The algorithmic god. How recommendation systems became our post-religion
Dear Reader,
I want to write about a specific claim that I think is more accurate than most people currently accept. The claim is that recommendation algorithms have taken over some of the specific social functions that religion used to perform, in ways that were not planned by anyone but that have specific consequences we are only beginning to notice.
I do not mean that algorithms are literally worshipped. I mean something more precise. Religion, historically, has performed several social functions beyond providing metaphysical narratives. It has told people what to pay attention to. It has told them what to value. It has shaped what they aspired to. It has structured their communities. It has given them frameworks for understanding themselves and their place in the world. In secular societies, these functions were supposed to be handled by other institutions. Schools. Journalism. Community organizations. Political parties. Cultural producers.
What has actually happened in the last two decades is that these secular institutions have weakened, and their functions have migrated to a different kind of authority. Recommendation systems. Feeds. Algorithms. Not everyone. But a substantial and growing portion of what used to be handled by religious and secular institutional life is now handled by systems that were not designed to handle it, that have no accountability structure, and that operate through mechanisms most users cannot see.
This is not a call for a return to religion. It is a specific observation about what has happened to the specific functions religion used to perform. And it is an argument that we should look at these functions carefully, because the systems now performing them are shaping us in ways we did not consent to.
Start with what religion actually did as a social technology, apart from its metaphysical content.
Religion told people what to pay attention to. This is easy to miss because we now think of attention as a private resource, but for most of human history attention was substantially shaped by shared religious frameworks. Prayer times structured the day. Holy books provided reading lists. Sermons directed collective focus. Rituals concentrated attention on specific themes at specific times. The person immersed in a religious community had many of their attention decisions made for them by the community.
Religion told people what to value. Not just in abstract moral terms, but in concrete detail. What kinds of people were worth emulating. What kinds of accomplishments mattered. What kinds of failures were shameful. What kinds of beauty were legitimate. These were not neutral preferences. They were transmitted through the religious framework and reinforced by community practice.
Religion shaped what people aspired to. Sainthood. Salvation. Enlightenment. Righteousness. These abstract goals gave people specific directions to move in, and the religious framework gave them specific practices for moving there. Whether or not the metaphysical claims were true, the aspirations organized life around specific goals that were shared with the community.
Religion structured communities. This is the most visible function. Religious communities provided regular gatherings, shared rituals, communal celebrations, and networks of mutual support. Most human societies, for most of history, were organized around religious communities. Even when people did not fully believe, the communities provided structure.
Religion gave people frameworks for understanding themselves. Who they were. What their life meant. What they were supposed to do. These were not private psychological questions to be worked out individually. They were shared questions with communally-provided answers, transmitted through the religious framework.
All five of these functions were substantially social technologies, whatever else religion also was. And when religion lost its grip on modern life, the question was what would take over these functions.
The traditional answer, from the Enlightenment onward, was that secular institutions would take over. Schools would tell people what to pay attention to. Journalism would shape what they valued. Political parties would give them frameworks for understanding themselves in relation to society. Cultural institutions would provide aspiration and meaning. Community organizations would structure their social lives.
For a while, this worked. The middle of the twentieth century, in the West, was a period when secular institutions were strong and religion was weakening, and the social functions previously handled by religion were largely handled by other institutions. Not perfectly, but substantially. Most people had some combination of school, work, political affiliation, cultural participation, and community involvement that structured their attention, values, aspirations, and self-understanding.
What happened next was that these secular institutions themselves weakened. Trust in traditional media declined. Political parties became more divisive and less trusted. Community organizations, particularly local ones, lost members. Schools became more contested. Cultural production fragmented. The specific institutional structure that was supposed to handle the social functions of religion after religion weakened turned out to be more fragile than expected.
The specific vacuum this created was substantial. People still had the underlying needs. They still needed something to tell them what to pay attention to, what to value, what to aspire to, how to structure their community, and how to understand themselves. And the institutions that were supposed to do this were failing.
Which is when the algorithms arrived.
Recommendation systems entered into this vacuum without being designed for it. Their designers thought they were building tools for finding relevant content. They were, in a technical sense. But because they were solving a technical problem in an environment where the underlying institutional structure was already weak, they ended up performing functions that no one had assigned them.
Let me walk through what recommendation algorithms actually do now, mapped onto the functions religion used to perform.
They tell you what to pay attention to. Your feed is the answer to the question "what should I look at right now?" It is provided by an algorithm that has decided, based on data about you and about the world, what will hold your attention most effectively. This is a decision about attention allocation that used to be made by you, or by institutions you trusted, or by religious authority. Now it is made by an algorithm that you did not choose and that you cannot see the reasoning of.
They tell you what to value. Not directly. But the specific content that shows up in your feed shapes what you consider important, admirable, worth aspiring to. If your feed is full of people succeeding in specific ways, those become the ways you consider success. If your feed is full of people criticizing specific things, those become the things you consider worth criticizing. The algorithm does not have opinions in the human sense, but the specific weights it applies to what to show you have the effect of transmitting a specific set of values.
They shape what you aspire to. Related but distinct. What the algorithm shows you shapes the specific futures you can imagine for yourself. The person who is fed hours of content about entrepreneurship starts to see entrepreneurship as a natural aspiration. The person fed content about fitness starts to aspire to physical transformation. The person fed content about political activism starts to aspire to a specific kind of political engagement. The aspirations you develop are downstream of what the algorithm decides to show you.
They structure community. This is where the effect is most obvious. Your feed determines who you interact with, who you consider your peers, whose opinions matter to you. This has replaced, for many people, the specific structuring role that religious communities used to play. Your online network is your community. And the algorithm shapes it by deciding whose content you see and whose you do not.
They give you frameworks for understanding yourself. This one is the deepest. What the algorithm shows you shapes how you think about who you are, what you are doing, what your life means. If the algorithm surfaces content that treats certain identities or situations as important, those become the categories you understand yourself through. If the algorithm treats certain concerns as marginal, those become concerns you do not consider central to your life. The specific self-conception you have is substantially shaped by the specific stream of content you have been consuming.
All five functions. Not entirely, and not for everyone. But substantially, and for a growing portion of the population. The recommendation algorithm is now doing what religion used to do, in these specific functional respects.
The specific reason to call this a religion, or a post-religion, is not metaphysical. It is functional. Whatever performs the specific social functions religion used to perform is, in a functional sense, doing religion's work. And what performs those functions now is the algorithm.
There is a specific quality to how people relate to algorithms that reinforces this. Most people do not know how the algorithm works. Most people trust that the algorithm knows what it is doing. Most people have the specific experience of feeling that the algorithm has understood them in some way, that it knows what they want, that it is offering them what is right for them. This is an experience of relating to an authority whose reasoning is not fully accessible, whose judgments are treated as broadly reliable, and whose specific verdicts about your attention are accepted with only modest resistance.
This is, structurally, the specific relationship people used to have with religious authority. Not literally God. Not literally worship. But the specific texture of trusting a system whose inner workings you cannot see, whose judgments you nonetheless treat as authoritative, and whose recommendations shape the specific quality of your daily life.
I want to be careful not to overstate this. People also complain about algorithms. They notice when the algorithm gets things wrong. They occasionally do things to try to influence it. The relationship is not identical to the historical relationship with religious authority. But it is close enough, in the specific functional sense that matters, that the comparison is worth taking seriously.
Let me tell you what I have noticed about my own relationship to the algorithms I use.
I use LinkedIn for work. I have watched, over the last several years, how the LinkedIn algorithm has shaped my sense of what I should be doing professionally. Not because I take it seriously as advice. Because it is constantly showing me examples of what other people are doing, and those examples become the reference points for what I consider possible.
If the algorithm consistently shows me successful founders, I start to think of founder success as the natural benchmark. If it shows me thoughtful long-form content, I start to think of long-form thoughtfulness as the norm for professional discourse. If it shows me hustle-culture content, I start to feel behind if I am not hustling. The specific content shapes my baseline expectations, whether or not I agree with the specific examples.
This is a small effect. It matters at the margin. And multiplied over years of use, it adds up. My specific professional aspirations are not identical to what they were before I used LinkedIn. Some of the differences are downstream of things I chose deliberately. Some are downstream of the specific content the algorithm decided to show me. I cannot fully separate the two.
The unsettling thing is that this is happening to everyone who uses these platforms, and it is happening at scale, and the specific direction it is happening in is not chosen by anyone. It is the emergent output of an optimization process aimed at engagement. The engagement is what the platform cares about. The specific effect on your aspirations is a byproduct that the platform is indifferent to but that you are living with.
There is a specific consequence of this substitution that I want to name, because I think it is the most important one.
Religion, whatever else it did, was accountable in a specific way. Religious communities had visible leaders. Religious doctrines had visible texts. Religious practices had visible traditions. If you did not like what your religious community was doing, you could see clearly what it was doing, and you could argue about it, and you could leave, or reform it, or start a new one. The authority was legible, contested, and negotiable.
Algorithms are not accountable in any comparable way. Their decisions are not visible. Their reasoning is not accessible. Their leaders, insofar as they exist, are the engineers who tune the systems, and those engineers are not identified, not selected by users, and not answerable to them. The specific authority that shapes what you attend to, what you value, what you aspire to, and how you understand yourself, is opaque and unaccountable in ways that religious authority, for all its problems, was not.
This matters practically. When religions have caused harm, there have been mechanisms for confronting them. Reform movements. Theological debate. Institutional pressure. Political action. When algorithms cause harm, the mechanisms are much weaker. Users can complain, but the complaints go into an interface that is essentially non-responsive. Regulators can intervene, but the specific technical decisions are made by engineers who move faster than regulation can keep up. The specific structure of accountability that religion developed over millennia has not been developed for algorithms, and there is no clear path to developing it.
Which means that if algorithms are indeed doing religion's work in the functional sense, they are doing it without the specific structures of accountability that made religious authority survivable. This is a genuinely new problem, and one that our institutions are not yet equipped to handle.
I want to close with a specific reframing.
The specific question worth asking, if you want to think clearly about your own relationship to algorithms, is what kind of authority you want to have shaping your attention, your values, your aspirations, and your self-understanding. If the honest answer is that you would prefer these to be shaped by human institutions you can see and criticize, then the current situation is bad for you. If the honest answer is that you are fine with these being shaped by opaque optimization systems whose interests are not yours, then the current situation is fine for you.
Most people, if they think about it clearly, come out on the first side. Which means that most people are living under a specific kind of authority that they would not choose if they were choosing directly. This is not a small thing. It is the specific situation of being religious about something you do not endorse.
The question of what to do about this is hard. Individual practice helps but does not solve. Structural change is slow and contested. What is available in the near term is at least accurate perception. You can see the situation clearly. You can name what is happening. You can hold your own relationship to the algorithm at some distance. This is not enough. It is more than nothing.
The specific quality of freedom that the Enlightenment thought it was building requires the ability to think outside the frameworks that shape you. That freedom is being eroded, in ways that most people cannot see, by systems that no one specifically designed to erode it. Recognizing this is the first step to preserving whatever remains.
Next month I want to write about AI manipulation more specifically. Because the algorithmic god I have been describing is one form of manipulation, working through recommendation. There are other forms working through more direct means, and understanding them is important for anyone who wants to see clearly what the current AI systems are actually doing. Stay with me.
— Transmission Sent —
Niklas Hanitsch
Reference materials
- Charles Taylor — A Secular Age (2007)
- Yuval Noah Harari — Homo Deus: A Brief History of Tomorrow (2015)
- Byung-Chul Han — The Burnout Society (2015)
- Frank Pasquale — The Black Box Society (2015)
- Neil Postman — Amusing Ourselves to Death (1985)
- L.M. Sacasas — The Convivial Society newsletter
- Robert Bellah — Habits of the Heart (1985)
- https://www.newyorker.com/culture/cultural-comment/what-does-the-algorithm-want
- https://en.wikipedia.org/wiki/Religion_and_technology
Continue reading
- Dopamine loops: How social media hacked human motivation
- The attention economy is eating consciousness
- Surveillance capitalism in 2026: The concrete infrastructure
- Where does the sacred live in the digital age
- Ritual is not superstition. It is technology for the soul
Frequently asked questions
Are algorithms actually a religion? Not literally, but in a specific functional sense they perform some of the same social work religion used to perform. They shape what people attend to, what they value, what they aspire to, how their communities are structured, and how they understand themselves. Whether this makes them religion depends on how you define religion, but the functional overlap is substantial.
Who controls what the algorithm shows me? Nobody in particular. Recommendation algorithms are optimization systems tuned to maximize engagement or similar metrics. The specific decisions about what you see are made by the algorithm, based on data about you and about content. Engineers at the platform tune the algorithm's parameters but do not typically decide specific outputs. The result is that there is no clear person or institution to hold accountable when the algorithm produces bad outcomes.
Why is algorithmic authority worse than religious authority? Religious authority, whatever else you might say about it, was visible and contestable. You could see who the leaders were, argue with the doctrines, and leave or reform the institution. Algorithmic authority is opaque. You cannot see who made the decisions, cannot argue with them, and cannot easily leave without giving up the platform entirely. The specific structures of accountability that religions developed have no parallel in the algorithmic world.
Can I escape the algorithm? Partially. You can use fewer algorithmic platforms. You can choose non-algorithmic alternatives when they exist. You can build habits of consumption that do not go through algorithmic feeds. What you cannot do is fully escape while remaining in normal digital life. Escaping is a partial and effortful practice, not a clean choice.
What is the political solution? Regulation of recommendation systems. Requirements for algorithmic transparency. Support for non-algorithmic alternatives. Public investment in the institutional structures that used to perform the social functions algorithms now perform. None of these are easy or quick. All of them are being discussed but have not been substantially implemented.