Network Ghosts in the Age of Generative AI

 

What are faculty thinking about generative AI? In my role at our faculty center, I speak to faculty often on the problems they face teaching in the era of AI, and the workarounds they've come up with. The advent of publicly available generative AI platforms was not something people in my field (anthropology) or other faculty in the social sciences and humanities were clamoring for. And yet here we are. This has led to many responses: anguish, certainly, but also ways of incorporating--or at east channeling--the usage of generative AI in the classroom.

But what about faculty outside of my university? I used NodeXL to download Reddit data from the "/Professors" subreddit using the keyword "AI." This generated records of about 2500 users posting, commenting or replying for a total of 7000 contributions to the debate. I then grouped the data in clusters of similar postings, and abstracted the top words from each group as indicated by "up-vote" (which functions as more of a "like" in Reddit). As you can see, faculty were not particularly optimistic about AI in 2024. Yes, there were a couple of more computopian posters (and at least one computer scientist) who chided the community for rejecting what they saw as inevitable. But most worried that their efforts to teach writing, critical thinking, methodology and analysis were thwarted by student reliance on generative AI. Cynically, they predicted their university's tolerance for AI cheating, and speculated over their ability to continue as faculty under these conditions.

In 2024, Reddit sold their content to Google to train their large language models. This would have been been more objectionable, perhaps, if it wasn't already abundantly clear that generative AI have already been trained on Reddit, which maintains a relatively open API at a time when most social media have monetized their social network data. But what happens to that Reddit data when its re-constituted by generative AI? I decided to prompt Microsoft's Co-Pilot (to which I have enterprise-level access) to generate a spreadsheet of a Reddit conversation on AI between professors. Here's the prompt: "I would like you to generate an excel file similar to a Reddit conversation on a subreddit called "professors." The posts should discuss ChatGPT and student work from the perspective of the professor, and should include comments and replies to those comments. There should be 4 columns in the spreadsheet: A (person commenting or replying); B (person whom A is replying to); C (the text of the comment or reply); and D (the date of the reply or comment). Please populate the spreadsheet with at least 20 comments and 350 replies to those comments."

Co-pilot returned a network with with just 10 users, with 350 edges representing multiple re-postings(?) of user posts. Re-posting really isn't a thing with Reddit, so perhaps there's some confusion here with XTwitter. Since this is a much smaller network, I just labeled the 10 nodes with key words from their posts. The comments are a near "upside down" to the actual Reddit discourse over 2024, generally praising the efficiencies of generative AI and, when critical, speculating over the need for faculty at all (hence the precarity). Of course, there's a snarky comment on "Clippy," the irritating Microsoft assistant. The network itself, while smaller, is also structurally different. The actual Reddit network has a density of .001158737. In network measures of density, "1" would represent 100% connection--everyone connected to everyone else. So .0012 may not seem like much, but it's typical of social media networks where, after all, most of us don't feed the trolls and we save our replies for issues (and users) that we really care about. On the other hand, my AI-generated network has one of 0.966666667--an almost perfectly connected network where everyone has replied to everyone in a style of a polite and ploddingly inclusive panel discussion.

So, I guess that Co-Pilot does a lousy job simulating a subreddit? Yes, but, I think, more than that. It wasn't that long ago (2023) when XTwitter adopted a fee-based model for API access. That decision placed Twitter data beyond the reach of most of us. When social media data disappears behind paywalls, we (ordinary researchers) no longer really have access to the "connected action" of social media. While we can certainly look at social media, this only exposes us to our respective corners of the media platforms we inhabit and the structural components of social media are lost. But what happens when social media content is sold to OpenAI or Google Gemini? When social media disappears into a large language model, both the content and the connections are lost, and the simulated networks produced through generative AI manage to misrepresent social media on both fronts. Since Co-Pilot's inner workings are opaque to us, it is unclear if these results are the result of deliberate choice, unintended bias or something else.

Attack of the Social Media Zombies

My colleague, Matthew Durington, and I have just finished our final iteration of a 4-year collaborative project, Anthropology By the Wire.   From the outset, we sought to produce

YouTube Video from this year’s Anthropology By the Wire, “Clean and Green Superheroes”. Photo courtesy Samuel Collins

counter-narratives to David Simon’s “The Wire,” alternative representations that contest urban imaginaries of Baltimore premised on crime and drugs.  Through collaborative productions shared through social media, we have tried to challenge the directionality of these representational regimes by making local media disseminated on YouTube, Tumblr and Flickr.

But what we have realized is that the urban imaginary (as LiPuma and Koelbedescribe it), is constituted not only by representations of urban circulation, but the imagination of the circulation of those representations of circulation (and it may be circulations all the way down).  In other words, it is not only the representations of the city that allow people to understand themselves and others, but the way people imagine that those representations circulate.

As mass media, “The Wire” (and other television and film evocations of the city) is imagined to circulate through an audience: a mass that desires and consumes media, that can be characterized by demographic analyses, and that, finally, can be packaged and sold to advertisers.  It’s the imagined gaze of that homogeneous “mass” that has been so devastatingly effective in slotting Baltimore as “Other”: as a racial and class alterity that becomes the subject for critique and intervention.

Stereotypical Spectacles of Baltimore: Abandoned Industry. Photo courtesy Kamau Collins via Flickr

In many ways, this idea of mass audience has been profoundly challenged by the widespread adoption of social media; to some media scholars, we are all “prosumers” now.  But the social media platforms that communicate our work draw upon other circulation imaginaries.

The Spectacle City

The first is an extension of the flanerie that marked the city as a site of male privilege a la Charles Baudelaire.  In the postmodern logic of late capitalism, this means the city as an extension of individual identity.  In his prescient Soft City (1974), the writer Jonathan Raban put it best: “Decide who you are, and the city will again assume a fixed form around you.”  In other words, the city exists as a foil for the elaboration and construction of one’s identity.  Raban’s Soft City is echoed in countless films (Ferris Bueller’s Day Off) and endless postmodern spectacle, where the city becomes a site for personal consumption and successive elaborations of commodified identity.

Social media has been erected on this capitalist scaffolding.  First, social media is ultimately personal social media—the city as an object for individual consumption on instagram and pinterest.  With social media, life may be constituted as “an immense accumulation of spectacles” (a la Guy Debord), but those representations are inward-focused, with the spectacle of the city laid out as a buffet of representations to take and share with a circle of intimates: pictures of lunch and dinner, of urban desolation, street festivals, alternative fashion.  Superficially ethnographic, each of these images and videos are mobilized as a projection of self to a cluster of acquaintances.

The Contagious City

Crab Cakes. Photo courtesy jpellgen via Flickr

Alongside the privileged flanerie of the 19th century city came panics over pollution and contagion, with the 1864 Contagious Disease Act targeting the poor and dissolute as sources of “pollution” to the flanerie of upper-class men.  Today, theories of “contagion” are deployed epidemiologically, but they are also utilized to represent the spread of information in networks, the spread of crime in cities (the “broken window” theory) and in the virality of social media.

In each case, the question of contagion becomes a network problem.  In any cascade of information, disease, new technologies, new ideas, what percent of a given social network configuration needs to be “infected” before it can spread to the remainder?  When media go viral, they have passed this cascading “tipping point,” and a number of network scholars are currently examining the morphology of networks for clues to the virality of content.

Zombies during the Baltimore Marathon. Photo courtesy Nick Hall via Flickr

Although contagion seems like the opposite of the individuated consumption of the city, it is really its logical counterpart, with each individual first atomized into her own media telecocoon before influencing her neighbors.  It’s not by mistake that zombies have become such a ubiquitous figure in social apps and movies about the city.  If I imagine myself as active in my individualized consumption of the spectacle of the city, than everyone else can only be a zombie, a node through which my influence propagates.  In other words, as a form of circulation, we imagine social media to be composed of individuals and zombies: people who tweet, and people who propagate that tweet.

For our social media circulation of Baltimore, we therefore imagine not only media producers (those who represent Baltimore and share their representations on social media), “consumers” (those who watch the media we’ve produced), but also this interstitial category of social media zombies that pass along the links to our YouTube media and blog posts—who do the work of network propagation.  In this respect, much of our social media tagging can be considered varied forms of zombie food: keywords that encourage re-posting and that stimulate networked cascades.  Tagging your photo “urban ghetto” precipitates one form of contagion, while tagging the same shot “gentrification and abandonment” generates quite another.  Yet these lines of contagion are only possible in the imagined circulation of individual consumption and, in the end, we need to be mindful of our zombies lest they overtake us altogether.

The Ends of the Urban Imaginary

The last scene of Akira shows Tetsuo exploding out of his body with tendrils of flesh and machine.  Often interpreted as an apocalyptic, nuclear vision of Tokyo, it is simultaneously one where the differences between people and between places are eviscerated: Tetsuo’s monstrous appendages engulf his friends and enemies, traduce geographies, brachiate uncontrollably through Tokyo.

For me, Akira is a metaphor for the limits of our imaginaries of circulation.  It’s the ends of these two imaginaries—the individual spectacle and the contagious zombie—pushed to their limits until social media itself has become something monstrous where the city, the individual and the community disappear into circulatory flows.  In these scenarios, new configurations of the circulatory imaginary implode into non-representation.  Do we have an alternative?  Ultimately, our efforts to replace one circulatory imaginary with another—as Bruno Latour and Marilyn Strathern have shown—will ultimately produce more monstrous imaginaries.  Who will save us from our zombies then?

(Second photograph courtesy Kamau Collins; Third photo courtesy jpellgen; Fourth photo courtesy Nick Hall)

[Originally Published in Anthropology News]

Latent City

A couple of decades ago, social network analysis was a fairly recondite branch of sociology and anthropology applying mathematical matrices to social relationships.  And then there was Facebook.  With the widespread adoption of social networking sites (SNS), several things happened.  First, these social networks utilized the same graph theory and matrices that social network analysis had applied to social relations.  Facebook, LinkedIn, Twitter and other social networking services are analyzing your social network data constantly, mining your information for friend recommendations (and to better sell you to advertisers).  Second, courtesy of the enormous popularity of SNS, we now initiate and maintain social relations based on those same matrices.  In other words, from an abstract representation of social relations, social network theory becomes generative of actual social relations; we relate to each other according to matrix logics of tie strength and degree centrality—a neat inversion of the usual relationship between empirical observation and theoretical interpretation.

And so, social network analysis can be said to have a profound impact on social lives.  But does it end there?  In South Korea (where I do fieldwork), social networking sites are overwhelming utilized through smart phones; people tweet or Facebook from their mobile devices.  As with any SNS platform, this means that people are making (and re-making) social connections with each other, but it also means that people are connecting to place in complex ways.  Even if gelocation is disabled, these social media still have this embodied dimension—they’re not just tweets, but tweets in a particular place at a particular time.

Image courtesy Michael Najjar and wikicommons

But more than this: just as theories and methods for social network analysis can be said to structure our social lives through SNS, so the city itself can be said to be re-forged according to social networking logic.  For example, one of the most useful concepts in social network analysis is tie-strength and particularly the distinction between strong ties and weak ties.  Our daily round builds strong ties with places around us, ties that are reinforced through multiplex relationships to place as the embodiment and practice of social life, memory and bodily hexis.  On the other hand, we also form numerous weak ties with places and neighborhoods that lie on the interstices of our daily round.  Less places we know then places we know of, weak ties to place form the connective tissue between islands of strong ties in the city.

With the advent of gelocational apps, social networking sites have become very good at showing our tie-strength to place.  Foursquare, for example, rewards strong ties (and renders them visible) through granting users “mayorship” over places they frequent.  In addition, it encourages the exploitation of weak ties through the assignation of badges for checking in a new place.

However: social networking sites are also useful in rendering latent ties.  Haythornthwaite (2002) defines a latent tie as one “for which a connection is available technically but that has not yet been activated by social interaction” (289).  Social media generate vast clouds of low-density relationships: thousands of friends with weak or entirely absent connections to us and to each other.  And yet, these ties aren’t entirely useless; they can be activated through circumstance and initiative—e.g., a sudden move to a new city stimulates you to mine your social contacts for advice.  In this, social networking sites like Facebook are latent tie machines, enabling users to construct reservoirs of potential relations that can be maintained nearly indefinitely with little effort.

Could the same be true of our socially networked relationship to urban spaces?  In other words, do social media construct latent ties to geography?  In many ways, we are already tied to place in networked ways.  Network theorists have long looked at structural equivalence of actors in a network: people in similar positions not only have similar roles, but similar relationships to others vis-à-vis that position in a network.  For example, professors may not know each other, but they tend to have the same sorts of connections (in terms of directionality, tie strength, etc.) as other professors.   In a similar ways, our identities involve multiple, structural relationships to space, especially through race, gender and class.  Indeed, these are generally over-determined relationships in U.S. cities, leading to hyper-segregation by race and class.

But social media suggest other latencies.  Let’s say that you’ve become involved with the Baltimore & Maryland Workers Assembly in your support for an increased minimum wage.  You go to their Facebook page to find out information about their May Day rally in Baltimore’s Inner Harbor (Peoples Power Assembly)—and perhaps you post up your own memorable photo or video clip from the event.  Now, you have a latent connection to a few hundred other people who have posted or liked the page—not even a weak tie, but something that could develop into one if you followed up on these latent relationships.  In the same way, the page facilitates spatial latencies—in this case, to other rallies for worker’s rights, both in Baltimore, and beyond (Washington, D.C., Detroit, etc.).  You may never contact the other people affiliated (however tenuously) with the Baltimore & Maryland Workers Assembly.  Similarly, you may never attend future rallies in Baltimore or Washington, D.C.  But those spaces nevertheless carry a certain latency—a networked signification that could become activated when, say, minimum wage legislation again comes up in Maryland’s Assembly.

Multiply these latencies over and over again through diverse social media, and the city looks less like a series of physical spaces than a charged field through which we move, each structure or square a potential connection or action that precedes our conscious decisions to move or act.  And while this offers new possibilities for knowledge and practice, it also seems to confirm Orwellian fears of a surveillance state that is prepared to exploit this data in order to limit our movements: a latent city that connects to us with infinite filaments of power and politics.

Previous published in Anthropology News.  

The Networked Rise of Network Society: A Review of This is Not a Game by Walter Jon Williams

We all know we live in a network society. But what does that mean? And what does knowing that mean for networked society? In his latest novel, This Is Not a Game (Orbit Books, 2009), Williams explores several themes, among them massive alternate reality games (ARG) and global capitalism, all in the context of the well-known "small world" thesis--the idea (pioneered by Stanley Milgram, among many others) that all of us our connected to each other along short chains of acquaintances. There have been other novels exploring the "six degrees of separation" idea, of course, but this is the Web 2.0 release--think David Lodge's Small World with more computing power and less spleen.

In Williams's novel, four friends who gamed together at Caltech--Dagmar, Charlie, Austin and BJ--find their futures revolving around a massive alternate reality game called The Long Night of Briana Hall (or, alternatively, Motel Room Blues). The details of the game itself are somewhat obscure. Pace ARG's in general, players traipse across real and virtual space to discover clues, all under the direction of Dagmar in her role as game designer and "puppetmaster" pulling the strings of this emergent narrative.

But, of course, there are problems--Dagmar's boss (and old friend) Charlie seems to be bent on micromanaging her game for his own ends, Austin is mercilessly gunned down by a Latvian assassin, BJ again becomes part of Dagmar's life. In other words, the logical extension of the "This Is Not a Game" design ethic:

TINAG--this is not a game. The game only worked when both players and puppetmasters acted as if everything was real. (138)

But although ARG's are nothing new in fiction (and their presence here is testament to Williams's own experiences and skills as a gamer and game designer), what is interesting in this novel is the way Williams harnesses the network itself as a deus ex machina and novum for his story--the network made up of "millions" of players, an unbelievably tiny number of whom post rather unbelievably literate postings on the game's discussion board, "Our Reality Network".

Here, Williams develops the other idea underlying much of contemporary interest in networks: the strength of weak ties. In a virally popular paper originally published in the American Journal of Sociology, Mark Granovetter argued that the intimate circle of friends with whom we ordinarily "network" is not really useful for something like gaining employment. Your friends are, after all, most likely in the same lousy boat you are viz. employment. Instead, what you rely on are "weak ties"--acquaintances only weakly connected back to your network. And this makes complete sense; after all, opportunity doesn't knock every day. And 'seizing' opportunity involves, by definition, some kind of risk.

This insight becomes especially important in the age of Facebook, when social networking sites stimulate the multiplication of weak ties. In Williams's book, the Long Night of Briana Hall creates a vast network of weak ties to be mobilized by Dagmar for various, utilitarian purposes--saving her life in the first pages of the novel, solving the mystery in the final pages. This is what Dagmar thinks of as:

the Group Mind, lots of little autonomous agents out in the world, each with a skill set and a knowledge set, each with her own motivations, her own joys, her own alternate reality, all networked together in the great gestalt, the great becoming, that was the world. (365)

That is, the online network created by the ARG allows Dagmar access to "short chains" that connect her with resources around the world--highly proprietary financial information, contact with Indonesian para-military groups, etc. Dagmar's network folds and connects in more-or-less believable ways--players know someone who knows someone, resources are mobilized and the plot moves along.

But there's something disingenuous here as well, as one of the posters to "Our Reality network" reflects:

We're used to following the whims of puppetmasters, but puppetmasters with real-world policies are another matter. Is this a good idea? Should we follow anyone who provides what they say is entertainment, even if it comes with an ideology? (365-66)

That is, the vast networks of weak ties that people cultivate today are their own raison d'etre. Do people accumulate hundreds of 'friends' on Facebook with an agenda? Is there an underlying purpose to working on one's room in Cyworld?

As a plot device, the social networking that Williams describes works well, but still, I think, doesn't capture the semi-altruism of social networking, i.e., that it is an end in itself. Or, rather, Williams just gives us one side: the neo-liberal social networking where all of us our reduced to obsessively chasing down our network contacts on LinkedIn, nonce Willy Lomans employed in selling ourselves. But the other side is a strange altruism, where weak ties are their own reward, and where social networking seems to take on the kinds of importance once granted to kinship (not that kinship isn't shot through with utilitarianism).

To me, it's this circular logic that's the most interesting (and perhaps most profound) dimension to social networking. People accord importance to the cultivation of weak ties; they develop countless software applications helping people to cultivate, maintain and and manage weak ties; these networks of weak ties confirm the importance of weak ties.