The Outsourced Self, Updated
One of the most active areas of large language model (LLM) development is that of agentization, or basically, giving models the ability to do things. Generative LLMs take text input and generate text output, which already encompasses a whole lot of tasks: answering questions, producing creative material, summarizing or explaining concepts. But by default these models can take no other actions besides generating text, and that naturally limits their capabilities. Often people expect AI assistants to be able to do things like set reminders and alarms — no doubt because of the longtime existence of voice-controlled integrated assistants like Siri and Alexa — but in practice these actions are distinct skills that LLMs must be trained to do.
Still, there is hardly any doubt that that is where the market is moving. Full-fledged concepts of AI assistants involve AIs booking travel and ordering gifts on our behalf, negotiating rates, making decisions for us. Even without new actions, models have started to serve as surrogate consultants, therapists, librarians, and a whole host of other duties normally fulfilled by paid professionals. These trends, early as they are, reminded me not of any science fiction novels (my reading in the genre being pretty meager), but a book that I was first assigned to read in high school, The Outsourced Self by Arlie Russell Hochschild.
The book struck me as an odd choice to assign even at the time — it is about social and technological trends, and it was by no means inappropriate, but my classmates and I had no experience with many of the specific domains that are discussed in the book, like childcare and elder care, matchmaking, and family counseling. The concept is that many functions that were by necessity accomplished within the family or “the village,” the immediate physical community, typically on goodwill or in a bartering of services, have now been commoditized into services that one can, and often should, outsource to professionals.
The Outsourced Self was published in 2012, more than a decade ago. At the time, Hochschild was concerned with the commercialization of the most intimate moments in our lives. Her book details the rise of personal assistants, wedding planners, and surrogate mothers in functions that were once squarely in the domain of the home, but have moved into the open marketplace. In the last decade, some of the practices that Hochschild documents remain pretty fringe (nameologists who help you name your child, for example), but others have been increasingly normalized.
Various economic and social factors have contributed to immense demand for these services. Fewer young people live near their parents or extended family, needing to live where their jobs are; these same people are working longer hours than previous generations; public childcare options are extremely rare until kindergarten, a full five years into a child’s life; people are living and requiring medical care for longer at the end of their lives; and religious and community organizations that have filled some of the gaps in the past are by and large in decline. As Hochschild notes, there is outsourcing for convenience and by necessity, by the upper class and the working class alike, and both have stigma attached.
There are some tasks we have no trouble outsourcing without a twinge of guilt. No one I know would chastise me for using a laundry service, even if I might feel bad, since I’m perfectly capable of doing it myself. The outsourcing of childcare is incredibly common, and simply necessary for most families (at least in the United States), but it’s clear that there is more judgment and guilt associated with outsourcing care for children or elderly familar members vis-a-vis housecleaning or dogwalking. Some interviewees described the implicit narrative that to pay someone to, for example, visit your elderly father at his nursing home for you is lazy at best and immoral at worst. But the reality is that many people do not have the means to meet those needs otherwise, whether due to professional, financial, or emotional reasons.
In the best cases the market really does seem to fill a niche: the stressed-out daughter who lives across the country from her parent and has two children of her own can hire someone and give them a job, maybe even someone who will find joy and purpose in the work. In other cases, the incessant professionalization of intimate relationships seems to serve no purpose except to make people with disposable income feel inadequate. Hochschild writes, “The more anxious and isolated we are and the less help we receive from nonmarket sources, the more we feel tempted to fill the void with market offerings…Our imperfect, homemade versions of life seems to us all the poorer by comparison.”
The most extreme example presented to readers is Gloria, a young professional in Miami who attempts to convince Hochschild that everything is better when paid for, including friends. “You need someone to talk to, someone you can be completely honest with, and someone who will ask nothing whatever of you,” Gloria says. “Friends have problems of their own. And they have their own opinions. You need a blame-free zone so you can really look at yourself. And for that you have to pay.” I found the quote particularly sad, because although she might have a point about the value of blame-free introspection, she seems to fundamentally misunderstand what friendship is. Her description of what “friends” should do (listen to your problems) and shouldn’t do (have their own problem, or ask anything of you) does not resemble a relationship between equals: the way Gloria puts it, friendship really is a service.
Of course, the other side of this economy are the people these jobs are outsourced to, people that are often overworked and underpaid. Not the matchmakers for the 1%, of course, or the many professionals in the outsourced space that cater to the idle rich. But certainly the people who work in nursing homes and daycares, or who scrape together gig work. And so maybe it wouldn’t be such a bad thing if these roles, both exhausting and incredibly important, could be filled in another way. Hochschild investigates the interpersonal dynamics between a customer and a provider of “intimate” services, referring more or less to those that cannot be executed as a mere business transaction, but require the formation of an emotional relationship. But what happens if those dynamics are not interpersonal at all, but between a human and a machine?
You can’t hire a robot to babysit your kid (yet), but if you live in San Francisco, you can call a self-driving car to take you to your appointment, the airport, a night out. Thanks to the gig economy, you can hire someone via an app to bring you groceries, hang your picture frames, and yes, walk your dog. On some college campuses, robots are making deliveries; in Houston, drones are dropped Amazon packages onto driveways (but only nonbreakable items weighing less than five pounds, and only if the homeowner has first placed a target for the drone to aim at. Seriously.) In a way these services remind us of the failures of AI and robotics. They do use AI to, among other things, match customers to the humans actually performing these tasks, but it turns out that even mildly complex tasks in the real world require planning or dexterity that is well beyond current commercial capabilities.
But in other ways, AI is succeeding to a surprising degree. People already do use AI for not only the rote or trivial tasks, but the creative and emotional ones, too. Large language models, the type of AI that I work on, have been used to write best man speeches. They have outperformed doctors at producing scripts for discussing addiction treatments for patients, as judged by a panel of doctors. Yes, AI might have a better bedside manner than the average MD. People are increasingly turning to AI chatbots for entertainment, friendship, and even pseudoromantic relationships. Like outsourcing intimate tasks to paid professionals, using AI as a friend or romantic partner is stigmatized, to say the least. I am not a medical professional — far from it — but the practice also strikes me as potentially worrisome from a mental health perspective. Yet some studies have also shown that for people who suffer from loneliness, talking with chatbots can boost their confidence in social settings and improve their moods.
I would argue that a chatbot, even one highly capable of being an equal conversational partner, cannot be a friend. Neither is anyone paid to listen to your problems while never burdening you with any of their own. There is an imbalance in outsourcing by the nature of the relationship existing in a services market; there is an employer and an employee. And that doesn’t mean that these relationships cannot be mutually beneficial and respectful. Some long-lasting ones become something definitely more: a guardian who spends years caring for the same children, for example. In The Outsourced Self, the outsourcing of intimate tasks is shown to work beautifully in several cases, but what those have in common is the development of trust and connection between the parties involved. Therefore, even in a world where AI is embodied in robots that can move freely and easily, this type of outsourcing should warrant caution — the job requirements might require than we realize at first blush.
Hochschild interviews several longtime nannies and personal assistants who understand this well. One successful and experienced assistant observes the following about her role:
A lot of my clients have been thirty-something dot-com executives and they talk very fast. Rata-tat-rata-tat-rata-tat. “Call Jim at the office…tell him we need the order by eight a.m. tomorrow…” [She snaps her fingers, snap, snap, snap.] So I call Jim’s office. I’m friendly with his receptionist. I talk to Jim himself, answer questions, and make sure he’s got the message straight and is in a good mood about it. I respond to any hesitancy or resentment I sense in his voice. I’m patient. My clients outsource patience to me. And once they get in the habit of doing that, they become impatient people.
The Outsourced Self traces the takeover of the market, then encroaching from the professional world into our intimate, personal lives. Technology has followed that encroachment, and now not only are these services commercialized, they are increasingly automated. The promise of AI is liberation: LLMs and other tools, like countless technologies before them, claim to speed up or automate busy work and leave people with greater time for life’s greater pleasures. That future of freedom hasn’t yet manifested; instead, people by and large eke out productivity gains through outsourcing both human and machine. Instead, it seems like we might be headed for a future where everyone is impatient: we’ll have AIs to handle our phone calls and bookings, that can write our notes and send gifts on our behalf.
Hochschild’s study ends not with castigation — she knows there is no putting the genie back into the bottle — but with caution. “Confining our sense of achievement to the results, to the moment of purchase, so to speak, we unwittingly lose the pleasure of accomplishment, the joy of connecting to others, and even, in the process, our faith in ourselves.” Widespread outsourcing has resulted in the decline of specific skills (think of sewing and tailoring, for example) and maybe even broader traits, like the patience the personal assistant claimed to handle for her clients. Not everyone needs to sew their own dresses and grow their own food anymore, and that’s a great thing. But over the next few years it’s worth taking time to consider, if and when we let a model handle a task we’re used to doing ourselves, what it is that we’re really giving up.