Spurn artificial ideology

“Apocalypse now” identified the corrosive influence of new viral ideologies, created unintentionally by recommender systems, as a major AI risk. These may cause social collapse if not tackled head-on. You can resist.

Political actors of all sorts now recognize the king-maker power of recommender AI, and also its propensity for creating and empowering new, hostile ideologies. Influencing or outright controlling recommenders may now be the most important form of political power.

Governments, political parties, corporations, lobbyists, NGOs, and concerned citizens all pressure recommenders’ operators to censor messages from opponents, and to promote messages from themselves and their allies. That has been quite effective at forcing them to make changes. Mooglebook doesn’t care about the content of messages, so long as they cause ad clicks.

Censorship has, so far, not been very effective for suppressing ideological enemies. A recent mainstream survey found that 18% of Americans believe all three of QAnon’s central claims,¹ and only 38% reject all three entirely.²

In response, political actors have demanded harsher censorship and greater influence for themselves. As I’ve emphasized, what’s risky about AI is its ability to create pools of power faster than they can be checked by opposing powers. Whoever gains control over recommenders may achieve hegemony and institute a durable oppressive regime. I do not trust anyone with that power.

Social theorists and technologists have discussed many possible interventions into AI systems to curb their creation and direction of political power. Some combination of them may be effective, at least partly. I won’t discuss them here. I’ll make two more radical suggestions instead. First, discussed here, we can strengthen individuals, institutions, societies, and cultures against ideological takeovers. Second, we can outright shut down AI, instead of chipping away at edges of its harmful effects. The two following sections of this chapter are about that.

Ultimately, the problem is ideology as such. All ideologies are mistaken about many things, so each has to use underhanded tricks to blind you to realities that conflict with its narrative. However, with experience we’ve figured out how to keep the traditional ones mostly in check. Most of the world has driven the nastier ones (absolutist religions and twentieth century totalitarianisms) to extinction.³ We’ve learned to live with the relatively benign ones, in part by stalemating them against each other. Almost all countries now employ mixtures of the better features of socialism and capitalism. That works reasonably well, confining conflict to comparatively minor aspects.

We don’t yet have good defenses against the new ones. As of early 2023 QAnon looks like it has a meaningful chance of taking control of the Republican party. The next (still unlikely-seeming) step would be control of the United States government. No one knows what to do about that; the threat is both absurd and realistic. “Surely,” you’d think, “this mythology is so thin, incoherent, and transparently false that everyone will snap out of it.” Yet it continues to advance despite all the prophesies of its messiah “Q” having failed, and despite extensive internet censorship.

Anyway, deflating QAnon might miss the forest for the trees. Plausibly, AI can spawn and empower new ideologies faster than we can neuter them.

What you can do

Everyone can examine their own relationship with ideology. Do you have an ideology? Or does an ideology have you. Do you “belong to” a political party or religious group? Maybe, if so, you are a slave, as the term suggests.

Ideology’s trick is to convince you that it is True and Good, in absolute senses. Therefore, any evidence that contradicts the narrative can and must be ignored or explained away. Also therefore, you are Good to the extent that you propagate the ideology (and Bad to the extent that you question it).

Mostly, nobody wants to get free from the ideology that owns them, because it constantly assures you that you are Good, so long as you keep retweeting its messages. Everyone wants that cheap method for feeling morally adequate, and wants to avoid the painful shame of violating the sacred dictates.

That doesn’t sound like you, of course. Your political opinions are just common sense, well thought out, and obviously true. You are entirely rational about specific issues. Not at all like the Bad Tribe, who are mysteriously possessed by an evil brain-eating quasi-religious fanaticism. If you sometimes make strong statements online, it’s because their stuff must be stopped at all costs.

I have mostly avoided naming examples of the new ideologies, because I expect a substantial fraction of readers are possessed by one or another of them. I’ve used QAnon as the example because I’m guessing few readers of this book are owned by it. I suggest taking seriously the possibility that whatever ideology you are subject to appears similarly irrational and dangerous to anyone who isn’t. I suggest considering what that may imply about that ideology, and about you—not about its opponents.

The cultural assumption that everyone must “have” an ideology gives them much of their power. If you have any doubts about your ideology, you can reassure yourself that the alternatives are all vastly worse, so switching is out of the question. Or, if your faith deteriorates sufficiently, you may switch after all—and become a slave to Tweedledum rather than Tweedledee.

You can free yourself entirely, if you realize you want to. I explain how, step by step, in “Vaster than ideology.”⁴ You can turn the tables, and own ideologies instead. Many are useful for particular purposes in particular circumstances. You can recognize whatever actually is true and good in any of them. Standing outside and above all ideologies, you can even see why so many people adopt QAnon, and sympathize. The current American ruling elites, both left and right, are pervasively mendacious, corrupt, self-serving, incompetent, and entrenched. They do need replacement.

Institutions can prepare for memetic attack, constructing defenses in advance.

Functional institutions operate systematically, so their actions are guided by explicit principles.⁵ Under ideological attack, institutional authorities may abandon principles in order to please groups who have, or claim to have, extraordinarily strong, unprincipled feelings. For example, it’s too common, when an employee gets accused by a social media mob of violating some community’s sacred beliefs, for executives to bow to pressure and fire the person within hours. Dismissal might eventually turn out to be entirely justified; but immediately throwing a potentially innocent member to the wolves just to avoid bad PR is unjust and cowardly.

Leaders can stiffen their spines by putting explicit policies in place ahead of time. For example, company policy could be that, in such a situation, the employee can at worst be suspended for a month pending investigation and judgement according to a defined process, and then fired if appropriate. In response to ideological condemnation, executives can say “we sympathize deeply with your howls of pain, but unfortunately our hands are tied to the mast for the next thirty days.” Probably by then everyone will have forgotten about the whole thing anyway.

Some more examples:

• Government science agencies can adopt a policy of never crafting recommendations on the basis of how they might be ideologically misinterpreted, deliberately twisted, or opposed by culture warriors, and committing instead to always relaying the best current evidence.

• Universities can adopt a policy of strict ideological neutrality, and not allow members to hijack their institutional power to suppress messages they don’t like. More than a hundred have already adopted the “Chicago Principles,” along those lines.

• Companies can also adopt policies of ideological neutrality, and can prohibit employees from engaging in political activism irrelevant to the company’s mission while at work.⁶

Such measures have short-term PR costs, but in the longer term may be the only way to preserve a functional institution. The credibility of public health agencies has been severely, perhaps permanently, damaged by their PR-motivated lies during the covid crisis.

Fear of short-term costs deter action; few institutions have prepared adequately. If principled, defensive measures became common, artificial ideologies might be revealed as loudmouth weaklings, incapable of doing significant damage to anyone willing to stand up to bullying. On the other hand, the manifest deficiencies of the traditional ideologies may empower whatever opposition AI conjures up.

Research may help. Social scientists can investigate the dynamics of memetic conflict, the reasons people adopt artificial ideologies, and interventions that may loosen their grip at the individual, institutional, and society-wide levels. Such research risks further developing further offensive capabilities. However, currently there is far greater incentive for offensive weapons development (millions of people constantly crafting new tactics under the direction of AI-generated ideologies) than for defense against them, so the balance of power is on the side of attackers. Neutral understanding seems more likely to reinforce defensive capability than to further aggravate the problem.

Funders, such as NGOs dedicated to democracy promotion and to strengthening social capital, can support such work.

The longer-run solution is broad cultural understanding that being a slave to any ideology is bad for you and everyone else. This observation is startlingly rare; nearly everyone assumes that you must belong to one. However, it’s obvious once pointed out. Most people are also tired of ideological conflict, and understand that it’s destructive; yet persist because the alternative seems to be surrendering to the Bad Tribe. I am optimistic that anti-ideological understanding could spread rapidly once it gets going.