Many of you wanted to see even more personalized recommendations ... [Now], we've started inserting items selected just for you inside the Recommended items section. This is great if you've got interests that are less mainstream. If you love Lego robots, for example, then you should start to notice more of them in your Recommended items.Sadly, no additional details appear to be available. In my usage, there were rare gems in the recommendations, but a lot of randomness, and a strong bias toward very popular items. The lack of explanation -- why was this item recommended? -- and lack of a way to correct the recommendations likely will make people less forgiving of these problems. I also saw recommendations for items I had already read; items you have already seen always should be filtered from recommendations.
For more on that, you might enjoy some of my previous posts on this topic, such as the Mar 2009 "What is a good recommendation algorithm?" and the much older Dec 2006 "The RSS beast".
Update: A couple weeks later, Google launches Google Reader Play, a StumbleUpon knock-off that recommends web content based on what you say you like. Googler Garrett Wu writes that "it uses the same technology as the Recommended Items feed in Reader." In my usage, it had the same problems too.
4 comments:
I wonder how they account for the fact that most people mark all as read otherwise the feed becomes unmanageable.
I only really read about 25% of the items that come in but I mark them all as read when I am done so I know what is really new next time.
I just tried it this week-end and I found it very poor like you said with an high bias for most popular feeds. Sadly, I'm not looking for the next slashdot feed but more for the tiny little blog with few valuable posts highly related with my area of interests.
The lack of explanation -- why was this item recommended? -- and lack of a way to correct the recommendations likely will make people less forgiving of these problems.
I've been saying the same thing for years about regular web search results. Going abstract for a moment, getting a search result and getting a recommendation is not all that different. You run an Amazon query, and get recommendations based on the combination of that query plus your personal history. You run an web search and get results based on a combination of that query plus your personal history.
So shouldn't web results offer the same explanation? Why was a web result retrieved? And why is there a lack of ways to correct the results that I am seeing, when my web search results are bad?
Give people a way to do that, and you'll have given them the tools they need to make their search experience better. Why are we not given those tools, when we want them?
Well, also thinking back to your article about user noise in their own ratings - plus the fact that it's known in the search field that most users are awful at search terms...
I really hope they are taking into account some sort of best practices in terms of personalizatoin for the user.
Otherwise, you're going to end up with the same problem that results in current "top" lists or "most" categories (viewed, rated, reviewed etc). Common users don't know what to look for or where to go, and they instead follow the line of ants. This artificially floats content to higher views or to the top lists.
A lot of the problem with improving search and recommendations is the "gaming the system" you reference. Instead of focusing on "fooling" users into clicks or views, or simply using top line "rating" type data - it's important for search to not only become more real time, but also more deep in the data it considers for recommendations (actual clicks, time spent viewing, saved/fw'd links, click paths).
That means replicating the actual patterns of SUCCESSFUL experiences per user, not in general across the board. That's a direction most back end developers aren't thinking.... but users are. The problem is filtering out the user noise of what they say they want and offering backend categories that instead fool them into getting things they definitely want and probably will want in addition. But, this really needs some active "pushing" (i.e. a human editor) to adjust for the human users and their bad search habits on the other end.
Post a Comment