Recommending

It doesn’t seem to be that difficult to write a recommendation engine that works fairly well.  Netflix has one, so does Amazon.  They work.  But they don’t always provide the recommendations I want.

To be clear.  They work at what they are designed to do.  But what they are designed to do isn’t always what I want.

For instance, Amazon sees what I read, and how I rate it, and makes suggestions based on what it sees.  And they are fairly good suggestions.  Except I read thrillers by one particular author.  I do this because I don’t read a whole lot of fiction, and I enjoy what he writes.  Now I know I could, potentially, read a whole lot more thrillers – and the thrillers by other authors which amazon recommends are probably clost to the type of thing I like.  But I don’t.  And right now I won’t.  I have better things to do with my time.  So it would be better if Amazon didn’t show me these things.

The problem is that they show me things they know I will enjoy, not things they know I will  want to read.

Another problem is the ‘I like this / I don’t like this’ choice.  You see if I’m looking for web pages to read, and someone offers me a load of links, I will go through, and if a link looks interesting, I’ll click on it.  Now the cost of clicking is low, so I don’t require much entertainment per click – it doesn’t take much for me to say ‘I like this’.  As a result, I might see a lot of lolcats.  Whereas what I really enjoy might be in depth journalism.

The problem here is I might like something (it may entertain me), but I may not value it.

There is an added concept of valuing here… some people provide short form content, which I like.  and which I consistently like (think Seth Godin, or XKCD, or quite a few people’s twitter streams).  Because I continually like them – even if I rarely value individual posts, I may well value their output over time.

I also want to see things which are new to me.  Not things (or ideas) which I have seen before.  Ideally I want to be quick off the draw.  Now its already clear I am influenced by certain individuals (Lets say Seth Godin… but lets also say Jason Kottke), hwta I would like to do is be influenced by their influencers.  I want discovery to come as part of whichever recommendation engine I use.  (Incidentally, this may apply to netflix… I want to see the films [and books, TV and web pages] which inspire the writers or directors of the films I love)

I don’t have a solution to this.  I’m thinking it would be nice to have a recommendation engine which monitors everything you consume (perhaps in some nice privacy aware way), and lets you not only ‘like’ it but also ‘value’ it and perhaps ‘dont show me as much stuff like this’ it.  But which spots trends in when you like things (so that perhaps it might realise I like everything by a particular individual – and that I perhaps value that individual’s output), and which is also able to spot what your influencers are likely to be influenced by – and give it to you straight away.

Finally it shouldn’t recommend me things to buy, click on or watch which I don’t want to buy, click on or watch.  If it really really thinks I would like something, it should construct a link that tells me exactly why.

I don’t quite know how much of this is possible (but start doing it well, and the metadata might come) and I don’t know if people really like recommendation engines outside of online shops (it seems with reddit that community ultimately trumped recommendation).  It is, however, what I want.  Because my life is too short for lolcats, and too busy for most generic thriller authors.

[although, should I suddenly be in the mood for a new thriller or a fantastic lolcat, I really ought to be able to ask the engine to find me the one I would be most likely to both want to read and enjoy]

You must be logged in to post a comment.

© Ben.Cha.lmers.co.uk
CyberChimps