Slight tanget, but it might be interesting to some on search results and how Google's search results can sway users.
Googles been able to use ephemeral experiences (like autocomplete and answer boxes) to influence users, especially undecided voters. The research, which was reproduced by a German team, is believed to have influenced 2+ million people in the US election alone --
That testimony is highly misleading. Taking the highest estimates possible, which are based on essentially a conspiracy theory, the number of influenced voters is much lower than 2 million.
He claims they moved 2.5 million votes, not voters. This relies on an assumption of straight ticket voters and about 20 votes per voter. He doesn't correct people when they make the votes/voters mistake.
And even that new number, 100k voters swayed is not really well founded. He makes a couple of overlapping claims, one of them I looked into deeply and found that using his own papers, a reasonable way of expressing it was that 8-10k additional votes for Clinton was a reasonable upper bound, as he claimed, might be affected.
And the number was likely lower (and was only that high because more people identify as democrat than republican in the US).
(I work at Google, but take interest in this mostly because it's just terrible abuse of mathematics)
Googles been able to use ephemeral experiences (like autocomplete and answer boxes) to influence users, especially undecided voters. The research, which was reproduced by a German team, is believed to have influenced 2+ million people in the US election alone --
take a listen on the congressional testimony http://naplay.it/1157/37:25 (recommended listen speed is 1.5X)