No, Google did Not Rig Indian Elections

"Did Google fix Lok Sabha elections?" asks the India Times. Computer Business Review appeared more certain with the "How Google search results are influencing elections" headline for its version of the story.

Even the Daily Mail joined in, asking "Could Google fix an election?"

Only one thing: Google is not "fixing" the Lok Sabha elections. The company is absolute in its denial. "Providing relevant answers has been the cornerstone of Google's approach to search from the very beginning," a spokesman told the Guardian. "Our results reflect what's on the web, and we rigorously protect the integrity of our algorithms. It would undermine people's trust in our results and company if we were to change course."

Indeed, Google has no intention of doing so; and its search results have not influenced the Indian elections beyond providing links to information that is on the web.

So why the furore? The stories all lead back to a press release put out on 13 May by the "American Institute for Behavioral Research and Technology", an independent – that is, unaffiliated with any major university – research organisation based in California. That release was headlined "Could Google have fixed the Lok Sabha elections? A landmark new study in India shows it's possible".

It's a pretty bold and arresting claim. Even a more detailed publication of the research had a similarly dramatic headline: "Democracy at risk: how voters in the 2014 elections in India were manipulated by biased search rankings." Again, this seemed to imply that this manipulation had actually occurred.

But it hadn't. On closer examination, both the media release and the longer publication explain that what the study actually looked at wasn't whether Google is actively conspiring to overthrow democracy, but what would happen if it did. And those are very different things.

"In the new study, participants were randomly assigned to groups in which search rankings favored either [Indian prime ministerial candidates] Mr Kejriwal, Mr Gandhi, or Mr Modi," the press release (PDF) explains. "Real search rankings and web pages were used, and people were asked to research all the candidates just as they would on Google. The only difference between the groups was the order in which the search results were displayed."

Even then, the nature of the research remains unclear. It's only once you read the actual paper that the actual basis for the claims falls into place. (The paper is not yet published in any academic journal, and is only online in draft form. Some would consider it rather premature to issue a press release.)

The original study involved creating fake search results for the Australian premiership, and artificially biasing them towards Julia Gillard or Tony Abbott. This was literally done by just making the first page very pro-Gillard for one set of viewers, and shoving Abbott to the back; and vice-versa for the other half.

Sure enough, if you massively weight the information people receive in favour of one candidate, they go on to like that candidate more.

Except, oddly, the original academic paper contains no mention of the Indian elections at all.

So how has India come into it? The authors appear to have repeated their experiment online – as India went to the polls – by recruiting people living in various Indian states through Amazon's Mechanical Turk (which finds volunteers who will carry out tasks for small payments). They performed the same test with 2,000 paid volunteers, and found much the same results: by taking undecided voters and presenting them with mocked-up search results pages, they showed that people were influenced by how the results seemed to portray the candidates – to the extent of shifting their allegiance towards or away from one or the other.

However, this is not the same at all as saying that Google has done any such thing, nor even that someone who tried to manipulate its search results could achieve the same. So no, Google did not, and isn't going to, fix or influence the Indian elections beyond linking to what's out there.

In a coda to the press release, Robert Epstein, one of the researchers (who has had his own, quite separate, run-in with Google in the past over a malware warning relating to his personal site), says: "Of particular concern is the fact that 99% of the people in our study seemed to be unaware that the search rankings they saw were biased. That means Google has the power to manipulate elections without anyone suspecting they’re doing so. To prevent undue influence, election-related search rankings need to be regulated and monitored, as well as subjected to equal-time rules.” It's a remarkable suggestion – but highly unlikely ever to be adopted in any country.