Carlo Zanni: Where is the art side?

Derek Curry & Jennifer Gradecki: The CSIA could be described as a work of conceptual art, tactical media and critical engineering. As a conceptual artwork, the form of the CSIA is determined by information about actual surveillance systems we found in FOIA files, ethnographies, and leaked documents. These documents provide the concept that drives the work. As tactical media, the CSIA appropriates the forms and methods employed by intelligence agents in order to open up their secretive practices to analysis. Whereas Duchampian appropriation often reframes an image/object/video as an artwork to be evaluated based on formal qualities rather than use-value, we believe that art always stakes out a political position and can have a social use-value. As a work of critical engineering, the CSIA aims to provide users of our app with practice-based experience with systems of mass surveillance in order to reveal ways that technology shapes behavior.

We also feel that interaction with, and the presentation of, data produces a type of aesthetic experience that is related to ‘qualculation.’ Qualculation pertains to the material process and set of practices that allow people to make qualitative rational judgments: it includes the spatiotemporal arrangements and metrics needed for making value-based decisions. The CSIA is an interface that provides qualculative tools that allow the user to calculate the threat a Twitter post may pose to national security. This subjects conversations on Twitter to a new type of rationalized judgment that frames how those posts are interpreted. We have been calling the creative manipulation of these types of judgments ‘qualculative poetics,’ and think this could be a new phase in the development of analytic conceptual art. This sometimes results in the questioning, or disruption, of a dominant, rationalist logic similar to the tradition of conceptual artists who used concepts from linguistic philosophy in their work to explore the limits of language as a tool for communication.

DC | JG: Where are you pulling the news data from: is it a live data feed?

CZ: Yes it is a live data feed. We restored the piece after it stopped working some time ago. This time the feed comes from abc news. World, world news, travel, technology and Entertainment categories. The piece is quite old and has to be periodically fixed.

DC | JG: Is there a conceptual reason for your choice of news sources?

CZ: No, in theory the more the better.

DC | JG: We noticed that the conversations in the game are focused on world news events, while the images that come up while shoveling seem more random, pertaining to celebrities about half of the time. Is there a tactical reason for this, or is it the results returned from a newsfeed?

CZ: I think also the conversations contain mixed subjects. Percentages depend on the call done at each game reload. Entries and percentages change all the time, but sometimes it feels that one is more prominent.

DC | JG: Why did you choose to use the form of a game? Do you find that a game offers a type of interactivity, aesthetic, or engagement that other forms of media do not?

CZ: Game was hot back in the days. It was in the air. It was just very natural. The "game" medium was the right tool for the type of work I wanted to do.

CZ: I did try the twitter post checker. But I wasn't able to get any "suspicious" posts out of all the people I wanted to kill. Could you give me an example of a post being flagged as suspicious by Agent Bayes?

DC | JG: The Agent Bayes algorithm works by creating weighted vectors from a corpus of tweets labeled as threatening or not threatening. The corpus was created by evaluating and labeling actual Twitter posts using the same criteria that an intelligence analyst would. This includes having the ability to follow through with the threat. On Twitter, there are very few examples of legitimate death threats. If someone says they are going to kill someone, it is usually a facetious response to something, and quite often a reference to video game play.

There are, however, plenty of legitimate DDoS threats on Twitter--which is something intelligence analysts look for. A post that would be flagged as suspicious is, “I am going to DDoS your XBox game console.” While intelligence agencies don’t pursue people for DDoS attacks on gaming systems, it demonstrates the capability and willingness to engage in a type of cyber attack that could also be waged against governments, law enforcement agencies, or financial institutions.

CZ: Have you thought that while people are using your site (that wants to accurately replicate the behaviors of some NSA-like filter) checking their horrible posts about throwing bombs here and there, indeed they are being watched and flagged by a real agent? Should I expect Bayes' pal knocks on my door?

DC | JG: People are at no greater risk using the CSIA than they are with other sites. In fact, most US social media sites have agreements that allow intelligence agencies gain access to their user’s data: XKeyscore is one example of this. According to files leaked by Ed Snowden, if an intelligence agent has someone’s email or IP address, they can use XKeyscore to perform real-time surveillance of that person’s entire internet activity, including search terms and website visits, the content of emails, private messages, and social media chats. So, in all likelihood, the data in our system has already been collected and reviewed by governmental and private intelligence agencies before we ever see it.

Our intention is to allow people to engage in the process themselves to see what is already happening. We believe that these systems frame social media posts in a way that implies suspicion, and we want to expose the ways in which these systems result in false positives. For example, if you are an intelligence agent reviewing posts collected by these systems and you are presented with multiple images of terrorist propaganda followed by a post that seems violent, but actually turns out to be song lyrics, will you be able to see it objectively? How can legitimate plots be distinguished from propagandistic threats without falsely identifying innocent people?