Institutions for angels

2.1 Adverse selection

If a movement has no gatekeepers and an inclusive ideology, the recruits we get are not necessarily the recruits we want. Even with good intentions we should be wary of assuming that the people who turn up are like us. They turn up because the movement attracts them, and the things that attract people are not the same as the things that make a movement work.

Some are attracted by the movement’s substance — these are the recruits we hoped for. Some are attracted by the opportunity to extract status, attention, or resources from a low-friction social environment — these are the recruits the movement’s existence will inevitably produce. The ratio between the two depends on what the movement looks like from the outside, and on what it offers the people who join it.

Worked example: A friend of mine has been involved in a squatting movement, occupying empty houses that have been abandoned because climate flood risk makes them uninsurable. The movement is open to anyone who wants to join, and it has been successful in providing shelter to people who would otherwise be homeless, many of whom are pro-social community-builders. Open membership is the whole pitch — the people they are trying to help are exactly the people other organisations exclude. But radical openness to the excluded means many people join the squat who imperil the overall mission: aggressive sovereign citizens, sexual predators expelled from more exclusive projects, people with untreated substance dependency who steal group assets.

There is no general clean resolution here. The moment we erect a barrier to entry, we risk reproducing the problem of exclusion and judgement that we were trying to solve in the first place. The trade-off is structural, and what we can do is acknowledge it explicitly, and budget for the costs of openness according to how much of it we desire.

2.2 Rage cascades

Invasive arguments are effective at recruiting people to a movement that is dominated by growth and recruiting. Even if I do not want a rage-based recruiting strategy as a movement founder, those arguments may come to dominate if incidental rage-based recruiting outpaces other means. Possibly worse: rage-based recruiting tends to drive out the non-rage recruits — peaceable sorts may find it unpleasant to spend their free weekends with militant hardliners. This might be one of the factors leading to the tendency of revolutions to devour their children. (Related: purity cascades.)

If I have to recruit on the internet I need prominence, and the cheapest way to acquire prominence is to have a movement whose most visible proponents specialise in public denunciations and outrage. Even if I do not want to recruit that way, my movement may gain members that way unless I can tightly control how its members recruit. Even if I onboard members with strong social norms against rage-based recruiting, my opponents can rage-recruit for me — there is some evidence that this has been attempted at scale, even if reliable quantification of how effective it has been is harder to come by.

What this implies about social movements generally: constant weed-like growth in the most toxic wing of any movement that exists for collective struggle against some Bad Thing.

The internet is disproportionately likely to recruit the least useful and even counterproductive membership for various reasons. On that intensely contested ground, virtue signalling is selected for at the expense of effortful activism. Clicktivism is so cheap that it might reproduce much faster than face-to-face organising. Within the category of virtue signalling, the types of signal that get most traction are likely to be the most infuriating ones — shaming, policing without consent, trolling — alienating the opposition or creating new ones.

Accusations of virtue signalling have acquired a conservative valence on the internet, but the dynamic is not restricted to either conservative or progressive politics, AFAICT.

2.3 Entryism

A second-order problem: not every wrong recruit is wrong by accident. Some turn up on purpose, with a programme, knowing exactly what they are doing.

Entryism is the deliberate infiltration of an open movement by people who hold its trappings while subverting its values. The classic Trotskyist tactic — the Militant Tendency working inside UK Labour through the 1980s, openly described in their own internal documents — is the textbook case. The pattern recurs across the political spectrum. Alt-right operatives joining mainstream conservative organisations to push them rightward. Anti-vaccine activists joining wellness communities. Pro-natalist movements joining secular feminism. Various ideological groups joining Effective Altruism to redirect its donor base. The diagnostic is that the entryist arrives with prior commitments and an agenda for the host, not as a recruit to the host.

What makes this work is once again openness. A movement that cannot exclude anyone for any reason is one that a sufficiently committed group of outsiders can potentially take over. The defence has to come from somewhere other than gatekeeping at the door, since the door is the whole point. Strong inner circles with their own membership criteria might help (see movement design § hygiene and inner circles). So might keeping high-status roles rotating, so that capturing any one role does not amount to capturing the movement itself. So does transparency about ends — entryists prefer not to be seen, and a movement that periodically articulates what it is for makes their cover more expensive to maintain.

This is, I think, distinct from the Geeks/MOPs/Sociopaths model, although they are cousins. The sociopath in that ecology is opportunistic — they show up once a movement has gathered status and try to extract some of it. The entryist is more programmatic — they show up while the movement is still small enough to control and valuable enough to want to control. They bring a specific intent to redirect the movement’s trajectory from within, often as part of a larger strategy.

One observed entryist tactic is intra-movement purity enforcement: by making the host’s existing leadership untenable on the host’s own terms, the entryist faction creates room for itself. Whether this generalises into a self-reinforcing rage cascade or stays a localised tactic, I do not know.

The diagnostic above does not, by itself, distinguish predatory entryism from legitimate movement-broadening. Suffragists joining temperance societies, civil rights organisers working through mainline churches, women breaking into male-dominated unions, environmental advocates joining utility regulators, LGBTQ+ politics moving through mainstream legal organisations — each arrived with prior commitments and an agenda for the host. Each looked predatory at the time. Each reads as legitimate broadening in hindsight. Which it was is often only legible decades later, and only to historians.

The predatory framing stays as the default here because this notebook is about failure modes. The complementary view — that the same affordances are how movements legitimately reform from within — sits over in pluralism and agonistic conflict, and in the civil society and social capital traditions of Tocqueville and Putnam (and in deliberative-democratic theory more broadly, which I do not yet have a notebook for). One person’s entryism is another’s coalition expansion.

2.4 Controlled opposition

A third-order problem. Controlled opposition — the idea that a movement is, secretly, a sham, run by the people it claims to oppose, designed to absorb dissent into a harmless channel — is an overused concept on the internet. Most accusations of controlled opposition are wrong. The accusation pattern is itself a known failure mode of the conspiracy radicalisation pathway: once a person decides their movement is being betrayed from within, they have a frame in which any disagreement, any compromise, any failure to escalate is evidence of treason. Wherever I see the term applied to a contemporary movement, I assume the user has joined a cult of suspicion until shown otherwise.

That said, the thing the term refers to does happen, just less often than youtube cookers think. State-run cases — Cointelpro plants in 1960s civil-rights and antiwar groups; yellow unions designed by employers to forestall actual unionisation — are well-documented, but they were expensive operations conducted by intelligence services and capital with significant resources behind them.

More common might be the emergent version: a movement drifts into a safety-valve role without anyone planning it. The professional class of activists, NGOs, paid spokespeople, and mainstream-media-friendly faces ends up performing dissent in a way that does not threaten anyone, and the movement never recovers a wing capable of doing so. No one decides this; the equilibrium just settles there, with each individual locally optimising for survival within the institutions that fund them. (See Schelling-Goodhart for one model of how this kind of equilibrium gets locked in.)

Either version is recognisable as a move in a coalition game: a player whose payoff depends on the opposition being incoherent has every incentive to keep it incoherent, deliberately or by passively benefiting from arrangements that produce incoherence. The action question — how to manipulate the partition’s value function so a preferred partition is stable — has no clean general theory yet, but its empirical instances are everywhere.

I expect this dynamic could get more common in the AI era, in two distinct ways. Intentionally: coherent personas with their own posting history and apparent communities are about to become something a junior comms staffer can spin up over an afternoon. The cost of fielding a credible counter-movement that does not, in fact, oppose you approaches zero.

Incidentally: attention-economy platforms already select for performative-contrarian voices that occupy the opposition slot without delivering opposition’s substance — voices that are loud, photogenic, easy to argue with, and bad at coalition-building. LLM-generated content might amplify that selection pressure. A future where most of the visible opposition to anything is, in effect, controlled — controlled by a recommender system rather than a controller — seems structurally cheap to arrive at, even if no actor wants it.

2.5 Elite capture

Olúfẹ́mi O. Táíwò, On the Uses and Abuses of Identity Politics:

Elite capture, he explains, is a concept that emerged from the study of developing countries. It initially referred to the tendency of the upper class to gain control over foreign aid; in other words, the rich get richer. But the concept has also come to encompass the ways that elites appropriate political projects and monopolize attention.

Elite capture, Táíwò says, is “not a conspiracy” but rather “a kind of system behaviour.” Systems are a major theme of the book, a theme Táíwò develops by drawing on the philosophy of games. Another motif is his impatience with the symbolic gestures and efforts to avoid “complicity” that have come to take precedence, in his view, over actual political outcomes.

See also his Boston Review piece: Identity Politics and Elite Capture.

2.6 Purity cascades

c.f. Schelling-Goodhart.

Which norms are stable? Which are at the start of slippery slopes? I used to think no incremental change to group norms could lead to runaway extremism (too many arguments about slippery slopes have turned out to be bullshit), but I was demonstrably overconfident about that, because it does happen sometimes.

The question is what the exact dynamics are that encourage and discourage groups tilting into increasingly radical positions. I have witnessed too many people I know tumbling down the conspiracy radicalisation pathway and seen social norms shift in ways that ultimately seem to undermine the project. Sometimes the outcomes are destructive; sometimes just a waste of time, e.g. euphemism treadmills.

There are famous examples of escalating purity tests in movements throughout history. People like citing the French Reign of Terror; things can go off the rails.

I still think this example can take us too far, though — some weird stuff happened in the peak-woke era, but no one did actual mass-guillotining over microaggressions. Careers were destroyed though. Interesting case study: The OwnVoices Movement and the policing of who is allowed to present which identities in their characters. Who gets to define authenticity and appropriation?

Regina Rini relates the problems of calibration to control systems‚ to air conditioner settings, which is a helpful metaphor.

2.7 “Structureless” implicit hierarchy

A classic early essay in this area is Jo Freeman, The Tyranny of Structurelessness:

… to strive for a structureless group is as useful, and as deceptive, as to aim at an “objective” news story, “value-free” social science, or a “free” economy. A “laissez-faire” group is about as realistic as a “laissez-faire” society; the idea becomes a smokescreen for the strong or the lucky to establish unquestioned hegemony over others. This hegemony can be so easily established because the idea of “structurelessness” does not prevent the formation of informal structures, only formal ones.

I would refine this critique a little in light of the last few decades of research, but anyone who has tried to organise to make the world better will recognise this feeling.

Nöllke (2024) is a guidebook for playing the game of climbing invisible hierarchies.