Great Recommendations Are Half Placebo
scoopinion.com
Great Recommendations Are Half Placebo
1–10 of 15 posts
Re: Great Recommendations Are Half Placebo
#2Interestingly, I've found an opposite effect -- people tend to oversell their favorite movies, restaurants, etc., so a person's strong recommendation will often lead to disappointment, not because the movie/food isn't good but because my expectations were set too high.
Re: Great Recommendations Are Half Placebo
#3The only thing that ultimately matters to me as far as recommendators (whether human or algorithmic) are concerned is whether their recommendations turn out to be valuable to me in some way.
That's how you build trust: by making good recommendations. Everything else is secondary.
Re: Great Recommendations Are Half Placebo
#4At the same time, it's hard for me to trust algorithmic recommendations. I don't trust them because I know they're from a computer, and I find it difficult to believe that a computer program can "know" who I am.
Determining something that's likely to suit your tastes based on your browsing/buying/watching/reading/etc. habits isn't "knowing" you and doesn't require such a thing. I don't know if this is some sort of accidental anthropomorphism or what, but I've seen this in more than one place and it just doesn't follow.
Re: Great Recommendations Are Half Placebo
#5At the same time, it's hard for me to trust algorithmic recommendations. I don't trust them because I know they're from a computer, and I find it difficult to believe that a computer program can "know" who I am. Determining something that's likely to suit your tastes based on your browsing/buying/watching/reading/etc. habits isn't "knowing" you and doesn't require such a thing. I don't know if this is some sort of ac…
The whole point was that there's an emotional component to how good recommendations seem.
Re: Great Recommendations Are Half Placebo
#6A great recommendation engine should incorporate explicit feedback from users.
For example, Netflix knows what I've watched and how I've rated things. But it doesn't let me tell it things like "I categorically dislike horror movies", so it sometimes recommends those.
It's trying to guess something that I'd rather just tell it.
Re: Great Recommendations Are Half Placebo
#7Interestingly, I've found an opposite effect -- people tend to oversell their favorite movies, restaurants, etc., so a person's strong recommendation will often lead to disappointment, not because the movie/food isn't good but because my expectations were set too high.
This is a good point, and happens to me sometimes too in real life.
I've also found that some people read way too much into how well an automatic recommendation system should work. There have been a couple of angry feedback messages due to a guess being off.
Re: Great Recommendations Are Half Placebo
#8A great recommendation engine should incorporate explicit feedback from users. For example, Netflix knows what I've watched and how I've rated things. But it doesn't let me tell it things like "I categorically dislike horror movies", so it sometimes recommends those. It's trying to guess something that I'd rather just tell it.
Agreed. I've been wondering how to do that so that it doesn't get tedious but stays effective. Blacklisting categories or sources works to an extent.
Re: Great Recommendations Are Half Placebo
#9> All additional information will make your prediction better
I don't think this is true... I find that there's a point where additional information becomes noise, and can make your prediction less useful.
Re: Great Recommendations Are Half Placebo
#10A great recommendation engine should incorporate explicit feedback from users. For example, Netflix knows what I've watched and how I've rated things. But it doesn't let me tell it things like "I categorically dislike horror movies", so it sometimes recommends those. It's trying to guess something that I'd rather just tell it.
Isn't that what this does? http://movies.netflix.com/TastePreferences