Assistant Professor, Information School, University of Washington.
intlib@ischool.washington.edu
Column for April 2004
I have seen the future of indexing.
Well, probably not, but that makes a catchy opening, don’t you think? I ran across an intriguing website the other day called the ESP Game. No Dionne Warwick or Miss Cleo, sadly, but fun nonetheless.
When the game begins, you are paired with another online player and shown a series of images, one at a time. You and your partner are asked to type in “labels” for each image you see. Sometimes, you’re also given a list of taboo words, which you can’t use.
The object of the game is to earn points by coming up with a label that matches one your partner also suggests. Games last for two-and-a-half minutes, with a maximum of 15 images. If you get stuck, you can suggest passing to the next image; if you both agree, you move on but come back to the passed-over images if time is left. Otherwise, you can’t communicate with your partner at all.
Here’s what I’ve learned so far from this game: First off, it’s pretty addictive. The two-and-a-half minutes flies really fast, and it can be great fun trying to figure out the person you’re playing with. Sometimes I got into a rhythm and made matches like crazy; other times, I wasn’t so lucky. Simple words are best: descriptions of what you see, including shapes, objects, words on the image (if they aren’t ones that are forbidden), or colors. Nouns are much more successful than verbs, and don’t try to get cute or metaphorical.
Abstract or difficult-to-identify images are a bear, since there’s really very little to grab onto. After a game is over, you can see what your partner had suggested, which can prompt substantial and often comic disagreement.
From the experience, I’ve learned that there a lot of stupid people out there. Or maybe I’m too smart for my own good. Or too much of a librarian. Whatever. I have serious reservations about several people I played with, given what they did—or didn’t—suggest.
Harvesting search terms
So this is cute, but what’s really going on here? The fact is, the site is a research project out of Carnegie Mellon University, trying to get people to index images on the Web (though they don’t use the word “indexing,” preferring “labeling”). All the images are randomly harvested, and the researchers are studying these labels we’re all typing in.
The very thin “About” page lists the researchers’ motivations: improving accuracy of image searching, improving accessibility of Web images, and helping people block “inappropriate [e.g., pornographic] images.”
Fine. So what’s the lesson for us from this demented hybrid of indexing and the $25,000 Pyramid? I can tell you that normal people are not thinking about labeling the way we do, at least not in the context of this game. I doubt there is much consideration of term-discrimination values or the principles of sound indexing going on. People often choose the obvious, the easy, and the concrete.
Let’s assume this game actually takes off and becomes hugely popular. The site says that if that happened, “all the images on the Web can be labeled in a matter of weeks!” Would “common sense” indexing then drive out professional-level efforts, as some think has already happened with search engines and ready reference? Some kind of Gresham’s Law for intellectual control?
Probably not any time soon—I never saw more than about 30 people logged on at any given time—but these researchers are going to find out many interesting things here, and the project is the kind of thing worth keeping our collective professional eyes on.
I found out about the ESP Game in a brief news article that also mentioned SpellWeb, which reports the number of times variant spellings of a word are found by Google in an attempt to discern the “correct" spelling. In the January 2003 American Libraries I wrote a piece on “authority by community" that discussed sites like the Internet Movie Database that rely on users for authoritativeness. These sites seem more like “reality by consensus” to me. (As Lily Tomlin observed, “After all, what is reality anyway? Nothin’ but a collective hunch.”) Is this an early example of a new, democratic, popular notion of information description, based on mass interest and appeal? Or maybe it’s part of a slide dumbing us down toward the lowest common denominator . . . but that’s another story.