Stanford Center for Internet and Society

Marissa Mayer, Google’s Vice President of Search Product and User Experience, has said that “search is in its infancy.” If you want a preview of how Internet search might change over the next five to ten years, I encourage you to check out Weegy—“an online artificial being, powered by an advanced search engine and live experts.”

Weegy uses a combination of standard search techniques, low-level artificial intelligence, and crowd-sourcing to answer user questions in fields as diverse as “Parenting & Family” and “Electronics.” Weegy is (very) far from perfect, but does begin to leverage what I consider to be the search technologies of the future.

My predictions about next generation search—made on the basis of recent communications and design literature and comments by industry leaders such as Mayer—are as follows:

1. Next generation search interfaces will be social. People will search the Internet by “asking” an artificial agent a question, often out loud, and will receive answers in narrative form. Speech-to-text technology is already there, and search engines (such as Wolfram Alpha and Google Squared) begin to present “answers” instead of just links.

2. Next generation search will blur the distinction between machine and human labor. Machines are good at certain things; humans at others. Technologies like those underpinning CAPTCHA and Amazon Mechanical Turk will help search harness the power of people to help overcome nagging software limitations.

Weegy integrates both of these features... just not particularly well. When you first arrive at Weegy, you’re struck by a lovely representation of a female face accompanied by a simple and largely intuitive search function. The user goes through several steps to get an answer from Weegy. First, you're given the option to select a topic. Second, you enter a question into a standard box. Weegy then generates an answer which the virtual face reads out loud. The site invites you to rate the answer in order to make Weegy “smarter.” If the answer is “Good,” nothing happens. If it’s “Bad,” Weegy apologizes and finds a human expert to try again.

The problem is not in Weegy’s interface design, which is reasonably elegant, but in the website’s actual performance. The animation of Weegy’s face is good (in Internet Explorer, the only browser that supports the feature), but the voice with which the virtual face reads search results is awkward and outmoded. Weegy also draws answers verbatim from a set number of sources (e.g., Yahoo! Answers and Wikipedia), which often use the wrong tense or originate in an altogether different context.

Whether generated automatically or by the human “experts” hired to answer questions by the website, Weegy results vary tremendously in quality. I asked Weegy for the largest city in the U.S.; she nailed it. I asked her how many robots were in Japan. She read from part of an article about Japanese attitudes toward robots that did not answer the question. I marked her answer as “Bad” and got a live expert. The human’s answer was closer to the mark: “750,000 industrial robots in use around the world, about half are developed in earthquake-prone Japan.” Note that this answer, too, is somehow off—it uses stilted grammar, narrows the question arbitrarily, and includes extraneous information about Japan’s seismic activity.

Weegy is not unique in leveraging real people to answer questions. ChaCha performs this function faster and better, apparently vetting its participants with greater care. Nor is Weegy the only search engine ever to use anthropomorphic design (RIP, Ms. Dewey). But it is the first forum to combine these trends in any meaningful way. Weegy may, as the website speculates, “become the most respected authority in the world on all subjects.” Today, however, Weegy is just a herald of search wonders to come.