I took a trip to gnooks.com today - a site where you can get book selections. I had never heard of it before today, and it’s probably not the best place to get an accurate recommendation, but it’s the one stumbleupon threw at me.

I’ve always been curious how sites like this will inform me. As someone who knows a fair deal about books, literature and common authors, I assume that I can predict some of the authors I’ll like. Short stories come into play here as well - there’s not as much of a time commitment in reading a new author in these cases and recommendations pile up quickly.

So how accurate will the site be? And how is the information created?

I tested it to see. I entered John Steinbeck, Jonathan Safran Foer and Michael Chabon as some of my favorite authors. The first response? Nicole Krauss. I recognize her from shopping at Target, but that’s it. “I don’t know it,” I say.

Next: John Kennedy Toole. Love him and his one, Pulitzer Prize winning book. I like it.

Next: Gore Vidal. No thoughts either way; “I don’t know it.”

Glen David Gold. Joseph Mitchell. “Don’t Know.” Zadie Smith. Roddie (sp) Doyle. “I like.”

Here’s the inherent problem with sites like this. These authors aren’t actually recommendations. They are authors with similar fan bases.

Naturally - a cold, unfeeling computer can’t make a recommendation. It only knows data - it’s unable to ask follow up questions, and it’s unable to understand why you picked one book over another. It’s based not on themes or writing styles or anything specific - it’s simply saying “well, some people liked this, and they also happened to like this, so there you go!”

Each visitor makes connections between authors, and these connections are saved in the database. If I, for instance, enter the authors mentioned above, the database will make a connection between these three. If the program suggests Dave Eggers and I say “Yeah, like him,” then there’s another connection. The four authors are connected.

When the next person comes in and types in Michael Chabon, there’s already three authors connected to him. More connections equals a higher collaboration with readers.

I understand the concept, and it’s not a bad idea - it’s the entire basis of Amazon’s recommendations. What Amazon has going for it is information from millions of transactions. But even then, it’s the same concept on a larger scale. Sure, I like John Steinbeck, and I happen to like Michael Chabon, but it’s not necessarily the case that everyone who likes John Steinbeck will like Michael Chabon.

It’s the same for sites like Pandora. Music with the same fan base is listed as a recommendation. But it’s not based on your personal tastes, as a recommendation should be, but on a group of

When it comes down to it, sites like this are fun and interesting. But that’s it. Sure, you might learn a little bit about how the general public feels about an author or musician. But you can never get a real recommendation unless you talk to someone. A person. Who understands the connections not on a data analysis level, but on an emotional level.

Computers might be better data analysts than humans. But they’re not necessarily smarter.