Do I want an app that tells me what I like?

This week’s CBC tech column is all about the double-edged sword of online personalization. There’s a copy up at cbc.ca, and one below for posterity.

===

Last week, a Vancouver-based app-maker called Zite launched a new iPad application of the same name that it bills as “a personalized iPad magazine that gets smarter as you use it.”

As a member of the so-called Generation Y, I am, of course, a narcissistic egomaniac with an affinity for anything that promises to shape itself in my image. So, of course, I downloaded it.

Here’s the idea: Zite brings together stories from across the web — blog posts, magazine articles, stories from news websites — and filters them by your particular interests, creating an up-to-the minute personalized reading list just for you.

The underlying technology was developed at the University of British Columbia’s Laboratory for Computational Intelligence.

Zite’s fundamental innovation is that it tracks how it’s being read.

After narrowing down the subject areas, it shows you a number of stories it thinks you might like, then it tracks how you interact with them.

“We have an underlying philosophy that ‘you are what you read,’ explains CEO Ali Davar.

“When you’re on Zite, and we see you bypassing articles that we’ve recommended to you, that tells us as much as when you select an article. So, when you do that continuously, through time, we learn something about you.”

Staying on track

Zite also pays attention to the form and content of what you’re reading. Is it a long article? A short article? Who wrote it? Does it come from a particular political viewpoint?

By tracking your reading habits, Zite tries to give you more of what you’re interested in, and less of what you’re not.

This tactic will sound familiar if you’ve ever bought something from Amazon, used a TiVo, or watched Netflix.

All of these services track people’s behaviour, then use that information to give them more of what they like.

When I tried the Zite app, I really did have the sense that it was learning about me.

And while I recognize that these recommendation services can be useful, part of me can’t help but worry. Specifically, I’m concerned that online personalization will perpetuate my bad or lazy habits.

For instance, I spend a lot of time reading gadget blogs. Arguably, too much time. Over the past two weeks, there’s no question that I’ve read more reviews of the iPad 2 than necessary.

I know others with similar vices: Hollywood gossip blogs or obsessive sports coverage. My question is: Do I really need a tool that will help me find yet another iPad 2 review? Or would I be better off reading something new and unfamiliar?

Comfort versus challenge

For another perspective on this, I called Ethan Zuckerman, a researcher at the Berkman Centre for Internet and Society at Harvard University.

As he put it, “personalization is absolutely a double-edged sword. You can imagine it being a force to challenge you, and push you towards things you might not otherwise have read.

“You can also imagine personalization cocooning you in a world of familiar, unthreatening, unchallenging, but copacetic news.”

Zuckerman frames this tension as “comfort versus challenge.”

And it is the first part that worries me. That by giving me more of what they think I want, these personalization tools might actually narrow my worldview. They might cocoon me in the comfortable to keep me coming back.

When I asked Zite’s Davar about this, he told me his company’s technology is focused on what he calls discovery.

“It’s not simply about getting more of the same. It’s actually quite the opposite. The challenge is to give you the things that you wouldn’t typically find if you went out and looked for yourself.”

From a technical point of view, this is apparently a difficult feat. There are two very different answers to the question why didn’t you read that.

One is, I didn’t read it because I know I won’t like it. Another is, I didn’t read it because I didn’t know it existed.

Training computers to tell the difference is hard, but Zite believes they’ve found a way, using a secret algorithm sauce.

Zuckerman, on the other hand, told me that “anytime someone is providing algorithmically-organized information, there are some politics behind it. And you really owe it to yourself to think about what those politics are.”

The personalized recommendations that come from Netflix and Amazon are generated by proprietary algorithms. We don’t know exactly how they work, but we do know their objectives: to sell more stuff, and keep subscribers watching.

Zite’s algorithm is also proprietary and Davar says it took “a lot of money and a lot of time to develop.”

Zite’s service is currently ad-free, but Davar told me the company plans to add advertising and a subscription model to generate revenue.

It seems we’re headed towards a world of increasing personalization. As such, it’s important to think critically about personalization technologies.

When faced with an algorithmically-generated recommendation, we need to ask questions like why is this being recommended to me. Is it making me comfortable, or is it challenging me? And who’s getting paid?

For me, it’s about recognizing your own personal blind spots, and not necessarily trusting a computer algorithm to help fill them.

Now, if you’ll excuse me, I have to go read another iPad 2 review.