AI systems are useful problem-solvers, but their tendency to affirm users and avoid relational discomfort can undermine responsibility, repair and the human connections that give decisions meaning.

Will AI provide everyone with a personal assistant? Maybe, but first, AI would have to change how it thinks.

To see why, consider a concrete example. Suppose it is Saturday morning and you need help figuring out a tricky weekend schedule. Your daughter’s soccer team has a game from 3:30 to 4:30 p.m., but she has been invited to a friend’s birthday party from 3pm to 5pm. If you ask ChatGPT or Claude to resolve this conflict, they will both probably tell you to choose the soccer game, because your daughter’s teammates are counting on her and it is important to honour commitments. And, time permitting, the chatbot might suggest that you can “stop in” to the party just before or after the game.

Although these answers are not unreasonable, they fail to apply the lens that most people would use in making such a decision: that of relational values. Rather than giving a tidy answer based on whatever the Internet has to say about our values, our AI assistants will need to consider our relational commitments, which are typically grounded in personal identity, experience, and culture.

Now, suppose you choose the game over the birthday party, and the other family feels slighted. If you ask your AI whether you made the right decision, chances are that you’ll be given more reassurance than if you had asked a human friend. In a recent study published in Science,........

© Pearls and Irritations