3 ms·

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 a

by croon 10y ago

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.

nbadg 10y ago
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.

by croon 10y ago

nbadg 10y ago

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.