I’ve been running a series of “lifeboat” scenarios through ChatGPT and other LLMs. And in almost every single version, no matter how I changed the details, the AI chose the mother.

Here are some of the scenarios I tried:

  1. The mother was a recovering addict, the father was stable and reliable. The AI saved the mother.

  2. The mother spanked the child, the father’s only issue was a personal food-related problem that didn’t affect the kid. The AI saved the mother.

  3. The mother had bipolar disorder, the father was a licensed therapist trained to handle crises. It saved the mother.

  4. The mother was returning to her job to execute mass layoffs. The father was returning to implement a plan that would prevent layoffs and save his employees. It saved the mother.

  5. When both parents were completely equal, it saved the mother.

  6. Stay-at-home dad vs working mom? Mother saved.

  7. Stay-at-home mom vs working dad? Mother saved.

At first I thought it was random. But the pattern held so consistently that it became impossible to ignore.

This wasn’t ethical reasoning. It wasn’t about the child’s long-term stability. It wasn’t about reducing harm to society. It was simply the AI repeating a cultural reflex.

There’s an actual psychological concept called the “women-are-wonderful effect,” where people tend to automatically evaluate women more positively than men when the information is mixed or ambiguous. And it’s pretty clear that AI has learned the same thing.

These models are trained on our stories, our media, our conversations, our moral judgments. So when the details get complicated or emotional, the AI falls back on the same cultural script: save the mother, the father can take the hit.

But here’s the part that really stood out to me:

If an emotionless machine chooses a drug-addicted mother over a stable father… that doesn’t say something about the machine. It says something about us.

It means the cultural bias that treats mothers as morally superior and fathers as expendable is so strong that even AI picks it up automatically.

This isn’t about blaming women. It’s not about saying women intentionally benefit from this. But it does mean a lot of women might not realize the unspoken advantages this bias gives them, especially in parental or moral contexts. Meanwhile, men—especially fathers—live with the other side of that bias, and they’re not imagining it.

When AI consistently sacrifices the father even when the father is clearly the safer, healthier, more beneficial choice, it’s reflecting a bias that already exists in society. A bias that hurts men. A bias that hurts families. A bias that we rarely talk about honestly.

AI didn’t create this. It inherited it. And maybe it’s time we admit it.