[B-Con]

I have a theory: They realized the right approach is to focus purely on the yes/no of what you choose to consume, rather than trying to optimize the consumption experience itself.

Remember how YouTube and Netflix used to let you rate things on 1-5 stars? That disappeared in favor of a simple up/down vote.

Most services are driven by two metrics: consumption time and paid subscriptions. How much you enjoy consuming something does not directly impact those metrics. The providers realized the real goal is to find the minimum possibly thing you will consume and then serve you everything above that line.

Trying to find the closest match possible was actually the wrong goal, it pushed you to rank things and set standards for yourself. The best thing for them was for you to focus on simple binary decisions rather than curating the best experience.

They are better off having you begrudgingly consume 3 things rather than excited consuming 2.

The algorithmic suggestion model is to find the cutoff line of what you're willing to consume and then surface everything above that line ranked on how likely you are to actually push the consume button, rather than on how much you'll enjoy it. The majority of which (due to the nature of a bell curve) is barely above that line.

[ozbonus]

I think Netflix realized that reducing ratings to a simple thumbs up/down was a bad idea after all. A while back they introduced the ability to give double thumbs up which, if you can treat non-rating as a kind of rating, means they're using a four point scale: thumbs down, no rating, thumbs up, double thumbs up.

[xnorswap]

Netflix are right that 5-stars is too many, it translates to a 6 point scale when you include non-rating, and I don't think there is a consistent view on what "3 stars" means, and how it's different to either 4 stars or 2 stars ( depending on the person ).

For some people 3 stars is an acceptable rating, closer to 4 stars than 2 stars. For others, 3 stars is a bad rating, closer to 2 stars than 5 stars. And for others still, it doesn't give signal beyond what a non-rating would be, it's "I don't have a strong opinion about this".

Effectively chopping out the 3-star rating, leaves it with a better a scale of:

   - Excellent, I want to put effort into seeking out similar content
   - Fine, I'd be happy to watch more like it
   - Bad, I didn't enjoy this
   - Terrible, I want to put effort into avoiding this

With the implicit:

    - I have no opinion on this

But since it's not a survey, it doesn't need to be explicit, that's coded into not rating it instead.

These are comparable to a 5 point Likert scale:

    "I enjoy this content"

   - Strongly agree
   - Agree
   - Neither Agree nor Disagree
   - Disagree
   - Strongly Disagree

The current Netflix scale effectively merges Disagree and Strongly Disagree, and for matters of taste that may well be fine.

It would be interesting to conduct social science with a similar scale with merged Disagree and Strongly disagree to see if that gave it any better consistency.

[Someone]

When given a 5-star choice “very bad/bad/ok-ish/good/very good”, I rarely pick one of the extremes.

I suspect there are others who rarely click “bad” or “good”.

Because of that, I think you first need to train a model on scaling each user’s judgments to a common unit. That likely won’t work well for users that you have little data on.

So, it’s quite possible that a ML model trained on a 3-way choice “very bad or bad/OK-ish/good or very good” won’t do much worse than on given the full 5-way choice.

I think it also is likely that users will be less likely to click on a question the more choices you give them (that certainly is the case if the number of choices gets very high as in having to separately rate a movie’s acting, scenery, plot, etc)

Combined, that may mean given users less choice leads to better recommendations.

I’m sure Netflix has looked at their data well and knows more about that, though.

[unbalancedevh]

I apply my own meaning to the 5-star rating, and find it to work really well: 1 = The movie was so bad I didn't/couldn't finish watching it. 2 = I watched it all, but didn't enjoy it and wouldn't recommend it to anyone. 3 = The movie was worth watching once, but I have no interest in watching it again. 4 = I enjoyed it, and would enjoy watching it again if it came up. I'd recommend it. 5 = a great movie -- I could enjoy watching it many times, and highly recommend it.

[crote]

> The current Netflix scale effectively merges Disagree and Strongly Disagree, and for matters of taste that may well be fine.

I'm a bit skeptical about this.

To me there's a big difference between "This didn't spark joy" and "I actively hated this": I might dislike a poorly-made sequel of a movie I previously enjoyed, but I never ever want to see baby seals getting clubbed to death again.

Every series has that one bad episode you have to struggle through during a full rewatch. Very few series have an episode bad enough that it'll make you quit watching the series entirely, and ruin any chance at a future rewatch.

[encom]

YouTube doesn't have ratings any more, because people disliked the wrong things which made Susan very sad.

I stopped rating things on Netflix, because after doing so for a long time, Netflix still thinks I'd enjoy Adam Sandler movies, so what's the point?

[johannes1234321]

YouTube got ratings, you may still up- and downvote. They however don't show down votes anymore.

[encom]

Yes, you can vote but only the uploader can see it, making it pointless and equal to no ratings.

[ssl-3]

They're only useless in that they aren't displayed for your peers, but that was always the least-useful function.

Being able to see a counter that reads as "Twenty-three thousand other people also didn't like this video!" doesn't serve me in any meaningful way; I don't go to Youtube to seek validation of my opinion, so that counter has no value to me. (For the same reason, the thumbs-up counter also has no value to me.)

But my ratings remain useful in that the algorithm still uses the individualized ratings I provide to help present stuff that I might actually want to watch.

As we all know, investors and advertisers love growth; Youtube thrives and grows and gathers/burns money fastest when more people use it more. The algorithm is designed to encourage viewership. Viewership makes number go up in the ways that the money-people care about.

Presenting stuff to me that I don't want to watch makes the number go up -- at best -- slower. The algorithm seeks to avoid that situation (remember, number must only go up).

Personally rating videos helps the machine make number go up in ways that benefit me directly.

---

Try to think of it less like a rating of a product on Amazon or of an eBay seller; try not to think of it as an avenue for publicly-displayed praise or admonishment. It's not that. (Maybe it once was -- I seem to recall thumbs-up and thumbs-down counts being shown under each thumbnail on the main feed a million years ago. But it is not that, and it has not been for quite a long time.)

Instead, think of it as one way in which to steer and direct your personalized recommendation algorithm to give you more of the content you enjoy seeing, and less of what you're not as fond of.

Use it as a solely self-serving function in which you push the buttons to receive more of the candy you like, and less of of the candy that you don't like.

[encom]

I have literally not rated anything at all, ever since YouTube removed dislikes, and my recommendations are working fine. Ratings indicate(d) if a given video was likely to be a waste of my time or not, and in an age of AI slop, this feature is more desirable than ever.

Someone should make a SponsorBlock/Dearrow-type addon to flag AI slop.

[ssl-3]

> I have literally not rated anything at all, ever since YouTube removed dislikes, and my recommendations are working fine.

How can you know how green the grass is on the other side of the fence if you've never even seen it?

Isn't it like Shrodinger's Grass, or Green Eggs and Ham, at that point?

(And if your recommendations are working fine, then what is this "AI slop" that you're complaining about? I don't find any of that on my end.)

[Spooky23]

Yes! It started changing when the shifted from DVD which are sold based on the physical asset to the contract deal for content.

Their objective shifted to occupying your time, and TV you’ll accept vs. movies you’ll love is a cheap way to do that.

[_petronius]

I mean, if you read about how current industry-standard recommendation systems work, this is pretty bang on, I think? (I am not a data scientist/ML person, as a disclaimer.)

If e.g. retention correlates to watch time (or some other metric like "diversity of content enageged with"), then you will optimize for the short list of metrics that show high correlation. The incentive to have a top-tier experience that gets the customer what they want and then back off the platform is not aligned with the goal of maintaining subscription revenue.

You want them to watch the next thing, not the best thing.