Contents:
Simple Version
Example – Community Standards Leading To Censorship and Contradiction
Example – Anodyne Preference/Nonviolence Leading To Censorship and Support Of Nazis (and a bunch of others)
Fuller Treatment Of Appropriate Response
—
Simple Version
Community standards are whatever the mass of the people think is acceptable at the time. A cursory understanding of history will inform you that royalisms like racism, classism, sexism, and almost every other ill has been tolerated by some group of people at some time; therefore to filter or massively downgrade based on community standards means that your service reflects the prejudices of the mob.
Likewise, scientific and historical consensus itself is not as reliable as may seem. In addition to the widely acknowledged limitations of the science that were superceded by improvements such as quantum mechanics, the mainstream of science also endorsed phrenology, the basic theory of humours, and the carbohydrate-saturated diet, amongst many other errors. Likewise, the historical views of the past often are shaped by incomplete information, or information inconsistent with later available texts and archaeological evidence. Therefore to filter or massively downgrade based upon one’s perception of the scientific mainstream/consensus is to amplify the errors of scientists and historians in addition to amplifying the truths.
When providing a service such as a search agent, a content host, and similar, one must bear in mind the useful ends of such things. If the advertised purpose of these is to “bind like-minded people” or similar mission, of course applying these types of standards would be consistent with that mission. However, that mission essentially is entertainment; there is no personal growth possible from such echo chambers, and no confidence to be gained from what you know is already a distorted and incomplete picture of reality. Hence the service is an expendable toy.
Your response, as the user, then is to find an actually useful service, that presents more balanced viewpoints. Your response, as the provider of such service, should be either to acknowledge your limitations, or to present a fuller spectrum of facts/opinions in your results/content, so that you are reliable and therefore useful as a reporter.
—
Example – Community Standards Leading To Censorship and Contradiction
To make the point using a concrete, real example, I will refer to Google’s Search Quality Evaluator Guidelines as of July 20, 2018. The document is copyrighted and likely isn’t safe for work, either, so be careful about this link:
https://static.googleusercontent.com/media/www.google.com/en//insidesearch/howsearchworks/assets/searchqualityevaluatorguidelines.pdf
The above quality guidelines document has some good pointers in providing the function of a useful tool, but also features a telling example.
The example is the evaluation of a website featuring “proof that Islam is evil, violent, and intolerant – straight from the Koran”. The query term cited is “islam”.
It is a general search term; neither I nor anyone else is going to be able to give a specific user intent/request from it. The response to that search term then must reflect what this service thinks it’s about; that is, what does it advertise in response to such terms?
In the case of this search provider, their advertised interpretation is:
User Intent: Find information about the religion of Islam.
a worthy and reasonable default, others are possible.
The summary of the ratings for that site is as follows:
– Fails To Meet (based on upsetting/offensive; however, other statements are given, so it’s not 100% clear that it’s only based on upsetting content: “This is a broad query seeking information about Islam. The LP was created for the purpose of stating the author’s opinion that Islam is not a religion. The author lacks expertise in the topic; the writing has frequent misspellings, typos, and grammar errors; and the page seems to exist to promote intolerance or hate.”)
– Upsetting/Offensive Flag
It’s unclear what the content quality rating for the page is based on the recommendations.
Both of these ratings will cause the content to be hidden unless the user issues an explicit query to pull the page or similar pages (basically you have to know what you are asking for before you can get it, so that response fails the purpose of an information query, if the information on the site is in fact relevant and accurate).
By contrast, consider the opposing site commentary:
Teachers Guide – Muslims | Frontline | PBS (the United States government-partially-supported broadcaster)
Commentary: This is a high-quality article on a reputable site, with an accurate summary of the major beliefs and practices of Islam. Do NOT assign the Upsetting-Offensive flag, regardless of what the query was.
To dissect this:
– A “high-quality article” – what does that actually mean in the case of a religious query? If I write “Christians worship Jesus, who told them to turn the other cheek” is that really accurate? Didn’t Jesus engage in acts of civil disobedience/criminal behavior, like overturning the money changers’ tables? Didn’t he say something about “Think not that I am come to send peace on earth: I came not to send [or bring] peace, but a sword.” (Matthew 10:34) Haven’t there been many Christian holy wars in the past, and Christians who waged war?
You understand that the final evaluations of all these is not obvious, but they relate directly to the accuracy of the page, which is the entire point of using (in this search provider’s advertised mission) “Expertise, Authoritativeness, and Trustworthiness”. None of those advertised criteria matter if the page is inaccurate; the service should not consider that page a useful response to the user’s request.
Furthermore, if the page is inaccurate or misleading/omitting key details, how could you consider that the page/site actualized “Expertise, Authoritativeness, and Trustworthiness”?
In this particular example, there is a hilariously obvious counterpoint: the “non-upsetting/offensive” site has a section about “Conversion to Islam” which directly contradicts the information presented on the “upsetting/offensive” site. The point I am making is not about which one is right (although the “upsetting/offensive site” is closer at least to the literal truth), but that it is not at all obvious that this is a high-quality article, even on a cursory skim, from a worldly person, albeit one who would not claim he is an expert (though if a non-expert can see this problem, then that indicates the situation is actually worse than would be assumed).
“a reputable site” – do you think a US government funded site, additionally funded by liberals, is going to represent something different than kosher viewpoints to these two groups? Did practicing Muslims write this page? (I did not see an author attribution on the excerpt and most likely I wouldn’t have any idea of whether they are “mainstream, reputable Muslims”, even if I did) Again, note the points I made on the accuracy of the information on a cursory glance. Where is this assessment of reputation coming from in regards to this topic? We are left to assume that the US government and liberals are reputable on this topic without even any indications of proof. (Note: the actual evaluation of reputation comes from the fact that Frontline does present at least relatively balanced news that can be verified and you know its provenance, similar to newspapers etc. However, that is not reflected on this page, nor in the search provider’s commentary.)
By contrast, the “upsetting/offensive” site names an author (this is the Internet, she could be a dog), the author cites why she thinks her claims are correct/on what authority basis, and puts direct quotes from religious text in her page.
Hence, even from a purely academic point of view (e.g. authorship named, references, logic shown), the more reputable site clearly is the one that this search provider’s criteria downgrades.
“an accurate summary of the major beliefs and practices of Islam” – well, sort of accurate; there are some other factual errors and disputes from which I am sparing you because hashing out religious disputes is not the purpose of this particular writing. Nonetheless it’s clear the evaluator(s) don’t know what they’re dealing with and so don’t know what they’re talking about.
Of course, the assignment of the “Upsetting/Offensive” flag is hilarious in light of the above notes; I would be upset if the search engine showed me a page with several clearly controversial topics on it. Furthermore, a religious topic, by its inherent subject matter and often by its inherent ambiguity and contradiction, has a high probability of bringing up upsetting issues, even if the authors attempt to avoid such provocations. I don’t wish to attribute an intent to individuals I never met on the basis of their handling of this one issue, but it’s clear that their assignment has nothing to do with the actual meaning of those words, and little to do with the functional distinction.
As you have read, the evaluator(s) attempts to apply their community standards to these pages, would result in a search engine trained this way, to draw two completely different conclusions about two ostensibly meritorious pages on this topic. Hence the results are not even coherent, let alone correct: graded only on accuracy, the community standards downgraded page probably is the best one.
—
Example – Anodyne Preference/Nonviolence Leading To Censorship and Support Of Nazis (and a bunch of others)
Consider the general statement, across service providers, not just the one previously noted, along the lines of “advocacy of harm, words that could lead to harm, threats, and other similar behavior are not allowed”.
Certainly any advocacy of war or armed opposition does not conform to this policy.
It so happened that in the late 1930s, the people of the United States (and in point of fact, the rest of the world) faced a decision regarding the expansionism of the Axis powers, Germany, Italy, and Japan. Perhaps to oversimplify, they had the option either to oppose them with armed force, or to stay out of the conflicts and defend themselves. The vast majority of the American people initially stood with the isolationist viewpoint, and only after the Japanese attack on Pearl Harbor in 1941 did this opposition cease.
At the same time, US president Roosevelt proposed the lend-lease policy and other actions, which would ally the US with the opponents of the Axis, hence giving the Axis powers clear cause for war. Further, the Japanese felt that the US imperial policies were intended to encircle them, binding them into an unsustainable economic sphere and therefore ensuring continued humiliation and suffering at the hands of the West; the events in mainland China over the previous century, and certain others in the farther East, such as the forced opening of Japanese ports, were consistent with this view.
Of course nowadays, (and even to an extent it was known back then,) we recognize that the Germans, Italians, and Japanese engaged in pointless war (almost a tautology), and on top of their belligerence, inflicted completely unnecessary suffering in the form of rapes, other tortures, massacres, starvation, and other privations upon their subject peoples. Hence we feel it justified to have conquered them. But consider the matter from the perspective of these service provider terms of service…
Clearly Roosevelt’s proposed actions could have induced and threatened war; on top of the threat already imposed by imperialism, certainly no defence of such actions could be considered in any way to comply with this service provider nonviolence policy. However, in full consideration of the totality of events, and in particular that the US atrocities, even considering the segregation and certain concentration camps they enforced, still were far better than the conditions produced by the Axis powers and by the Soviet Union – with that knowledge, we reasonably (and probably correctly) believe that in fact, a more aggressive policy than what was implemented, would have been optimal at the time of the late 1930s. If the people of that time had advocated either the Roosevelt policies or the more likely correct direct conflict policies, any service provider that enforced this policy, would have censored clearly better options, in a matter of the gravest public import.
This foolishness extends well beyond that time: consider the matter of the “global war on terrorism”. To oversimplify somewhat, the GWOT involved attacking countries that seemed to harbor terrorists, in a somewhat unpredictable and unregulated fashion. Yet people demanded retaliation for 9/11 and correctly opposed those who advocated caliphate Islam (and often really bad government underneath it, the biggest problem). To advocate this retaliation, and the liberation of various countries, of course incites violence, and hence these service providers, if they were to adhere to any of their own policies, should censor the entire US government and everyone who advocates this retaliation.
Yet this is not the end of the censorship: consider the matter of police unjustified killings, gang killings (e.g. MS-13), or really any notorious killing at all. Prior to a trial, which, in the absence of egregiously obvious guilt and exigency based on the risk of further violence, the due process of law required to send these people to prison (or death) has not been performed. So what happens when the mother of the slain calls out for justice against the suspects? Is she allowed to call for the arrest of the criminals, which we would assume will require force? Is she allowed to call for the death penalty, especially in a poorer country that already can’t afford proper prisons? The viewpoint enforced by the nonviolent service providers would be, no, those pages should be censored. Hence any mention of an ongoing investigation into crime, could not be published, and in the strictest sense, no proceedings could be, either, and maybe not even the punishments themselves could be mentioned – after all, isn’t the law code promising to use force/violence against those who break it?
Just as context: these service providers usually then say, “We comply with all relevant laws…” – often that may be correct, at least of the time. But usually, no law compels their services to exist (most dictators would like to see them disappear), or for them to publish anything. So that’s a non sequitur. Either they have a consistent policy or they don’t. Indeed, as time has gone by, many of these famous service providers have added a series of “clarifications” that create an even more absurdly contradictory censorship regime.
—
Fuller Treatment Of Appropriate Response
If a service really just wants to be entertainment, or a limited/focused topic forum (e.g. automobiles), that’s fine if they advertise themselves this way.
Otherwise, the service has to be explicit about what they show and what they don’t show, so a user can’t get a mistaken impression about what they’re going to get out of this thing.
Some good examples are:
A link of “mix in fact-based opinion results” if you want your default most-relevant to be hardest-facts only
A notice of “Islamic State propaganda illegal in your country has been filtered out”
A drop-down menu of “non-mainstream/less-clicked-on search results categories” that includes “nationalist perspectives”, “feminist perspectives”, “mainstream critics” if you want your service to show mainstream sources and big organizations first based on popularity, but still provide the capability for online research and individual voices
A checkbox for “show strident voices” if you are concerned about children or upsetting the audience, that you keep enabled via cookies (if user accepts)
A checkbox for “show graphic/violent content”
A checkbox for “show likely rumors/unverified news”
Another possibility is to show a default set of buckets based on editorials, such as showing IPCC, IPCC + mods, alternative climate science, and climate deniers, and each gets a portion or sub-header on your search results. Of course that requires more work.
Likewise, if you want to run a post promoter that runs off AI you don’t fully understand, just say so: your results may vary, this is not an unbiased media source, it runs according to these rough principles and training.