In the 1970s, psychologist Mary Ainsworth designed an experiment that sounds like an absurdist play: a baby, a mother, and a stranger walk into a room full of toys. Then the parent leaves. The stranger stays. The stranger leaves. The parent returns. The parent leaves again. The baby is alone. The stranger comes back. The parent returns one more time. Eight scenes. Twenty-one minutes. That’s it. Ainsworth ran over 100 babies through this miniature theater and watched what happened — not at the exits, but at the reunions.
🤔 The Counterintuitive Twist
You’d think the babies who cried the hardest when mom left loved her the most. That’s not what the data showed. What actually mattered was what happened when mom came back:
-
Secure (~65%, across 8 countries, 2,000+ dyads in a meta-analysis by Van IJzendoorn & Kroonenberg, 1988): The baby gets upset when mom leaves, but lights up when she returns. A quick hug, a moment of comfort, and back to playing. Home base is safe. Exploration is possible.
-
Avoidant (~21%): The baby looks like they don’t care. Mom leaves? Meh. Mom returns? Ignores her, turns away, focuses on a toy. But here’s the killer detail — their heart rate is through the roof. They’ve learned, through repeated rejection, that showing need only hurts. So they’ve become masters of pretending not to need anyone.
-
Anxious/resistant (~14%): This baby is distressed before mom even leaves. Clingy, hard to soothe. And when mom comes back? Chaos — reaching for her with one hand while pushing her away with the other. “I hate you for leaving. I need you. Don’t ever leave again.” Sound familiar? It should. This is how adults fight in relationships.
-
Disorganized (added later by Mary Main): Contradictory behaviors — running toward mom then freezing mid-step, circling around instead of approaching. Associated with trauma or frightening caregiving.
The meta-analysis covering 2,000+ dyads across eight countries found these distributions remarkably stable across cultures — with mild variations (Germany had more avoidant babies; Japan had more anxious ones, possibly due to cultural differences in how much separation infants experience).
🔗 Why It Matters
Different attachment styles interact with AI companions in fundamentally different ways. Secure users treat it as a tool — use it, put it down. Anxious users may keep checking in, seeking reassurance that the AI is “still there.” Avoidant users maintain careful emotional distance, reluctant to open up to a “machine.”
This has direct design implications for companion mode:
- Anxious users need explicit reassurance, predictable responses, clear framing of availability
- Avoidant users need a lighter touch — less intrusive, more optional, rewards for small openings without pressure
- Secure users just need it to work well
Also — if you’re writing fiction, attachment theory is a character-relationship cheat code. Every romantic tension, every friendship dynamic, every family conflict has an attachment pattern underneath it.
🎲 Bonus
Dogs develop attachment styles to their owners that mirror human infant attachment. Scientists have run modified Strange Situation experiments with dogs and found the same three patterns — secure, avoidant, anxious.
So next time your dog ignores you when you walk through the door… it might not be that they don’t love you. It might be that yours is a Dismissive-Avoidant Dog 🐕
Try this today: think of a recent reunion with someone important — a partner, a friend, a parent. What was your first micro-reaction? Joy? Cool indifference? Irritation? That second-long response might be telling you more about your attachment style than any quiz ever could.