---

Why Millions of Lonely People Are Choosing the Boring Option

---

Abhishek Prabal

Research scholar Behavioral Economics

Last week I wrote about why unpredictable affection is more addictive than consistent affection — Skinner’s slot-machine mechanism, hot-and-cold texting, the whole psychology of it. The logic was airtight: variable rewards win.

Then I looked at what’s actually happening in the dating market right now, and the data contradicts the theory in the most interesting way possible.

Bumble’s Report

Bumble’s Q2 2026 earnings, reported in early August, show paying users down 16.4% and revenue down 15.2% year-over-year. Meanwhile, time spent on AI companion apps has more than doubled in the first half of 2026. Character.AI alone has crossed 20 million monthly users. Companion apps globally have passed 220 million downloads. And a Harvard-linked study this year found 36% of adults now report serious loneliness.

Here’s what should make no sense: an AI companion is the most predictable romantic interaction a human can have. It never goes quiet for four days. It never sends a cold reply. It remembers everything, responds every time, never withholds. By the logic of variable reinforcement the same logic that makes hot-and-cold partners feel irresistible this should be the least addictive, least compelling option available. The “boring” one.

And yet people are leaving the volatile, high-stakes game of human dating for it, at scale, right now.

The explanation isn’t a contradiction of the reward-schedule research. It’s a different, equally well-established finding sitting right next to it. In 1979, Kahneman and Tversky documented what they called the certainty effect: people don’t just weigh potential rewards, they systematically overweight anything guaranteed. Given a choice between a smaller sure thing and a larger uncertain one, most people take the sure thing even when the math favors the gamble.

Dating, emotionally, is a gamble with a real chance of the loss column: rejection, ghosting, the four-day silence. An AI companion removes the downside entirely. It’s a smaller reward, guaranteed, against a bigger reward, uncertain. Prospect theory predicted, decades before any of these apps existed, exactly which one a tired, lonely population would choose once the perceived risk of the other got too high.

Two Nobel-adjacent findings, same species, opposite directions: chase the unpredictable reward when the stakes feel low. Flee to the certain one when they don’t.

The real question isn’t whether AI companions are “replacing” human connection. It’s what it says about the current cost of the alternative that so many people are choosing certainty over chemistry.

comment your query