An experiment conducted by American Institute of Behavioral Research and Technology (AIBRT), a research institute in California, has revealed that biased search engine rankings could alter the voting preferences of undecided voters in recently concluded Lok Sabha elections in India. Using data collected in India between April 2nd and May 12th 2014, the researchers deliberately manipulated the voting preferences of more than 1,800 undecided voters (age 18 to 70) and were able to shift votes by an average of 12.5 percent to favored candidates, according to a report in the Washington Post.
In earlier online and laboratory experiments, the AIBRT researchers had found that biased search engine rankings could swing the voting preferences of undecided voters by 15%, when the participants have little or no awareness that they were being manipulated. The researchers call this phenomenon the ‘Search Engine Manipulation Effect’ (SEME).
In the experiment, the researchers recruited undecided voters who had not yet voted through print advertisements, online advertisements and online subject pools and encouraged them to sign on a web portal. Once they answered some general questions, the subjects were presented with Kadoodle, a fictitious search engine, and encouraged to search for information on the major candidates in the election – Rahul Gandhi, Narendra Modi and Arvind Kejriwal. To avoid bias, people’s preferences were pushed equally toward all of them.
But Kadoodle was manipulated. Each of the voters was randomly assigned to a group of equal size that supported one of the candidates. All the top 10 links that Kadoodle generated featured Web pages favoring that candidate; at the same time, the ranking of favorable links to the other two candidates fell. After viewing the search results for about 10 or 11 minutes, the participants were queried on their voting preferences. Eligible voters from 26 of India’s 28 states were participated in the study and the results were presented as follows:
- Social Impact Measures – The possible impact of election-related search rankings manipulations on a large scale was measured with three indicators:
- VMP (Vote Manipulation Power) – This refers to the percentage increase in potential votes for a candidate after people have been exposed to search rankings. The study showed there was a 12.2 increase in VMP. Even low percentage increase like 4.0 can have a significant impact on a close election.
- WMP (Win Margin Power) – WMP, the “threshold win margin below which control of an election might be guaranteed” was found to 2.9 percent. Below this win margin, manipulation of search-ranking might guarantee the outcome of a two-person race. It is estimated that more than 25% of national elections worldwide are won by margins below 3%.
- PMP (Psychological Manipulation Power) – PMP is the proportion of voters who were apparently unaware of the manipulation of search results. In this study, PMP was 99.3 percent, which means a very high proportion of participants could not detect the manipulations. So a percentage of 100 means that no one detected the manipulation.
- Candidate Data – These statistics show how voting preferences changed for each candidate after exposing to search rankings that favored that candidate. Participants presented with search rankings favoring Mr Gandhi, the number of people who decided to vote for him increased by 28.3%. For those presented with search rankings favoring Mr Kejriwal, the number of voters for him increased by 12.0%. For participants who were shown search rankings favoring Mr Modi, the number of people who decided to vote for him increased by 8.1%.
- Demographic Data – The study also presented data on selected demographic groups listed in order from the most vulnerable to the least vulnerable to manipulation, based on VMP scores for each group. It was seen that some groups are especially vulnerable. A shift was evident in the voting preferences of 19% of women over 35 were and those of 18% of voters who were unemployed.
The study suggests as more and more people get access to internet worldwide, the power of search engine rankings to determine election outcomes will grow, which poses a threat to the element of ‘free and fair election’ in a democracy.
Says Dr Robert Epstein, lead researcher in the study and Senior Research Psychologist at the American Institute for Behavioral Research and Technology in California, “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”.
The Washington Post report says that some outside experts agree that prominent search engine like Google has the ability to alter how people and events are viewed. But, skeptics of this study argue that voters typically have a range of information sources beyond the data obtained from search engines and that they are influenced by several other factors including party allegiances, ethnic and religious affiliations and potent issues. They also added that major search engines such as Microsoft (Bing) and Yahoo have incentives to prevent the even the appearance that they may be manipulating elections as they fear user backlash.
In their response to the AIBRT study, Google officials said that the cornerstone of search engine giant’s approach to search has always been to provide relevant answers. If Google were to change course, this may undermine people’s trust in their results.
What relevance does this have search engine optimization? Well, is quite obvious that Google does manipulate search results with Penguin and Panda. Google only indexes and shows pages with distinct information, valuable content and legitimate link-building practices. Over-optimization simply results in penalization instead of front-page coverage.