Dr. Robert Epstein on politics: what the evidence says · JRE #1768
SUBJECT: POLITICS
The short answer
Epstein claims his research found Google search bias shifted 2.6 to 10.4 million votes toward Hillary Clinton in the 2016 election without voters' awareness. Epstein and colleague Ronald Robertson published peer-reviewed experimental work in PNAS (2015) showing that biased search rankings can shift undecided voters' preferences in controlled lab settings.
Not a true/false call. Every claim is logged with its sources; read the exhibits below.
we knew from the experiments we had run that that was enough bias to have shifted over a period of time among undecided voters somewhere between 2.6 and 10.4 million votes without anyone having the slightest idea
What the evidence says 01 / RECORD
Epstein and colleague Ronald Robertson published peer-reviewed experimental work in PNAS (2015) showing that biased search rankings can shift undecided voters' preferences in controlled lab settings. The specific 2.6-to-10.4-million-vote figure for the 2016 election, however, comes from a separate, non-peer-reviewed white paper and later congressional testimony, not from a published, peer-reviewed study; it extrapolates from a small 2016 field observation of 95 subjects (21 undecided) over 25 days, using an unpublished computational model. When the same 2.6-to-10.4-million figure was cited (via a distorted "up to 16 million" framing) by then-President Trump in 2019, PolitiFact's review found outside experts identifying major methodological problems: the study lacked a clear definition of "biased" search results, relied on crowdsourced subjective bias ratings, excluded Gmail users based on a speculative theory that Google was intentionally showing them unbiased results, and never publicly disclosed the formula used to extrapolate from a few dozen subjects to a multi-million-vote national estimate. Independent researchers said there was insufficient information in Epstein's materials to verify that his model validly translated observed search-result bias into vote totals, and Google called the claim inaccurate and previously debunked. As of this writing, Epstein's vote-shift estimates for 2016 have not been independently replicated or published in a peer-reviewed venue, and the methodology remains contested.