Late in 2012, I began to wonder whether highly ranked search results could be impacting more than consumer choices. Perhaps, I speculated, a top search result could have a small impact on people’s opinions about things. Early in 2013, with my associate Ronald E Robertson of the American Institute for Behavioral Research and Technology in Vista, California, I put this idea to a test by conducting an experiment in which 102 people from the San Diego area were randomly assigned to one of three groups. In one group, people saw search results that favoured one political candidate – that is, results that linked to web pages that made this candidate look better than his or her opponent. In a second group, people saw search rankings that favoured the opposing candidate, and in the third group – the control group – people saw a mix of rankings that favoured neither candidate. The same search results and web pages were used in each group; the only thing that differed for the three groups was the ordering of the search results.
To make our experiment realistic, we used real search results that linked to real web pages. We also used a real election – the 2010 election for the prime minister of Australia. We used a foreign election to make sure that our participants were ‘undecided’. Their lack of familiarity with the candidates assured this. Through advertisements, we also recruited an ethnically diverse group of registered voters over a wide age range in order to match key demographic characteristics of the US voting population.
All participants were first given brief descriptions of the candidates and then asked to rate them in various ways, as well as to indicate which candidate they would vote for; as you might expect, participants initially favoured neither candidate on any of the five measures we used, and the vote was evenly split in all three groups. Then the participants were given up to 15 minutes in which to conduct an online search using ‘Kadoodle’, our mock search engine, which gave them access to five pages of search results that linked to web pages. People could move freely between search results and web pages, just as we do when using Google. When participants completed their search, we asked them to rate the candidates again, and we also asked them again who they would vote for.
We predicted that the opinions and voting preferences of 2 or 3 per cent of the people in the two bias groups – the groups in which people were seeing rankings favouring one candidate – would shift toward that candidate. What we actually found was astonishing. The proportion of people favouring the search engine’s top-ranked candidate increased by 48.4 per cent, and all five of our measures shifted toward that candidate. What’s more, 75 per cent of the people in the bias groups seemed to have been completely unaware that they were viewing biased search rankings. In the control group, opinions did not shift significantly.
Over the next year or so, we replicated our findings three more times, and the third time was with a sample of more than 2,000 people from all 50 US states. In that experiment, the shift in voting preferences was 37.1 per cent and even higher in some demographic groups – as high as 80 per cent, in fact.
We also learned in this series of experiments that by reducing the bias just slightly on the first page of search results – specifically, by including one search item that favoured the other candidate in the third or fourth position of the results – we could mask our manipulation so that few or even no people were aware that they were seeing biased rankings. We could still produce dramatic shifts in voting preferences, but we could do so invisibly . . .
We published a detailed report about our first five experiments on SEME in the Proceedings of the National Academy of Sciences (PNAS) in August 2015.
Read the rest at
https://aeon.co/essays/how-the-internet ... r-thoughts