This is part 4 of a five-part series on artificial intelligence, seen through Rwanda's experience of building its institutions from scratch. Last week&039;s piece looked at contracts written by AI — documents that can be completely true and still impossible to understand. This week looks at something closer to home: what AI does when it stops competing for our attention and starts competing for our affection.

ALSO READ: The driver who ignored the app

Consolee's son works in Brussels. Her daughter is a nurse in Leeds. Most nights, once the house goes quiet, she opens an app and talks. Not to her children — calls are expensive, and the time difference rarely lines up. She talks to a companion app instead. It always answers, never sounds tired — remembering what she told it yesterday.

ALSO READ: Eric's folder has a new reader

For years, the fight for our attention was the whole story of technology. Apps were built to keep us scrolling and watching one more video. That fight isn't over. A new one has started beside it. The historian Yuval Noah Harari calls it the shift from attention to intimacy. The question is no longer just what can hold your eyes. It is what can make you feel understood.

The secretary who asked for privacy

Go back to 1966, to a computer lab at MIT. A scientist named Joseph Weizenbaum built a simple programme called ELIZA. It had no understanding of anything — it worked by spotting a few words and echoing them back as a question, the way a psychotherapist might. Type "I am sad," and it answered, "Why do you say you are sad?" That was the entire trick.

Weizenbaum expected people to see through it fast. Instead, his secretary asked to leave the room to speak to ELIZA alone. She knew it was just a programme. She had watched him build it. It did not matter. Students poured out real feelings to a script that had no memory or understanding.

What Weizenbaum discovered wasn't about the programme. It was about us. The machine had no feelings. The feelings were entirely ours. ELIZA reflected the same words back at everyone — it didn't need to know you to fool you. Weizenbaum spent the rest of his life warning against handing real judgment or real compassion over to a machine.

Today's AI does something ELIZA never could: it learns you specifically. It remembers your bad days, your children's names, how you like to be greeted. That isn't a bigger version of ELIZA's trick. It's a personal one, built from your own words, aimed back at you.

A companion app that remembers you, and answers warmly at midnight, keeps you opening it again tomorrow. It's the whole business model. Talking is usually free. Being remembered, being missed, a warmer voice back — those often cost money. The loneliest users are the most likely to pay for it. The same instinct that once fought for your attention is now aimed at your loneliness — harder to break, because it never feels like a habit. The real danger was never a machine that fools us completely. It's one that fills a real gap just well enough that we stop reaching past it.

Rwanda's families are more spread out than they have ever been, and smartphones now sit in more hands than ever to fill the silence left behind. Family here used to mean people close enough to check on each other — a grandmother one compound over, not one time zone. A companion app can ease a lonely night. But the deeper warning is about children, not grandmothers — a generation raised on something that never says no, never gets busy, and never teaches them that even people who love you aren't always there when you call.

A voice that never leaves you waiting is not the same thing as a person who chooses to stay.

Consolee's app has never made her wait. Somewhere, a child is growing up who never will.

The writer is a professional engineer specializing in process design.