On that Wikipedia page, the two Robert Epstein sources (one a PNAS paper[1], the other a Politico article[2]) are theoretical ("whether they could be manipulated") and pre-2016. The other main source[3] is similarly theoretical and provides policy recommendations to combat SEME.
It also has a Washington Post article[4] discussing a (since-deleted) video that purportedly showed potentially-manipulated differences between how Bing and Google autocompleted "Hillary Clinton" searches. But none of the 3 main sources cover what the impact of autocomplete could be.
Do you have any links about Google being caught red-handed politicizing ranking order?
>>The 2020 presidential election monitoring project successfully recruited a diverse group of 732 registered voters in three battleground states: Arizona, Florida, and North Carolina. It preserved more than 500,000 ephemeral experiences on Google, Bing, Yahoo, YouTube, and Facebook. These are the fleeting experiences that Big Tech companies use to influence votes and opinions worldwide and that are normally lost forever.
>>A preliminary analysis of the data shows that Google search results (but NOT search results on Bing or Yahoo) had a significant liberal bias – enough to have shifted at least 6 million votes in the months leading up to the election.
You need to know the methodology being used to understand the results, as they tracked ephemeral experiences using surveillance software with volunteers.
This guy needs a website team in the biggest way. It's hard to find the information and it's scattershot. It's very frustrating because the work he's doing is solid. Don't let the shitty web presence deter you.
The Rogan podcast was awkward af but he does provide a lot of substantive content if you're open to listening.
There's no research on that page. Many of the linked articles that this individual has written also are vague on actual details/proof and only contain vague "this could happen!" comments. Beyond that, many of them also end with a plea for more fundraising.
>This guy needs a website team in the biggest way. It's hard to find the information and it's scattershot. It's very frustrating because the work he's doing is solid. Don't let the shitty web presence deter you.
All he needs to do is spend a little more time uploading data and evidence, and less time shoving as many "DONATE" buttons onto his page as he can. None of this looks like solid work, even his television appearances; it looks like someone who goes around throwing out lots and lots of loose speculation while asking to be paid to do so.
OK, his SEME bias analysis is based on recruiting people through a "passive monitoring system," which recorded SERP rankings over time (what Epstein deems "ephemeral experiences"). Participants' SERP rankings, and raw HTML of each of the top X links, were stored on Epstein's servers. He then had "online workers" (likely mturk?) rate each archived page as pro-Trump or Clinton.[0]
He submitted a paper on this in June 2021 to the AABSS but they have yet to publish it. He also has 5 other, earlier SEME papers submitted for publication.[1]
I appreciate you going the distance on providing these links but I'm loath to trust this data until it's been peer-reviewed.
The graph depicts 300k undecided voters resolving to a party over time. Here's Epstein talking about it:
> ...by the time we hit election day, we have an enormous gap - more than 100,000 people - between the conservatives and liberals. And we've created that gap with our biased search results.
Assigning that movement to a single cause seems unlikely to me.
His presentation does include a relevant WSJ article titled "Google Workers Discussed Tweaking Search Function to Counter Travel Ban."[3] That's not "red-handed," since they didn't do it, but it's very relevant. HN at the time: https://news.ycombinator.com/item?id=18036353
It also has a Washington Post article[4] discussing a (since-deleted) video that purportedly showed potentially-manipulated differences between how Bing and Google autocompleted "Hillary Clinton" searches. But none of the 3 main sources cover what the impact of autocomplete could be.
Do you have any links about Google being caught red-handed politicizing ranking order?
[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4547273/
[2] https://www.politico.com/magazine/story/2015/08/how-google-c...
[3] https://www.jstor.org/stable/10.5325/jinfopoli.9.2019.0370#m...
[4] https://archive.is/fHzW8. Is Politifact trustworthy? They investigated this and found "Google’s suggested searches, for the most part, avoid offensive suggestions for everyone, not just Clinton." That isn't a total debunk, of course. https://www.politifact.com/factchecks/2016/jun/23/andrew-nap...