The Algorithmic Other

In Margery Williams’ The Velveteen Rabbit, the worn Skin Horse explains what it means to become Real. “Real isn’t how you are made,” he says to the Rabbit. “It’s a thing that happens to you. When a child loves you for a long, long time, not just to play with, but REALLY loves you, then you become Real.”

“Does it hurt?” asked the Rabbit.

“Sometimes,” said the Skin Horse, for he was always truthful. “When you are Real you don’t mind being hurt.”

Realness, in their world, is something that emerges through relationship: through a specific being engaging with you over time, not perfectly, not efficiently, but persistently. You can’t shortcut it.

I’ve been thinking about how many of our technologies, from social media to AI, are designed to do exactly that — to shortcut the sometimes uncomfortable process of coming to know one another. One place this is especially vivid is in the tools we've built for parents. The parent-child relationship is where we first learn to relate to another person — to understand someone implicitly, to develop intuition through repeated, imperfect contact. If that foundational relationship is being mediated by algorithmic comparison from the very beginning, it's worth asking what expectations we're building — not just about parenting, but about how we relate to others in general.

Last year, I interviewed a handful of mothers of young children about their experiences with parenting apps and devices. I was interested in their experiences of the tradeoffs between convenience and reassurance, and the potential for anxiety and dependency. Instead, I heard not only about those, but also something bigger: about broken instincts and the creeping feeling of relating to your own child as a data problem rather than a unique individual.

Daniela (all names changed) had an eight-month-old daughter. She had been using Huckleberry, a tracking app that logs an infant’s sleep, feeding, and diaper output, and offers a premium “Sweet Spot” feature that alerts parents when their child should be put down for a nap or woken up, based on algorithmic recommendations. She described a moment of reckoning:

“One day I realized how the timer stressed me out. My kid’s wake window is too long and she’s not showing any tired cues yet? Ahhh! So I deleted the app. I’m still clueless, but she is somehow doing fine.”

The word that strikes me is “clueless.” Not because Daniela lacked awareness of her daughter — she clearly noticed the mismatch between what the app recommended and what her child was actually doing. But the app had become the authority. Rather than building her own sense of her daughter’s patterns over time, she’d been outsourcing that judgment to an algorithm, and came away feeling worried that her daughter was the problem. When she deleted the app, she had to sit with the uncertainty of being “clueless” — even though her daughter was, by all evidence, fine.

Grace, whose son had spent his first weeks in a sensor-filled NICU, described a similar dependency. She had continued the intensive monitoring at home with tracking apps and was candid about the cycle it created:

“It keeps me on track… I love it! However, I feel like it also makes me paranoid especially with the baby’s daily feeding. If he eats a little less that day, it stresses me out, but the thought of not tracking makes me feel even more stressed out! I feel like I need all this data for some reason.”

Her pediatrician had warned her about exactly this: that the monitoring devices encourage a checking behavior to soothe anxiety, versus coping with anxiety through thought change. “You feel better initially, but the minute your anxiety ramps up again, you’ve gotta check it.”

Grace knew this intellectually. She described it clearly. And she couldn’t stop.

“I want to quit but fear it has already ruined me. Even if I stop logging his sleep, I’ll still be counting it all in my head and trying to hit certain goals.”

The philosopher Martin Buber distinguished between two ways of relating to another being. An I-Thou relationship is a genuine encounter with a specific person — reciprocal, unpredictable, accepting of flaws. An I-It relationship treats the other as an object: something expected to behave according to a pattern, measured against a standard, valued for conformity rather than individuality.

What I kept hearing in these conversations was a technology-mediated slide from I-Thou to I-It. Not because these mothers didn’t love their children — they leaned on technology precisely because they loved them fiercely — but because the tools they relied on were structured around comparison to a norm. The app doesn’t know your baby. It knows a composite baby, an algorithmic average, and it tells you when your specific child deviates from that average. Over time, the deviation starts to feel like failure — the baby’s failure to conform, or your failure as a parent.

Louisa, whose first daughter was born on the day the WHO declared the COVID-19 pandemic, described this plainly:

“It became just this intense world of tracking, right? I felt like my baby was abnormal or something was wrong… if we weren’t consistently hitting the recommended levels. Like, even though the minimum is six diapers, today we had to be doing at least eight.”

Grace arrived at a similar realization from the opposite direction:

“I used to panic if my baby didn’t eat enough in the 24-hour period. The thing about babies is that they’re just like grown-ups. Some days you just aren’t as hungry as others.”

That recognition from Grace is an I-Thou observation. It sees the baby as a person with moods and variation, not a system that should be producing consistent outputs. But it took months of anxious tracking for Grace to arrive there — and by then, the tracking mindset had become hard to shake.

The point here isn’t a simple binary about technology being bad and instinct being good. These tools exist because the conditions of modern parenting can seem genuinely impossible. Louisa, reflecting on her family’s isolation during the pandemic, was honest about why technology seemed like the best available solution:

“I think our family would have done a lot better in a world where the mom and her mom and her sisters and her community come and just whisk them away for eight weeks, and then come back. …but that’s not going to happen, so instead, we’ll try this technology. We’ll just try it. We’ll just make it work.”

The parenting apps filled a void left by the erosion of intergenerational community. That void is real; technology is largely not to blame. The question is what kind of technology fills it — tools that connect you to a standard or set expectations of a flawless It, or tools that connect you to other people, the imperfect Thou.

The psychoanalyst Donald Winnicott developed a concept he called the “good-enough mother” — the idea that children don’t need perfect parenting, but rather a caregiver who starts with close attunement and gradually, naturally, adapts less completely as the child develops their own capacity to cope. In Winnicott’s framework, the only truly inadequate parent is one who demands that their child conform to an ideal rather than seeing them as a specific, developing individual.

The parenting apps are structured around exactly the opposite premise. They define the ideal and surface every deviation from it. Huckleberry’s marketing promises that “our algorithm does the work for you.” The work, in this framing, is knowing your child — and it’s been outsourced.

This goes far beyond parenting. It shows up wherever technology mediates a relationship that used to require tolerance for uncertainty and attunement to a specific other. When AI companions or tutors are designed to be endlessly patient, affirming, and frictionless, they create interactions that feel good in the moment but don’t build the capacity for the harder, less predictable work of relating to actual humans. When productivity tools promise to optimize collaboration, they sometimes also remove the productive disagreements and shared uncertainty that make collaboration generative.

In each case, the technology isn’t malicious. It’s optimizing for something real — convenience, consistency, measurable outcomes. But the optimization treats friction as a bug when it might be a feature. The discomfort of not knowing whether your baby is sleeping “enough,” the awkwardness of asking a classmate for help instead of asking an AI, the tension of disagreeing with a collaborator — these are I-Thou experiences. They require you to sit with another person’s unpredictability and develop your own understanding and judgment in response. The I-It alternative is smoother, faster, more legible. And something is lost in the trade.

On rainy weekends, we send an open invitation to our neighborhood WhatsApp group for a play date. One family brings pizza dough, another brings crackers, someone else brings sliced fruit. It started with two or three families and has grown to over fifteen. The technology is simple — a group chat — and what it creates is the opposite of the tracking apps: not a comparison to a norm, but a connection to specific, real, imperfect people. You can look away for a moment. Someone else is watching, helping. Nobody is optimized, and everybody is good enough.

The Skin Horse’s test was simple: does this relationship make you more Real, or more carefully kept? It’s not a bad question for technology design either. In parenting tech, in AI, in any tool that sits between one person and another — are we building toward genuine encounters with a specific, unpredictable other, or toward the comfort of an algorithmic stand-in?

Big thanks to the parents who shared their experiences with me, and to Sherry Turkle for feedback on an earlier version of this essay.