In the first class I mentioned peñabots as an example of government social media manipulation. Rather than define them myself I’ll lean on the Wikipedia page definition: “the name given to what analysts believe to be a network of automated accounts on social media used by the Mexican government to spread pro-government propaganda and to marginalize dissenting opinions in social media”. Though not the first institutionally-organized trolling operation they stand apart for their sophistication, their effectiveness at shutting down oppositional organizing, and their use in a nominally non-authoritarian government (although one with a deep history of media manipulation).
A lot of research has already been done on peñabots so I will just point to a few good places to get started. I first heard about them in Finn Burton and Helen Nissenbaum’s great book Obfuscation, which is a great research jumping off point for a whole host of different kinds of obfuscatory digital practices. The above talk from Chaos Communications Camp in 2015 is a great expose of original research. If watching something doesn’t feel right, Erin Gallagher has collected a trove of news and research articles collected on in a Medium page. If you only want to read one article, you could check out this one from Motherboard. Or if you are more into podcasts you could listen to this Reply All episode.
Now, for Americans in 2018, peñabots probably call to mind Russian government-sponsored trolling. For a really good early treatment on the Russian Internet Research Agency and their social media manipulation campaigns I would direct you to Adrien Chen’s piece for the New York Times Magazine.
For my transcription exercise I wanted to pick something that was
a) short enough so that I could transcribe the whole thing
b) interesting enough that I wouldn’t mind listening to it over and over again
c) spoken slowly enough that I would only have to pause and rewind ten seconds 500 times and not 1000.
I’d recently watched to MacKenzie Wark’s edutainment video on Chantal Mouffe for DIS magazine (now just DIS?) and decided that it met the required criteria, as well as having some arbitrary time limit for existing on the web (just 11 days left on the web) that gave my transcription a greater sense of archival weight. Ken Wark has always seemed a little unctious to me but I didn’t mind having to listen to his disembodied head talk for an hour or so as much as I thought I might.
You can see the original video here and see the transcript here. I mistyped “Mouffe” as “Moufee” the first time I wrote it and since I was typing in atom it autosuggested the wrong spelling every subsequent time. I decided to leave it in as an interesting artifact of the process.
For my computational media final project I wanted to make a game about Cambridge Analytica and about the current cultural moment in data science and AI. I have seen the rise of the imperative to make one’s identity known and intelligible for collection online by various state and corporate actors as a sinister reach to further institutional control over the everyday, but I also think that the practice of psychographics is a silly pseudoscience. I’ve also recently been interested in games that have an anti-interface with the simplest possible values (especially cookie clicker games but also internet quizzes and ClickHole ClickVentures) so I wanted to limit the interface to just the fewest possible elements.
You can play the game here.
And see the code here. It is run via express server serving a p5 sketch. All the interactive elements are dom buttons that “switch on or off” different arrays of text written to a p5 canvas. Note some buttons are confusingly labeled.
The song is Mirandolina by Burnier e Cartier (1974).
One of my favorite word games to play is to try to suss out the grey areas of meaning with seemingly well established words (e.g. what foot angle and distance of leaving the floor differentiates a ‘tap’ from a ‘step’, in what ways are the aesthetic associations of a crypt different from those of a mausoleum). Kind of a useless semiotics examing the edge case value of a word (how much mist on a misty morning differentiates a bog from a swamp).
Recently I’ve also had psychographics, Cambridge Analytica, and the ways in which social media and internet usage data are being used to group, differentiate, and make the global human population intelligible to various governmental and corporate actors.
Combining these ideas I’d like to make a game focalized through the perspective of a machine learning algorithm tasked with assigning personality and demographic profiles to users based on those users’ associated word choices (i.e. a user decides that a bog has 85% more mist than a swamp on a cold bracing morning in the early spring -> that user is open to trying new things and likely lives in a large cosmopolitan city within the Eurozone).
The game interface would be half personality quiz and half Clickhole-style-choose-your-own-adventure. The aesthetic I’m looking to create calls on clickclickclick.click and Human Resource Machine much more so than AI fearventures like I Have No Mouth But I Must Scream. My end goal, other than to create an amusing and engaging experience, is to make intelligible the ways in which machine learning is deployed to group and assign identities and to hint at methods of real world digital subversion.
Inspired by the near extinct website guestbook and completely extinct methods of phone phreaking, I wanted to make something that felt as unfinished and wide open with potential as the “old” web of 1992 – 2001.
Please participate: https://tinyurl.com/y7tvrusf
Communication between users in this space is possible but only through “hacks” of the intended use. There is no security and everything can be seen.
At this point I have gotten the basic Google home API working on my raspberry pi and have started manipulating the methods of interaction (changing the trigger word, changing the voices) and have started to work on building a specific Google home Action. Still to do is build the haunted dollhouse enclosure. Since the basic first interaction I want users to have with the haunted doll is to knock on the dollhouse door, I wanted to first set up a basic knock and open circuit.
The piezo would be connected to the inside of the door and the motor underneath or above the door, rotating an axle that controls the door hinges. I am not sure if a servo is best for this or if there is a better motor for this application. The piezo would detect the knock (any kind of knock should do) and trigger the motor to open the door.
Beyond this very basic interaction I have a few different ideas. The aesthetics and interaction in my mind are beginning to push into Zoltar territory, so I was thinking of adding a thermal printer printout of whatever fortune or advice the doll is giving. Additionally I would like to replace the trigger word with the knock, i.e. to connect the arduino to the pi and trigger the pi to start listening for a question once a user has knocked on the door. Additionally I would like to incorporate some movement in the doll, that was the biggest feedback that I received at user testing. I think an easy start would be to put two LEDs in the eyes so that the eyes light up. From there, with time and resources I could add additional movements, such as a head or arm movement.
I really wanted to work with the twitter API and to show data in real time. I started working with the word cheese just to test it out and I ended up sticking with it throughout the process.
You can observer cheeses here.
And see code here.
Recently I’ve been thinking mostly – but somewhat disjointedly – about haunted dolls as relational to AI and various societal-level fears. The doll as feminine artifact rendered creepy by autonomous movement and malicious intent is drawing parallels for me to visions of consumer-level AI – which frequently is embodied feminized labor – disrupting human civilization, both on the level of job displacement and as genocidal overlords. Many questions arise. At what depth of the uncanny valley does misogyny lie? On what level are we as a society afraid that factory workers and truck drivers are being emasculated by machines and at what level are we genuinely concerned about joblessness and the negative externalities of creative destruction? How much are baseless fears of AI – akin to fear of paranormal and unknown – tangled in with worthwhile critiques and concerns, and how should AI policy and practice be formulated taking this intermingling into consideration?
This is a puzzle that I want to play with so I’ve been wanting to make some kind of “haunted doll” and to put inside it a camera with face tracking and servos that allow for head movement to do the kind of “Scooby Doo” project but then also put in a microphone and speaker connected to a Raspberry Pi running the Amazon Alexa API. Voila, a creepy Amazon Echo.
I’m in the middle of changing hosts for my blog so I will keep this entry short. I made a game involving clicking DOM buttons… a lot.
You can see the code for the game here.
And you can play the game here.
For our midterm Ridwan and I thought it would be really funny to make a machine that claims to give a high five but actually fakes you out every time.
Aesthetically it reminds me a bit of the toilet hand from Zelda.
Inside it has an arduino, two metal gear servos on a pan/tilt bracket meant for camera tracking controlling the hand, an HC- SR04 distance sensor, and an LCD screen. Ridwan found a box template for a laser cutter and we laser cut the enclosure out of cardboard, making the necessary modifications for our sensor, hand, and screen.
We had quite a bit of trouble with powering servos and actually destroyed two servos by putting too high a voltage across them.