The Clockwork Critic

Disappointed by the movie or music recommendations you've been getting from human critics? Then maybe it's finally time to plug into the ever-improving (ostensibly) world of "recommender applications." Canadian writer Sarah Lazarovic explains that such "web-based applications seek to recommend music not through descriptive reviews, but through affinities calculated by a computer algorithm. Websites like Movielens.org and Filmaffinity.com endeavour to tell you what you will like by having you tell them what you already like."

There's ever more media out there to choose from, notes Lazarovic. "Defining by association," she writes, "involves abandoning micro-judgments (e.g. one critic's opinion, your brother's testimonial) for the world of global opinion -- computer-produced taste groups based on mutual affinities. We've long been in the habit of defining things in the context of other things -- so and so writes like so and so, someone sings like someone else. Sites like Musicplasma.com, Music-map.com [and the others] just take it to the next level, dispensing with the articles and adjectives."

Maybe you can never have too few adjectives, but I admit that I've gotten used to the articles.

Thanks to: ArtsJournal

Comments

[digamma]

The autorecommenders on Amazon and Netflix consistently recommend things to me which I dislike or in which I have no interest. YMMV.

[Twba]

I occasionally get an email from Amazon recommending something I was just thinking of buying. Because I've spent so many years living far from decent music retailers, I've bought enough crap from Amazon to make their recommendations program work fairly well.

[s.m. koppelman]

Must be a slow day at the CBC arts desk. Amazon and other major book and music retailers have been doing this for at least six? seven? years. The old Firefly, before Microsoft bought it to get the technology and shut down the site, did an especially good job once you rated a couple hundred artists and albums. That was about ten years ago.

Then as now, these systems only start doing a good job once you've indicated your ratings of at least a hundred or so items in a given medium, and in my experience, it doesn't cross media well. If you haven't rated a bunch of movies, all the data in the world on your music tastes isn't going to result in very good movie picks, at least at this point.

[ABC]

The problem with these applications is that they all implicitly assume that your preferences are objective, or at least consisten. Whether you like or hate a movie is highly subjective, and I don't mean just that one person might love a movie that you hate, I mean that the you that exists at this moment might have hated the very movie you loved last night.

Whether you like a movie or not is at least as dependent upon your mood going in as it is upon the intrinsic qualities of the film.

[Stevo Darkly]

I occasionally get an email from Amazon recommending something I was just thinking of buying.

In some ways, the Amazon recommendations are pretty spot on -- they often recommend things I already bought years ago from somewhere else.

There seems to be one bug in the way Amazon recommendations work, though. Many of the recommendations are based on stuff I bought someone else for a Christmas or birthday present, and have no interest in for myself.

[Yakov Smirnov]

Maybe you can never have too few adjectives, but I admit that I've gotten used to the articles.

Not me! I think articles are pain in ass! I never use articles when I am doing stand-up routine.

[Nathan]

The autorecommenders on Amazon and Netflix consistently recommend things to me which I dislike or in which I have no interest. YMMV.

At least your TiVo doesn't think you're gay.

[isildur]

ABC,

They generally don't function on the assumption that your tastes are objective, but rather on the assumption that your subjective tastes are similar to those of people who like the same things you do.

If all the data points are similarly subjective, you can usually build a pretty coherent recommendation list, assuming any given user supplies it with enough points of reference.

I get a lot of uninteresting stuff from Amazon's recommendations -- but they also recommended Richard K. Morgan to me. Best. Recommendation. Ever.

[SR]

"There seems to be one bug in the way Amazon recommendations work, though. Many of the recommendations are based on stuff I bought someone else for a Christmas or birthday present, and have no interest in for myself."

You can fix that by going into your Amazon.com recommendations page, clicking on the link that says "[X] Items You Own" (where [X] is a number) and then de-selecting items you've bought for someone else.

[Stevo Darkly]

Ah. Thanks, SR.

[Matt]

I've had a pretty consistently good experience with Amazon's reccomendations for books. Far less so than with reccomendations for music (which I don't buy much of) or movies (which I buy a lot of, but according to what most people would consider to be extraordinarily strange selection criteria). I freely concede that the failures of the system for me in music and movies are because of insufficient (in the former case) or misleading (in the latter case) data.

The difference between Amazon and a system like this is that Amazon's process is much more user-friendly. I buy things I want (which is what I go to the site for) and it automatically gathers data and attempts to perceive my taste...and if it guesses wrong, the worst case scenario is that I only get to do everything I actually intended to do at Amazon (that is, buy the books I went there to buy). I don't have to spend hours of my life specifically feeding it data about myself and hoping that the outcome of its analysis proves useful enough to justify that time.

Even if your TiVO thinks you're gay, it'll still record the shows you tell it to. What do these sites offer to users whose tastes they aren't good at interpreting?

[Josh Cohen]

There's an application we have at my office that works somewhat like this, but not for movies. (I'd say more, but we have a confidentiality agreement.) We never use it. It was programmed into our computer system and never completed.

People, on the whole, prefer humans to machines. That's why we all hold down (or repeatedly press) the zero button when we need to talk to a customer service representative. That's why we still have human beings reading news, weather, traffic, and sports instead of some digitized voice -- even though a digitized voice is possible, and even if the human is just a taped loop repeating the same stuff over and over. We'd still rather hear the human.

[HC]

It makes me curious whether any of you ever get annoyed by these online recommendation systems, telling the user what he is "expected" to like based on past behavior (strong intention to persuade/manipulate). What do you think?