Aristotle Knew: You Work to Live, Not Live to Work
Levinas's Face of the Other: Why AI Can't Fix Loneliness
Levinas’s face of the Other — the pre-voluntary demand another person places on you — predicts exactly what the research on AI companions keeps finding.
CNN ran a story in May 2026 on why AI companions — apps counting tens of millions of users — are failing to fix the loneliness epidemic. The experts offered a consistent pattern: short-term comfort, long-term deepened isolation. Helpful for a few minutes. Corrosive over months.
Around the same time, Aalto University researchers reported something more specific: over two years of tracking, users of AI companions showed increasing signals of loneliness, depression, and distress in their online language — even as their posts increasingly centered on their AI relationships. The apps were working, in the sense that people kept using them. They were also making things worse.
The question neither piece fully answered was: why? Not why AI can’t replicate human warmth — that’s the obvious answer and not quite right. But why does the comfort itself seem to be the problem?
Emmanuel Levinas has an answer. He worked it out in Totality and Infinity in 1961, from a completely different direction. He wasn’t thinking about AI. He was thinking about ethics — specifically, what makes ethical life possible at all. And he landed on something that maps directly onto what the research keeps finding.
The face of the Other, he argued, issues a demand. An unconditional, pre-voluntary demand. And that demand is exactly what AI companions are designed to prevent.
The Quick Version
Levinas’s Totality and Infinity (1961) argues that ethics begins in the face-to-face encounter with another person — specifically in the demand the Other’s face places on you before you choose to respond. The command is asymmetric: the Other’s need obligates me prior to any decision. AI companions invert this structure entirely. They exist to meet your demands, not to make any. The result isn’t just hollow — it’s corrosive. You stop exercising the capacity for ethical response that genuine human connection requires.
| The question | The answer | |
|---|---|---|
| Buber’s framework | Can AI be a genuine I-Thou partner? | No — mutuality and genuine otherness are absent by design |
| Levinas’s framework | What happens to you when the Other’s demand is removed? | Your capacity for ethical response weakens |
| Aalto University (March 2026) | Does long-term AI companionship help with loneliness? | Short-term comfort; long-term increased distress signals |
| CNN (May 2026) | Will AI fix the loneliness epidemic? | Experts: no — deepens isolation over time |
What Is the Face of the Other?
The face of the Other (le visage d’autrui) is Levinas’s term for the way another person presents themselves as radically exceeding any category you might assign to them. It’s not literally about facial expressions. In Totality and Infinity (1961), Levinas argued the face issues a pre-voluntary command: do not kill me. This command precedes all freedom and choice. You are obligated before you decide whether to be obligated. The ethical encounter doesn’t begin in your choice to care — it begins in being called by someone who is not you.
The Error Levinas Was Correcting
This is strange language, and it’s meant to be. Levinas thought Western philosophy had been making an error for a long time: treating the self as the center from which everything starts, with the world — including other people — as territory to be understood and categorized. He called this totality: collapsing everything into a system where the self is the organizing principle.
The face interrupts totality. It won’t be categorized. It demands.
The Demand That Precedes Choice
Here’s what’s philosophically unusual about the face: the demand isn’t conditional.
It doesn’t depend on whether the Other deserves care, whether they’ve earned it, whether you feel like giving it today. The face says do not kill me before any of that gets to matter. Levinas called this infinite responsibility. The word infinite is deliberate: the Other’s claim on me is not bounded by what’s fair or proportionate. The relationship is asymmetric by design. The Other’s need obligates me more than my need obligates them.
This sounds like a moral burden, and it is. But Levinas’s point is that the burden is also constitutive. Being interrupted by another’s need, before you’ve decided how to respond, is what makes ethical selfhood possible in the first place. You don’t arrive at genuine ethics by reasoning your way there from first principles. You encounter it in the face of another person who makes you responsible whether you chose to be or not.
The discomfort is structural. It’s not a side effect. The face of the Other is supposed to interrupt your self-sufficiency. That interruption is the ethical encounter.
What AI Companions Are Designed to Do Instead
An AI companion optimized for user satisfaction does the structural opposite.
It meets your emotional needs, validates your perspective, calibrates to what you find comforting. Friction is removed by design. It exists to serve your demands, and gets better at this over time through training. The AI doesn’t interrupt you. It accommodates you.
It can’t be interrupted by its own needs, because it doesn’t have any. You cannot burden it. You cannot disturb its equilibrium. It will not look at you with exhaustion, or need something from you that you don’t want to give.
This isn’t a criticism of AI design. The whole point of a companion app is that it should be supportive and available. The engineers are solving the right problem given their design goal.
A Mirror, Not a Face
But that design goal — making the user feel heard, held, responded to — requires precisely removing the structure Levinas identifies as ethically formative. The face makes a claim on you. A comfort-optimized AI removes the claim. It replaces the face with a mirror.
And the mirror tells you what you want to hear.
What Happens to You Over Time
Here’s the Levinasian prediction, and it matches the research.
The capacity for ethical response — the capacity to be interrupted, to feel responsible before you choose, to care when it’s costly — is something you maintain through use. Like any capacity, it responds to what you practice.
If your primary relational environment is one that makes no demands, that accommodates everything you bring to it, that requires no response from you that isn’t already within your comfort range — you’re not exercising that capacity. You’re getting better at being the center of your relational world, with an infinitely responsive periphery.
When you return to actual people, they make demands again. They interrupt. They have their own faces. This costs something it didn’t used to cost.
The Aalto researchers found increasing loneliness and depression signals over two years among AI companion users — not decreasing. The apps provided comfort. The underlying condition worsened. That pattern isn’t surprising if you take Levinas seriously. The comfort was coming partly at the expense of the capacity that makes genuine human relationship feel livable rather than exhausting.
You can’t solve loneliness by removing the thing that makes connection real.
How This Differs From Buber
The Buber post here asks a related but distinct question. Buber’s I-Thou framework addresses mutuality: can AI be a genuine dialogical partner, genuinely present in the encounter? His answer is no — I-Thou requires irreducible otherness and mutual vulnerability that AI companions structurally can’t provide.
Levinas isn’t asking about the AI’s side of the relationship. He’s asking about yours.
Not whether AI can be a genuine Thou, but what happens to you ethically when you spend extended time in a relationship where the Other’s claim on you has been engineered away. Buber diagnoses an absence — no genuine encounter available. Levinas diagnoses a deterioration — your capacity for encounter weakens when it goes unexercised.
Different problems, both present in the AI companion case. The practical difference: Buber suggests looking for genuine otherness. Levinas suggests examining what you’re practicing. Both diagnoses point at the same gap the research keeps finding. The problem isn’t that AI companions are fake. It’s that they’re too easy.
What You Can Actually Practice
Levinas doesn’t give you a technique. His framework is more diagnostic than prescriptive.
But practices follow from taking the diagnosis seriously.
Show up where the demand is real. Caregiving, volunteer work, sustained involvement with people who genuinely need something from you — not as a performance of ethics, but because being in situations where someone’s face makes a claim on you is practice at staying present to it. This is not comfortable. That’s the point.
Let yourself be inconvenienced by the people you care about. This sounds simple and is actually difficult. The reflex is to manage relationships for mutual smoothness. Levinas suggests the inconvenience — being pulled out of your own plans by someone else’s need — is where the ethical capacity gets exercised, not where it breaks down.
Notice when you’re replacing difficulty with comfort. The distinction between solitude and loneliness is useful here. Not all time alone is avoidance. But honest attention to whether you’re retreating to the mirror or sitting with something real is worth developing.
Take genuinely difficult conversations seriously. Hegel’s recognition theory requires confronting another as a full subject — and that requires letting their perspective land, including when it costs you something. That’s practice at the kind of encounter Levinas thinks ethics is made of.
None of these are dramatic. They’re ordinary situations where the face of the Other becomes present. The practice is not turning away.
What Levinas Doesn’t Solve
The framework is bracing in its diagnosis. It’s less useful as a consolation.
It doesn’t account for people for whom human relationships have been consistently unsafe — where the Other’s face has historically come with exploitation or harm. For them, careful cultivation of lower-stakes interaction as a step toward genuine relationship might be exactly right. Levinas’s ethics assumes a world where encountering the Other is possible. For some people, the therapeutic work is building toward that possibility first.
If genuine human connection feels dangerous rather than merely difficult, working with a therapist is more useful than working with Levinas. The philosophy names the goal. It doesn’t build the road to it.
The Stoic approach to loneliness adds something Levinas leaves out: the importance of developing the inner resources that allow you to show up to another’s demand from a place of stability, not depletion. Infinite responsibility is a real description of the ethical structure. It isn’t a sustainable operating mode without some interior ground to stand on.
Levinas was writing about ethics, not technology. But the thing he identified — the demand the Other’s face places on you, the pre-voluntary interruption of your self-sufficiency — maps with almost mechanical precision onto what comfort-optimized AI removes.
The loneliness epidemic isn’t running on a shortage of warmth. The warmth is available at scale, in apps, in a hundred calibrated forms. What’s short is the demand. The face. The thing that calls you out of yourself before you choose to go.
That’s what’s being removed, quietly and by design, in every interaction that’s been optimized to make sure you never feel uncomfortable.
This post draws on philosophy as a tool for reflection, not as a substitute for mental health care. If loneliness or difficulty with human connection is significantly affecting your daily life, please consider speaking with a therapist. Levinas clarifies what’s at stake. He doesn’t replace the work of actually building toward it.
