Hacker News Reader

Comment by aantix | Hacker News Reader

YouTube recommendations for kids shouldn't be driven by engagement (e.g. "oh, you watched 20 hours of MrBeast videos. You probably need another hour.")

If my kid watches an hour of unboxing toy videos, I shouldn't have to try and disable 3,000 channels of toy unboxings in an effort for that topic to never surface again.

The thumbs down button essentially has zero effect.

https://www.wired.com/story/youtube-dislike-button-mozilla-r...

Diversity of thought and topic is much more critical when they are kids.

View on HN

Todd Beaupré, the product lead for the YouTube homepage, says "We love exploration, have multiple teams dedicated to it."

But from my kids' viewing habits, I don't see it. The algorithm is seems super narrowly focused.

https://x.com/hitsman/status/1845168160691061121

Andrej Karpathy commented on the narrowness of the algorithm just the other day.

https://x.com/karpathy/status/1844449291282284925

I’d go further and say nobody should be subjected to this stuff. Engagement is a euphemism for addiction. Self improving addiction machines are a horrible idea. They’re actively bad for people and society for a laundry list of reasons.

Anyone who works on this stuff should be ashamed of yourselves. You’re modern day cigarette companies.

What do you feel is an appropriate and ethical way to surface content?

How would you go about writing the recommendation algorithm?

One thing I would like to see is an algorithm based on expressed rather than revealed preferences.

De-jargoned, that means clicking the like button means I want to see more things like that, and the dislike button means I don't want to see things like that. If I watched something all the way through, but clicked the dislike button, that means I don't want to see things that produced a similar response from people who tend to react the way I do.

This kind of algorithm is not going to increase watch time over the short term, and might not over the long term either, which makes it unlikely to be adopted by any for-profit service.

Remove the personalization aspect.

"The algorithm" was different in the early 2000's and all of our lives were richer for it. Before Facebook publicly embedded psychologists into the development teams and before the goal was more and more attention. The internet used to be great.

> How would you go about writing the recommendation algorithm?

I probably wouldn’t. Twitter, YouTube and Facebook were all invented before recommendation algorithms existed, and they worked great. Reddit still doesn’t use a recommendation algorithm.

Would they make less money without this stuff? Yeah, probably. But I think that’s a good deal for humanity.

I was really struck in The Social Dilemma how at the end they asked devs if they’d let their children use the platforms they’ve made. People interviewed overwhelmingly said no. If that’s the case, you know it’s bad. Just stop.

Reply on news.ycombinator.com