top of page

🧠 Synthetic social media

Writer's picture: Charley JohnsonCharley Johnson

Internet users, brace for the emergence of E-Swift



In May of 2022, I tried to look around the corner and imagine the future of social media. I argued that ‘creator coins’ or ‘social tokens’ might one day turn creators into stocks and fans into investors. This hasn’t happened yet, though many well-funded projects are still trying to make this future a reality — stay tuned. In the meantime, AI has replaced crypto as the technology du jour, and today we have a new future to contend with: one in which every creator has their own personalized AI chatbot, and it behaves like an online representative, with its own social media profile that interacts with users — and other bots. This future might not be too far away, so let’s dig in.


Let’s say Taylor Swift has a new chatbot, ‘E-Swift.’ Much of the training data that shapes E-Swift would come from Taylor’s social media posts. But E-Swift wouldn’t be trained solely on this data. It wouldn’t be enough to feel ‘real,’ so Facebook and others would crawl the web for additional data sources. This would leave E-Swift vulnerable to regurgitating or inventing something off-brand, or just plain misinformation. If I were a celebrity, I would want to know everything about my li’l AI agent — the data it is trained on, how the model was built, including the culture and incentives of the company that shaped it. Perhaps one bank-shot silver lining of this imagined future ends with celebrities and creators in solidarity with researchers advocating for data access and robust audits. A guy can dream ya know?

👉 Taylor Swift might also want to read this essay to understand why data access and algorithmic auditing are important but incomplete steps along the path from transparency to accountability.

But let’s take this celebrity AI agent concept further: what if these agents were continuously learning from interactions with their audience? Audience interactions would continually adapt to the way E-Swift responds and behaves. Of course, this leaves E-Swift open to exploitation by trolls, who will undoubtedly get the agent to broadcast heinous things in public forums.


Taylor Swift herself would have been sold the idea that they could align E-Swift to her values and voice, but in the real world, that’s impossible. If the audience is continually training the bot, you will get nice nuanced updates for some, and problematic outputs for others. But in either instance, creators lose control over the thing that matters most: their brand.

Now we’ve arrived at the crux of it: the existence of E-Swift links the audience and creator together in a networked performance. In this world, the inputs that shape the creator’s public performance change. Sociologist Erving Goffman went so far as to say that “the audience’s way of shaping a role for the performer can become more elaborate than the performance itself.“ What was once a one-to-many exchange of solo performances for likes and cheers has now become an ongoing conversation between fans and a representation of a celebrity that will never run out of time or energy. It has become a dynamic dance between many interconnected yet very distinct groups: engineers, technologists, past data inputs, new data generated by audience interaction, and Taylor Swift herself all contribute in a networked fashion. The performance emerges from the network.


One framework that might help us make sense of this phenomenon is something called ‘actor-network theory’ or ANT, which was conceived by critical theorist Bruno Latour. ANT contends thatin a man-object association both agents modify their actions symmetrically and it is impossible to think about the action of the parts separately.” Basically, Latour comes along and says that dualisms are dumb and that there isn’t a clean binary that separates ‘technology’ and ‘people.’ They’re entangled in one another. As a result, “no unmediated action is possible” but “action is a property of associated entities.” Or, in our case, E-Swift is the product of a network of people, data, and technology interacting with one another — it is not a monolith.


Part of the value of the ANT framework is that it makes visible labor inputs, that were otherwise invisible — we can’t just ignore the engineers, product managers, and trust and safety professionals in the making and curating of E-Swift’s performance. As Latour writes, ignoring all these groups forces us to rely on “delegated actions that themselves make me do things on behalf of others who are no longer here and that I have not elected and the course of whose existence I cannot even retrace.” So the decisions made by those building E-Swift will reverberate in its performance in ways that aren’t easily understandable or predictable.


But what about for the rest of us — the audience? The big thing that shifts in this world is our epistemology, or how we know what we know. As media theorist and critic, Neil Postman, wrote “Every epistemology is the epistemology of a stage of media development. Truth, like time itself, is a product of a conversation man has with himself about and through the techniques of communication he has invented.” Right, the type of media shapes how we know what we know. Postman was writing during the era of television, warning — it turns out, prophetically — about the transition from writing and printing to television and, in turn, the substitution of entertainment for truth. Toward the end of “Amusing Ourselves to Death,” Postman wrote:

“When a population becomes distracted by trivia, when cultural life is redefined as a perpetual round of entertainments, when serious public conversion becomes a form of baby talk, when in short, a people become an audience and their public business a vaudeville act, then a nation finds itself at risk; culture death is a clear possibility.”

The fact that we refer to ourselves and others as ‘an audience’ already reflects a prior shift in our epistemology. Currently, our information consumption is entangled in a passive performance. Individuals post, and many watch and react. But the introduction of AI-creator chatbots triggers a new epistemological shift towards a more active, networked performance. Social media activity from a celebrity bot would be produced collaboratively by a range of participants (audience, creators, designers, etc.), which reach so far and wide that it will be hard to truly make sense of whose inputs have resulted in which outputs. This will lead to less creator control and more opacity under the guise of ‘innovation.’ Does the bot seem to dislike people with ginger hair because that represents the feelings of the fanbase? The engineers who built the bot? Or is this how the person the bot represents actually feels? Or is it just the result of trolling? Who can tell! That’s the epistemology of a networked performance.

0 views0 comments

Recent Posts

See All

Comments


Let's Chat

I'm curious about

Find me online

Logo for the newsletter Untangled
Logo for The Atlantic Council
Image of book, AI Untangled
Logo for Facilitation Leadership Lab
Logo for Data & Society Research Institute
LinkedIn Logo
Award announcement for Disinformation Action Lab

© 2024 by Charley Johnson. All rights reserved.

bottom of page