Hacker News Reader

Comment by wvh | Hacker News Reader

Is this just part of the trend to dumb everything down to a polarising dichotomy? Because the trend to stop people from having a more complex and rational opinion than just picking sides doesn't seem to exactly improve discourse or debate, or inform and educate...

Wouldn't it be more beneficial to have fewer people give a more detailed review to describe what they like and dislike about something? What use is "good" or "meh" compared to for example "I rate this product low because the shoes are very narrow for my normal feet" so you can actual relate it to what you actually are looking for? One person's "meh" could be another's "good"...

From the article:

"This makes sense – giving a five-star rating takes some thought, especially for something like a movie or TV show.

A binary “yes or no” option is much easier for viewers to commit to, [...]

View on HN

I think one problem is that people have vastly different opinions on what 3 stars mean. For some it's quite bad, for some it's in the middle.

A binary option actually has three options. Did it prompt me to rate it as bad, or did it prompt me to rate it as good, or was it bland enough where I don't want to sway the rating of it at all?

If I rate a movie on IMDB as 5-6, and maybe even 7, the possibility of me recommending it to someone "depends". In that case I might as well not rate it at all.

Ratings are inherently tricky, because we all like different things. I don't think you lose anything by going to a like/dislike system. Recommendations should be based on aggregate maps and matching you with similar viewers that have a certain overlap in ratings. That's made easier with binary choices.

IMDB ratings are only partially informative to me anyways.

From my perspective, that shouldn't be a problem. Ultimately (as a user) the point of the rating is for me to get stuff that I like -- so it's all relative to me, regardless of individual rating differences.

Now, I realize that's almost certainly not the level of individualization offered by their ranking algorithm (I'm convinced there's way more of a "groupthink" dynamic), but that's partly my point: if your ratings and recommendation system isn't capable of supporting individual differences, then your algorithm is broken.

If you have a complex problem (personalized recommendations absolutely are complicated) then you shouldn't fix a broken algorithm (Netflix recommendations are, at least for me, a bit meh) by simplifying the inputs (switching from stars to up/down). You should fix the broken algorithm, by creating a better algorithm.

I realize it isn't that simple, particularly for a publicly-traded company whose stock price would absolutely be influenced by a decision to scrap part of their core technology, but from my perspective as a user, this is going to hurt my experience with the product.

I think it's more a reaction to a usage pattern of their star system. YouTube abandoned their five star rating system citing that most people voted only 1 or 5 stars with almost nothing in between. [1] I'd guess a vast amount of Netflix users do the same.

[1] https://youtube.googleblog.com/2009/09/five-stars-dominate-r...

YouTube is partly to blame for this. Since YouTube reported the average stars, if you saw a 4-star video and thought that it should be 3-star, then the optimal strategy was to vote 1-star to pull it down as fast as possible. If instead YouTube had reported the majority option (and publicized that fact) then it would have incentivized people to vote what they truly think.

Your suggestion has its own problems. Imagine two videos with a majority rating of three stars. Say 60% of raters have rated it that. But one video has the remaining 40% one-star, and the other has the remaining 40% five-star. Your algorithm will ignore this data, showing a completely wrong result that both videos are equally good.

Reply on news.ycombinator.com