‘Jules and Jim’ shows essential friction. Credit: MOMA
Much of what we find meaningful in relationships we might call friction overcome. Great art reflects this. In the 1962 François Truffaut movie Jules and Jim, a French and a German soldier who are best friends are separated by World War I; in Richard Linklater’s Before Sunset, released in 2004, two lovers separated by distance and coincidence struggle to close the gap; in the most moving of Hergé’s Tintin books, the eponymous boy reporter treks all the way to the Himalayas to save his friend Chang from the wreckage of a plane crash. These and numerous other texts of the kind remind us that overcoming difficulty is what makes these relationships meaningful.
Real-life friction was valuable, too. Only a few decades ago, we had to memorize phone numbers and addresses, write letters, make phone calls. And there was a romance to this. Knowledge of another person was earned, not taken for granted.
But things are changing. Last month, Meta debuted its agentic personal assistant, Muse, largely a clone of earlier open-source Clawdbot agents, which became popular with programmers and entrepreneurs last winter. Muse can send email, book travel, fill out forms, shop, and schedule appointments. It can order from Facebook Marketplace, set goals for you, and make phone calls on your behalf.
During the rollout of Muse, Mark Zuckerberg inverted the old logic of Facebook and Instagram, which once offered to “bring people together.” Now the technology is intended to save users time and energy so that they can be better fathers, mothers, children, and friends, in real life. Let the agents handle the small stuff: finding recipes, ordering ingredients, remembering birthdays, remembering dietary restrictions, and so on. Weirdly, Meta conflates learning about people with doing more thankless and impersonal tasks like ordering plane tickets or filing quarterly taxes. Convivial tasks like cooking and being with people are conflated with trivial tasks like dealing with byzantine bureaucratic processes. Maybe for some, there is no distinction anymore.
From a cybersecurity perspective, the mass deployment of autonomous agents poses profound risks both for individual users and for digital systems, particularly financial systems. But it is also worth questioning the new proposition, even in the ideal scenario in which these systems are totally safe and work as intended, and assuming we use them as intended. Will we actually become better friends and family members? Will society, an automated society, a frictionless one, function better than alternatives past and present?
There is reason to be skeptical. Tech companies have always promised that frictionlessness and ease would lead to more connection; and that these digital so-called connections would lead to better lives.
The release of Muse is a tacit admission that this was never true: being on a screen locks you into a screen; having a lot of digital friends doesn’t mean you have a lot of friends. The technology is merely making the same promise in a different way — you’ll be more connected if you use this. The core belief that friction and inefficiency in human relations are bad hasn’t changed, but merely the proposed solution.
Most rational people would agree that there are many small tasks that we would prefer to be automated. And most would also grant that getting away from the screen and living in the world, away from the predations of an increasingly complex digital ecosystem, is good. But we must also be aware that our reliance on digital tools often has weird secondary and tertiary effects, including dependence on the tools that are supposed to make us free. The Jevons paradox states that gains in efficiency tend to increase, not reduce, total consumption of a resource; Xerox means more incentive to copy pages; the more roads we build to decrease traffic, the more congested traffic becomes, and so on.
Meta hasn’t merely invented a nifty tool that will do some previously annoying tasks for you. Meta is suggesting that we use Muse to solve an interpersonal problem, alienation and loneliness, that has been partially caused by other Meta products.
There’s something counterproductive about giving an agent the task of reminding me who someone is or how to connect with them or what their needs are, because asking an agent to do that means I haven’t done it. It implies that I am not really willing to do that, I am not willing to internalize facts about another person. I am outsourcing that critical function to the machine; the agent performs the acts of caring, and we take the supposed credit for it.
It is true that the soul is not built up by doing comparative searching on airline tickets, but just as surely something is lost when the work of actively participating in our relationships is outsourced to an autonomous algorithm.
According to the philosopher and critic of technology Ivan Illich, technologies are disabling because they incentivize us to stop exercising physically, mentally, or both, in service of a task. Every time we adopt a tool, there is a trade-off: what we gain in speed and efficiency, we lose in terms of personal capability. In this sense, Zuckerberg’s claims are classically Illichian. He’s saying that Muse will save you time, stop you from wasting time exercising your brain on pointless tasks, and give you the freedom to connect with people — but he’s eliding that anyone with a screen and social media has already been disabled; their real-life social links are already atrophied.
So whether we’re applying the Jevons paradox or just thinking about disabling technologies, there’s very little good reason to assume that Muse, or any other kind of autonomous agentic helper, will incentivize us to return to the world and to formal, convivial relationships, just as there’s similarly little evidence that the time saved by Uber, DoorDash, or Hinge has been meaningfully applied to anything other than more scrolling and swiping.
Zuckerberg’s promotional frame for Muse is that life has gotten too busy and complex to maintain and nourish friendships; but its hidden assumption is that most people don’t really have the personal tools for doing so anymore. Technology is presented as the solution to the disabling technology.
Autonomous agents do present some modicum of relief from the digital quotidian, but the quantitative gains in time saved will have no qualitative impact unless we acknowledge the alienated, lonely state that previous waves of digital technology have left us with and distinguish between instrumental uses for autonomous assistance and human practices that should be carried out by humans. Zuckerberg’s cozy picture of our assistants helping us know each other and socialize more likely promises to further disable us and further immiserate us; it is more dystopian than it seems on its face, as with almost all Meta products — though Meta products may already seem quite dystopian to some.
It might be useful to think of the practice of friendships and relationships as an art that can’t be gamed or optimized and can’t be handed over to an intelligent algorithm. If an assistant is handling taxes and flights, that is fine, but make it a good reason to head down to the bar or the park or the cafe to try to have a conversation with another person, leaving all screens at home. If we can still find the cafes without our phones, of course, is another question. If we can still form friendships, find romance, write letters, and fall in love with the challenging and unknown is another question as well, and of a higher order entirely.
