This article is reprinted from Bill Dembski, design theorist William Dembski’s Substack, with his permission.

In 1902, Vladimir Lenin published What Is to Be Done? This political tract changed the Russian communists’ approach to their movement in Russia. Lenin argued that spontaneous worker uprisings were insufficient to bring about socialism. He insisted on the need for a highly disciplined, professional revolutionary party to lead the working class. He also emphasized the role of “vanguard” intellectuals in raising political consciousness, guiding the proletariat beyond trade-union demands to a full revolutionary struggle against the Tsarist regime. I wonder if something similar is needed for a full revolutionary struggle against the Googlist regime.

Google will not admit wrongdoing

First off, let’s be clear that Google is highly unlikely to admit doing anything unfair or underhanded in its search rankings. For instance, it dismisses the work of Robert Epstein as “nothing more than a poorly constructed conspiracy theory.”   Defending the firm to the Washington Post, a Google spokesperson added, “We have never re-ranked search results on any topic (including elections) to manipulate political sentiment. Moreover, we do not make any ranking tweaks that are specific to elections or political candidates, period. We always strive to provide our users with the most accurate, relevant answers to their queries.”

The original use of the term “conspiracy theory” to disparage a position can be credited to the CIA. It was coined when the agency was attempting to deflect attention from its complicity in the assassination of JFK. The CIA memo of 4/1/1967 titled “Countering Criticism of the Warren Report” includes these sentences: “Conspiracy theories have frequently thrown suspicion on our organization, for example by falsely alleging that Lee Harvey Oswald worked for us. The aim of this dispatch is to provide material for countering and discrediting the claims of the conspiracy theorists, so as to inhibit the circulation of such claims in other countries.”

As I read Google’s denials, I’m reminded of the famous scene in the film A Guide for the Married Man (1967) where an experienced philanderer (Robert Morse) teaches a would-be philanderer (Walter Matthau) to deny any and all infidelity even if caught in flagrante:

Google has internalized this lesson.

Regarding Epstein’s research, let me urge readers who think he may be blowing smoke to look into it and then come to their own decisions about it (I provided links to his research here). Don’t take my word for it — but also don’t take Google’s word for it. Some have criticized Epstein’s work for extrapolating too far beyond his data. But even if the effects are not as extreme as he claims, he would at worst be somewhat overstating Google’s influence over voting preferences and democratic processes. That influence would still be real and palpable.

What we can learn from an insider

Epstein has worked on unmasking Google as an outsider. He operates as a reverse engineer who scrutinizes what Google is doing in public and draws conclusions from there about how Google is distorting search results to shape people’s preferences. But Google has also faced unmasking from inside, notably by former Google employee turned whistleblower Zach Vorhies, whom I discussed here. In his 2021 book Google Leaks: A Whistleblower’s Exposé of Big Tech Censorship, Vorhies details the political bias that infects Google. He recounts the moment it became clear to him that Google was betraying its objectivity in internet search.

It was the aftermath of the 2016 US presidential election. In response to the election result, Google leadership determined that they would do everything in their power to prevent populist and low-information voters from deciding future presidential elections. To that end, they would implement “machine learning fairness” in Google searches to steer users toward “good information” (promoting the right sources holding the right political views), and effectively censoring “bad information.” Most users would remaining completely unaware of this sort of paternalistic nudging (compare Nudge (2021) by Richard Thaler and Cass Sunstein).

How Google overrides its page ranking  system

Google leadership became convinced that the 2016 presidential election was unfair because of fake news (mis-, mal-, and disinformation). Consequently, Google needed to provide an appropriate counterbalance.

 To dispassionate eyes, “machine learning fairness” is simply putting your hand on the scale to achieve the balance you want. In effect, ut is a way for Google to override its basic algorithm. Google’s basic algorithm is PageRank, which is supposed to order websites in response to user queries based on the quantity and quality of links to those sites. But Google doesn’t just unleash its algorithm and let it have its way. Instead, it adds overrides that deliver results at variance with the algorithm when the algorithm produces results that Google doesn’t like.

According to Vorhies, one way Google overrides its algorithm is by instituting blacklists that block results when some term on the blacklist appears in a search query. He provides pages of blacklisted terms that he was able to find in Google internal documents while he still worked there. Another effective way to override the basic algorithm is to add a layer of machine learning that simply skews results in favor of some preferred ideology or business outcome (“machine learning fairness”).

Still another approach to overriding the algorithm is simply to impose a manual penalty, downgrading individual sites so that they rank poorly in Google search. Taken to its logical conclusion, such a manual penalty can mean deindexing a site so that it doesn’t even appear on Google, thus banishing the site to oblivion. Now I don’t mean to give the impression that such manual penalties are never warranted. Obviously spammy sites and sites depicting and advocating extreme violence would qualify for deindexing. But what we’re talking about is nothing like this. We are concerned with widely held viewpoints that Google is deliberately suppressing.

To the computer-science purist, what I’m calling an override is of course ultimately also part of the Google algorithm. Even manual overrides need to be entered as part of Google’s codebase. Still, these overrides are essentially addenda and exceptions to what would otherwise be a conceptually clean algorithm that handles search queries fairly without built-in bias. These overrides can be compared to epicycles in the old Ptolemaic cosmology. As observations continued to pile up and contradict Ptolemy’s theory, epicycles within epicycles (exceptions within exceptions) needed to be added to the theory (and in this case, to Google’s algorithm).

Next: What are my recommendations for reining in/reforming Google?

Here are the previous instalments: The evilization of Google—and what to do about it Part 1: Understanding Google’s dominance over the internet. Nothing is totally evil. Still, there’s enough evil in Google that it is, for now, more on the side of Darth Vader than Obi-Wan Kenobi.

The utter dependence of online businesses on Google. Part 2: An SEO business needs to please Google or else it is dead in the water. Google’s re-presentation of information created by others makes it less likely that users will visit them. Thus Google’s business expands at their expense.

Image Credit: gustavofrazao - Adobe Stock

A potential chink in Google’s armor: Loss of legal immunity Part 3: Currently, Google is legally protected from the consequences of frequent copyright violation. One outcome of the resulting ad clutter is that, unless you are top of the search, you’re likely wasting time trying to make money off organic search.

Becoming a slave to Google: How it happens Part 4: After an update, you always have to second guess what Google did. You become a reverse engineer who never sees under the hood. This leads to a mentality that always tries to second guess whether Google will favor some piece of content or way of expressing it.

Google’s power over online business: Monopolistic and extravagant Part 5: To help people understand Google’s power over online businesses, I ask. What if the road on which the local Walmart is located suddenly became a dirt trail? It’s as though Google is doing everything in its power to throttle organic traffic, driving that traffic to their sponsored ads, increasing its bottom line.

How Google search rank disrupted the alternative health industry Part 6: Google’s bias is as real as ever, but the firm has deflected it to an impersonal policy that gives an air of objectivity. Having a Wikipedia entry gets you instant credibility even though so much of Wikipedia is substandard and politically biased.

How Google can control and manipulate public opinion. In Part 7, Robert Epstein, former editor of Psychology Today, offers some disturbing information on how search engines can be manipulated for political purposes. Epstein, working through web projects, also provides a sobering assessment of how much Google monitors what average Internet users are doing online.

and

What public policies can help us achieve a less biased internet? Part 8: Robert Epstein proposes a number of changes worth considering. Robert Epstein is the former editor of Psychology Today who, as I discussed yesterday, has studied the ways in which Google can manipulate public opinion.