The problem with ogregores?

During a discussion about tech & mythology, William Kuskin challenged me to explain what problem I was trying to solve. Here’s an attempt to answer the question. 

The problem

The problem is rooted in the existence of greater-than-human techno-social entities that profoundly affect us. 

I call them ogregores. An ogregore is a highly connected group of people and tools with a common purpose. It's a techno-social structure that acts in ways that cannot be directly reduced to its constituents. The term is inspired by egregore, defined by Wikipedia as “an occult concept representing a distinct non-physical entity that arises from a collective group of people.”

The problem arises because we focus on people, especially high-status ones, rather than groups that effectively act as individuals. For example, both history and the daily news focus on exceptional individuals. That means that we don’t pay enough attention to ogregores. We fixate on the trees and miss the forest, which impairs our understanding of ogregores and our ability to predict how they will act.

This deficiency is a problem because ogregores can command substantial resources and can project a great deal of power. 

In other words, there are powerful aliens in our midst that we are not well-equipped to see. Some people may experience ogregores as monsters; H.P. Lovecraft comes to mind. Cf. “ogre” and “orc” on Etymonline.

The challenge, in other words, is to “see” ogregores. 

“See” is short for “making accessible to the senses and the understanding.” One could say that the question is how to make ogregores sensable [sic]. I’m open to many modes of perception including stories, analysis, performance, images, and sculpture. The means can be digital, analog, or some combination; interactive or static. Cheshire & Uberti’s delightful Atlas of the Invisible is a not entirely dissimilar project. 

Notes on ogregores

Corporations, government agencies, and collective efforts can (but need not) be ogregores. Not all social groups are ogregores. My current minimal conditions for an ogregores are: (1) Constituent people share a social characteristic; (2) there are 2-way social ties between members; (3) the group as a whole has agency (e.g., it is differentiated, autonomous, interactive, and adaptive); (4) it acts and delivers outcomes.

Ogregores are not people, though people are a necessary ingredient. Other ingredients include tools like norms, protocols, data records, and processors. 

Ogregores such as technology companies and states are powerful, pervasive, and mysterious. These are the attributes traditionally ascribed to deities. Ogregores differ from us in their nature and character—they are aliens, and some have godlike power.

Leaders and their actions are not sufficient to explain the behavior of ogregores but may be necessary. Exceptional individuals can make a big difference in the world, especially in the behavior of groups. It may be necessary to include those individuals in explanations. However, it is not sufficient since group dynamics also play a role. 

Mythological thinking can help make sense of ogregores because myths are a well-worn way to make sense of social phenomena. Archetypes can also help us make sense of our relationship to ogregores. In an interview with Nick Hornby, David Simon said, “But instead of the old gods, The Wire is a Greek tragedy in which the postmodern institutions are the Olympian forces. It’s the police department, or the drug economy, or the political structures, or the school administration, or the macroeconomic forces that are throwing the lightning bolts and hitting people in the ass for no decent reason.”

How ogregore action might be dangerous

Here are some examples where several similar organizations (putative ogregores) engage in similar behaviors. Actions are driven by mixture of powerful individuals, context, incentives, institutional structure, culture, etc. I will focus on cases where charismatic leader effects are less salient and actions can plausibly ascribed to the group. The actions of Tesla and Apple might be credited to Great Founders  but police departments, social media companies, and Android smartphone makers all behave sufficiently alike to make group action plausible. Loosely speaking, I’m looking at “species” of ogregores suggested by patterns of behavior.

• Most media are funded by ads. Ad rates depend on engagement. The need for engaging content motivates platform designs that encourage polarization and misinformation. 

• The WSJ reported that TikTok serves up sex and drug videos to minors (Sep 2021). Parents of young girls have sued TikTok, claiming it knew or should have known that its product was “addictive” and was directing children to dangerous content (NYTimes, Jul 2022).

• Instagram causes significant teen mental-health issues (WSJ, Sep 2021).

• Social media suppress some kinds of content but not others. A few companies thus act as gatekeepers to the public square. More generally, the algorithms of dominant content platforms shape popular culture through the content they promote. (“Algorithms” is metonymy; it includes business models, designer values, code architecture, advertiser priorities, etc. This is not new; the same happened in the Fifties with broadcast networks.) 

• Police departments everywhere disproportionately kill marginalized young men (Brookings, Dec 2020).

• Forced assimilation of indigenous kids at residential schools led to many unreported deaths (NYTimes, Mar 2022). This was not isolated behavior but an institutional disposition.

• There is recurring sexual abuse of minors in religious and other community settings (Phys.org, Aug 2020).

• Student bodies at several universities have silenced speech they consider dangerous (Forbes, Sep 2021).

• Supposedly not-for-profit health systems pay their CEOs tens of millions of dollars, exploit loopholes in federal law to get cheap drugs but don’t pass on savings, and don’t help—and sometimes burden—indigent patients (NYTimes, Feb 2020).

Why it’s hard to “see” ogregores

It’s hard to grasp the aggregate interactions of ogregores that have many different touchpoints with individual humans. It’s also hard to make sense of their actions when they are private or not easily visible to the public. Thinking in terms of ogregores may be especially useful in such cases of “single intent, diverse outcomes.”

To set up the contrast, here are two examples where it’s easy to see ogregore action:

• Fast food outlets have standard menus; all customers in a region see the same thing. Their impact on health and nutrition may be bad (a la “Super Size Me”) but we can all see the workings (or at least, the outputs).

• Interest rate changes by central banks are specific, uniform, and public. The same goes for antitrust actions.

Here are some examples where the actions are both diverse and obscure:

• There are roughly 18,000 police departments in the United States. Law enforcement oversight is hard at the best of times, and this hyper-decentralized structure makes it even harder to observe systematic problems like patterns in police killings. Up to 2014, the year Michael Brown and Eric Garner were killed, there was no comprehensive, national information source. There are now several databases including the FBI’s National Use-of-Force Data Collection (though participation is voluntary, and as of 2019 no data was available for outside review) and several news and non-profit efforts, but the sources are still sketchy and incompatible (Scientific American, 2019). [Many touchpoints; actions effectively private]

• The IRS did over 650,000 audits in 2021 (financeband). We have to rely on its description of what it did and why, since there is no 3rd party source of information on interaction of taxpayers with the government. [Many touchpoints; actions private]

• Most online content sites run on advertising. They are incentivized to maximize engagement which leads to personalized content. Each site visitor sees content unique to them. Even though there’s one algorithm, each user has a different experience. That makes it very hard to observe patterns in the content that is shown, let alone to document bad but relatively rare outcomes. In order to show that TikTok was serving up sex and drug videos to minors, the WSJ had to create dozens of automated accounts (WSJ, Sep 2021). [Many touchpoints; actions private] (Detailed behavior tracking is used to match content to individual users. Sites get paid the most for tailored advertising, which also incentivizes tracking. I’ll leave this aside for now.)

• This is speculative, but it might not be long before the experience of bias on content sites declines—because the AIs will tailor everyone’s content in a way that aligns with their ideology. Nobody will notice bias because it’s only other people who are prejudiced; I am wise and well-informed, and information that matches my views isn’t biased. [Many touchpoints; actions private]

• Amazon has monopoly power over most of its third-party sellers and many of its suppliers. For example, it can see the sales levels of every vendor, and reportedly uses private information to copy successful products. The 2020 House majority staff report on competition in digital markets documents many manipulative practices but had to speak to individual sellers and their trade associations. [Many touchpoints; actions semi-private]

These examples point towards two scenarios where ogregore action is hard to see:

• Large bureaucracies, especially decentralized ones, where it’s difficult to aggregate the experience of individual subjects, even though a single protocols (algorithm) determines the end user experience. 

• Highly personalized, algorithm-driven online content where each touchpoint has a different experience.

(For more on this topic, see my October 2022 blog post “Significant but hard to discern.”)

Coda

Gabor Molnar noted that a providing use cases would strengthen the problem definition and could trigger new ideas about moving forward. That work for the future. The obvious cases are online content sites that manifest themselves differently for every user, and social media platforms.

Online content prompts two images. The first is a modern-day Proteus who doesn’t change shape as he fights the hero, but changes from one to the next. The Proteus of Homer could be forced to speak truth, but only if the hero could keep hold of him; he would keep changing to prevent this. The modern Proteus is different: he takes the one shape that the specific petitioner can’t pin down. (If one can drive engagement, you can drive anti-engagement.) Heroes can’t learn from the stories of previous battles since each fight is a completely new one.

The other image is of a mythic octopus with 10,000 tentacles, each tentacle with 10,000 mouths, inside which is an eye.  (Large online media sites run order 100 million concurrent user sessions.) Each mouth speaks to a different user, but also watches them… something like this: 

The image was inspired by Minoan black painting vases like this one, a Minoan “Marine Style” vase from the New Palace period, in the collection of the Heraklion Museum. (For a modern copy that allows a close-up look, see GreekArtShop.com.)