Top Google Engineer: Google Instant Has No Brand Bias

Amit Singhal Google

"What we do at Google and what we've done for years is to not inject any subjectivity into these algorithms," says Amit Singhal, Google Fellow and head of the company's search quality, ranking, and algorithm team. "We didn't want to introduce any bias into the mathematical modeling--our modeling is predicting, given a letter, what's the probability of completion."

Singhal is referring to Google Instant, the results-as-you-type feature the company introduced months ago that has been criticized for having a brand bias. When the system was first overhauled, the blogoshere quickly noticed that many Instant queries were dominated by major brands: type E, and the first suggestion is eBay, J gets you JetBlue, V for Verizon, and so forth. (For a rhyming-couplet run-down of our ABCs, head here.) Had these results been bought? Could they be? Singhal is unequivocal: No.

"We stayed true to our mathematical principles," he said in a recent interview with Fast Company. "Most people typing A are seeing Amazon, but that probability is predicting that most people typing A are going to complete to Amazon. If you type T, most people typing T will go to Target. That's the probability model. If you add R to it ("Tr"), most people are looking for a translation system. It's actually just pure mathematical modeling."

The only exception, Singhal says, is objectionable results. Words like "pornography" are excluded from Google Instant, as many noted at its launch.

ComScore search evangelist Eli Goodman backs Google's policy. Unveiling research at a recent search summit, Goodman revealed that searches in Google Instant are 15% more likely to have ads because more of the results are brand terms. However, he specified that Google is not "being gamed in any capacity," and that nothing in comScore's research shows the engine is pushing certain topics for financial gain.

"I've said this many times: My subjective opinion, though always true, is just my subjective opinion," Singhal says. "We try not to insert it into any of our search processes--we just stick with our mathematical models."

Follow Austin Carr on Twitter.

Add New Comment

1 Comments