Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
The Political Bias of Google’s Search Results
This past election season, Gallup found that Americans’ trust in mass media — newspapers, TV, and radio — had fallen to an all-time low. And that much should come as a surprise to no one (except, apparently, to members of the media).
But the electorate as a whole tends to view search engines — Google, Bing, and company — as distinct from the rest of the mass media. In fact, according to recent Google findings, 87 percent of the population turns to online search first when they have a question.
I’d posit that this implicit trust in search engines is due to the illusion of control that you feel when you conduct a web search. You feel like you have agency over which sources you view. You feel like you’re getting a balanced picture of reality by reviewing several results on the first page of the search engine results.
But are you in control? Are you getting a balanced view of the questions you search?
It turns out that the top five search results account for over two-thirds of all clicks, with the first three results garnering 55 percent of all clicks. This suggests that whatever Google’s algorithm propels to the top of the search results is what the public will perceive as being the most accurate and authoritative.
And that represents a huge vulnerability for a healthy democracy.
In their 2015 study, Robert Epstein and Ronald Robertson concluded that the order of search results can have an outsized influence on voter behavior — and in the event of a close election, the effect of this influence could even be profound enough to determine the entire outcome of the election.
All that was the impetus behind online-search marketer CanIRank’s new research into the question of whether there is a detectable political bias in Google search results, which was covered yesterday in The Wall Street Journal. [Full disclosure: I work with CanIRank and participated in the study.]
To conduct the study, CanIRank collected the top 40 search results (i.e., the first four pages of results) for a set of 50 politically relevant search terms such as “gun control,” “abortion,” “TPP,” “ISIS,” “Hillary Clinton,” “Donald Trump,” etc. In the months leading up to the election, these search terms garnered over 20 million monthly searches.
Next, we had a bipartisan panel of scorers rank each of the 2,000 search results on a 1-5 scale, where 1 represents content that demonstrates a heavy bias toward views aligned with the Democrat party or a progressive ideology; 5 represents content that demonstrates a heavy bias toward views aligned with the Republican party or a conservative ideology; and a 3 representing balanced or neutral content.
Some of the more unsettling findings from CanIRank’s research:
- Web searchers are 41 percent more likely to encounter a page with a Left or Far Left viewpoint in the top Google results, and 66 percent more likely to find left-leaning sites on page 1.
- 56 percent of page 1 results and 53 percent of overall results demonstrated a political bias.
- 11 of the 50 search terms analyzed contained no right-leaning results on page 1 of Google’s results. Those terms: abortion; minimum wage; nafta; Iraq war; campaign finance reform; global warming; marijuana legalization; tax loopholes; tpp; budget deficit.
No company in the history of the republic has ever wielded as much influence over how the public perceives important political issues. As CanIRank founder Matt Bentley writes, “This trend is particularly troubling when one considers that the employees of this private company do not reflect the ideological diversity of the country at large, and have consistently been amongst the largest donors to Democratic party candidates.”
Published in Technology
Really interesting!
Could CanIRank do any background research into optimization done by the various top sites?
Side note, glad to see you back!
“Google, where’s the balance?”
“Search Me”
Just one more way the left has command of the culture.
Thanks for the welcome!
Not sure this addresses you question entirely, but CanIRank evaluated the top rankers by approximately 200 factors that we know affect search rank including things like number of external links pointing to the website’s home page, number of external links pointing to the ranking page, domain authority, page authority, keyword density, related term density, page relevancy, content length, and so on.
We know (by Google’s own admission) that number of external links is one of the most important ranking factors. So we would expect to see that high ranking results for our political keywords would have more links relative to lower ranking results. This actually was not the case! As shown in the image below, we discovered that left leaning pages required significantly fewer external links to make it into a top 5 search result slot compared to a neutral site or a right leaning site.
Welcome back, Diane. A great return post.
That does answer my question, thanks.
Great post — I especially like that you looked at incoming links, which was part of the original “secret sauce” of Google’s search.
Diane, it’s good to have you back. Is there any chance for a Young Guns reunion with Troy, Meghan, and Keith?
Good to see you back.
I’m not sure how this is a search engine thing and not a content thing. NY Times, ABC, Wikipedia… all are popular already and have viewpoints. So I would expect an unbiased engine to turn up left-leaning at a far greater rate anyway.
Just as I would expect a handful of jellybeans from a jar of mostly black jellybeans to be mostly black.
No?
Diane,
This is excellent information. There are many many factors that can go into a search algorithm. This is providing an objective measure of the results irrespective of what is going on inside the black box. This is the kind of data that can’t be ignored. Confronting Google with this will be embarrassing for them. You don’t need to claim any expertise in search engines. Biased results are biased results.
Thanks for the post.
Regards,
Jim
Did you compare with other search engines, such as DuckDuckGoGo?
Holy crap. Diane Ellis. I was so flabbergasted to see you I couldn’t get through the article! Welcome back. Hope you stay awhile.
Now I gotta go read what you wrote!
Valid point. However, for our study, scorers evaluated each page individually. So for instance, of the 25 results from abcnews.com, two received scores of “2” for moderate liberal bias, while the 23 received scores of “3” for neutral. Of the 45 results from the NY Times, only 11 were scored as “1” or “2”, while the rest earned “3”.
Now, I should note: it’s my personal opinion that the study’s scorers –though they were bipartisan–erred on the side of scoring MSM material with neutral scores, even though I’d personally score them as left leaning or far left. But this only underscores the point that truly far left stuff outranked the right’s content. An example below – top 10 search results for the keyword “clinton emails” shows that the top 10 results on a keyword that should have represented a major vulnerability for the Clinton campaign average a 2.9 across the first page of results (just slightly left of neutral). Looking at these results, do you think the scoring is spot on, or would you have ended up with an average score below 2.9 for these 10 results?
By the way, I owe a debt of gratitude to @bereketkelile for helping us find someone to round out our panel of scorers. As you might imagine, evaluating 2,000 web pages is no quick task!
How does their traffic compare? Among TV news media, Fox News has as many viewers as all the hippie networks combined. Even if hippie websites far outnumber conservative websites, is there a site on the Right with sufficient traffic to make absence from early search results suspicious?
Does Google’s algorithm ignore news aggregators? If not, how does Drudge’s presence in search results compare to less popular aggregators?
It the engine reflects bias, it isn’t necessarily deliberate bias. For whatever reasons, hippies are often unaware of their own biases (per Diane’s point about scoring).
It’s up to us to get those results back in balance by providing compelling content. I’m happy to report that more than once, I have found a Ricochet post in the top five search results. If only I could remember what the search keywords were…
Sadly, if you’re using Google for your search this is more likely due to your previous search and browsing history influencing what Google thinks you want to see rather than what the best information might be. I try and use DuckDuckGo almost exclusively as a way to avoid this bias since they don’t save any of your browsing habits and my hope is that leads to more impartial results (though it’s probably still influenced by many of the same ranking biases highlighted in these findings)
This is no real surprise to me as computer science student. I have been experimenting with looking for alternate search sites like duckduckgo instead of google who I think is defiantly one of the worst when it comes to both corporate and political bias in search results.
I actually think the corporate bias is almost as bad for society as the political in some ways.
Great post though, love to see the details.
Hi Diane! Nice to see you!
People are suggesting going to another search engine. I can see the point. But it worries me.
Imagine a future in which it is not just that progressives and conservatives tend to read different sites, but they are using different search engines. At least now, a progressive might just stumble across a website with a different point of view, even if it appears on the second page of search results.
In my imagined future scenario, with two separate and parallel search engines (one used by progressives and one used by conservatives) the bubbles we live in would only get thicker and more impenetrable.
This would be worse for progressives, because at least conservatives are exposed continually to progressive viewpoints via TV and popular culture.
Diane’s back!! *runs to tell everybody*
#makericochetgreatagain
Google is reticent about giving details of their algorithms to avoid letting people “google-bomb” the results.
They’d rather do it themselves.
Doesn’t Google also adjust for the searcher based on previous searches? Would you and I get the same results?
Also, aren’t results influenced somewhat by aggregating all searches and clicks? Thinking about myself, I don’t really search for news. I would never think to search for news by topic. I might come directly to Ricochet or national review but I’d never go there via search.
I could imagine, though, being 15 or 20 and being more caught up in hot topics or trying to learn about a topic and clicking results that sounded familiar like CNN and overlooking unfamiliar like NR. Or even avoiding FOX because I heard they were evil.
IF our collective searches and clicks influence and IF conservatives and liberals have different click and search habits then a perfect engine would still lean left.
Really, I guess my question in a nutshell is, How could a Neutral, unbiased search engine do anything but return results that tend left?
So not only can we not believe what we read on the internet, now we can’t trust what the datalords tell us is on the internet? Unfortunately, too many people view Google, et. al. as a public service, as opposed to a business geared toward pages views, ads, profit, and apparently, passive propaganda.
But all that aside, I’m just thrilled to see a post from Diane!
Or, it might represent a convenient starting point for doing some research. Personally, I reject the whole idea that we should rely on the objectivity and accuracy of any journalist or blogger. The claim of objectivity and accuracy in second-hand accounts is always a myth. When I google, it is to find original source material, which is the only thing that I trust. If I click on a secondary source, it is with the full expectation that I will click again (and perhaps again and again) to get as close as possible to what really happened.
Spot on. Good information can be found, but one often has to go digging for it. If you’re conservative, it appears you have to work harder to find it, though.
Its good to see you back. Its better to see you front.
I think that one can accept that assumption, establish it as a norm or baseline, and still measure actual results against what one would expect to get granting your assumption. For example, if we could “expect” to get results tilting 60-40 left based on a neutral search engine, and get results that are 70-30, there’s more of a bias than is explainable by the sources. Probably not easy, but doable.
I agree with Casey here.
Yes the media is biased, yes it is possible that search engines are biased. However, if we all we conservatives do is complain about biases we are never going to win the culture. What we need to do is make the media and popular search engines biased toward conservatism. We need to strive to make conservatism and conservative ideas popular. That should be the end game.
Does anyone remember when Google was tagging pictures of African Americans as gorillas? I doubt that Google’s programmers let that happen on purpose. As a programmer, I have a hunch that the algorithms were developed using the programmers’ own photo collections (we have such pristine QA regimes), which included their White and Asian family and friends. But that’s OK, because the White and Asian testers proved that the algorithms correctly tagger their family and friends.
I can imagine the same thing happening here, due simply to the lack of philosophical diversity. Some data scientists say that machine learning is like money laundering for bias. You train the algorithm with your own implicit assumptions and blind spots, but when results come out accordingly, you say, “it’s not me, it’s the computer.”