that’s a big project. but a piece of it that could work for this api assignment is a twitter bot that does something based on nyc open data on housing stuff.

so far, i’ve reserved this twitter account:

Screen Shot 2016-02-10 at 4.21.23 PM

watching shiffman’s tutorials on twitter bots and node.js, i used node.js and a node package called twit to start making a bot for this account.Screen Shot 2016-02-11 at 10.14.23 AM

now i need to figure out what @nyclandlord will say and when.

i will talk to anyone

for our second assignment (first was a soundboard), we were supposed to do our own version of steve lambert’s “i will talk with anyone about anything (free)”.

i made these signs and stood outside the nyu student center with the first one.



it was raining, and i did not get much feedback. a few folks took photos. one guy told me hated his graduate program, but it wasn’t law school. he told me the rain created a problem with my sample. maybe he’s a stats or econ guy. at any rate, the rain turned my sign to mush so i went back inside.

then, i posted on my personal facebook page. a friend who’s currently in law school shared the post.

Screen Shot 2016-02-03 at 1.16.43 PM

i tried to post a message & photos on the nyu school of law facebook page, but their settings prevent me from posting photos even if i “like” the page. i ended up posting without photos, and my comment ended up on the left side—under page stats, “about”, “photos”, and “videos”. in terms of location on the page, it takes 2.5 full downward swipes on the mouse to get there. no one will ever see it.

Screen Shot 2016-02-03 at 1.43.24 PM

the columbia law school doesn’t have an official facebook page; the unofficial one doesn’t let you post. there is a “columbia law school” group within the columbia facebook page, but it requires a email address. i looked at the university of washington’s law school page. they don’t let you post at all.

no comment

this is interesting. where are the public fora for this kind of feedback? how do institutions constrain conversations in physical and digital space? how do they use their presence on social media platforms to support the idea that they’re interested in engaging with the public, while, at the same time, limiting the kinds of engagement that are possible?

one way we know we’re asking the right questions in public, in physical space, is that security gets called or the cops come. in digital space, when there isn’t even an opportunity to appear, to be silenced, i wonder how we might signal to each other that we are asking questions at all.