“We must take caution that we do not find ourselves in a world where our systems do not allow what they cannot imagine—where they, in effect, enforce the limits of their own understanding”.
Brian Christian, The Alignment Problem, 327.
Two years ago, a piece I wrote was featured on the cover of CT. It was about what seemed to me then to be a trend of “religious deconstruction”, which I took to mean “an attempt to reframe or reconsider theological opinions in service of the truth”.
As I wrote here, I think that piece was misguided. I’d like to expand a bit on how I think technological imaginaries have captured the goods of religious deconstruction and turned them into something else.
I see now that religious deconstruction is functioning not as a personal, private reflection on one’s own religious experience. Under such a view, an individual, after a process of questioning, might be led into a different form of belief that was richer, deeper, more communal and more historic. I think now that religious deconstruction is a species of what I have come to call “the technological imaginary”, whereby individuals are prompted and provoked toward forms of belief that are themselves predictions of a machine and products of an opaque algorithm. I fear that “religious deconstruction” is less a personal, private reckoning and more a symptom of a populous that can no longer think for itself but that is achingly and incessantly responsive to the prodding, prompting, and thin psychological rewards that social media delivers. It is not even immanentizing the eschaton that is the risk; it is automating it.
For our purposes, let us consider two goods at hand in any discussion of humans and computing. They are power and values.1
Machines have almost infinite power but no intrinsic values- only the values that they are granted through programming. There is no “instinct”, only functions that are programmed. These functions may over time appear to be “behaviors”, but they are simply the pathways that machines have been programmed to follow.
[This is one of the chief differences between machines and animals. Though animals lack human language, they certainly possess traits that look like “values”. We may return to this at another time.]
Algorithms are models that predict future outcomes based on past ones. They derive their predictions from assigning probabilities to past data; machines can “learn” what a picture is based on the past exposure of machines to similar pictures. In this way machines can be said to both “learn” and to “predict”, though what they are really doing is relying on past exposures to similar data.
The kind of algorithms that social media relies on are basically automated educated guesses. They are predictions that are then encoded and granted power to organize vast sums of data, and they organize this through a form of reduction that privileges the past behavior of individuals in order to predict future behavior. For this reason there is both a targeting and a narrowing that occurs.
The targeting occurs when predictive data is used to locate and identify future information that would be interesting to an individual, based on their past behavior. The narrowing occurs when other information that might be interesting, informative, or challenging is not shown to the individual, in line with their past preferences.
This is where the second good, power, comes in to play.
The sheer power of large language computing models far outpaces the human mind. Such models can consume and analyze data at a rate that humans can simply not match. The most powerful “minds” are now those with no values, or at least no values beyond those that they are programmed with. These are the “minds” that are deployed by advertisers to put before our eyes an endless barrages of the things we already desire. Whether or not there is a “Machine”, there is certainly a man behind it, and he is profiting greatly off of human lust.
When persons engage with these forms of machine learning, and particularly when they engage with them through social media and digital technology, their very personhood is being shaped by a machine with no values and unmatched power.
Insofar as machines work through a process of elimination to refine and predict our decisions, individuals are increasingly having their future engagements predicted due to their time spent with the interface. When utilized by social media companies, algorithms do not merely “predict” future behavior, like I predict the weather based on my past knowledge of early February. They use predictions to shape future behavior, and then reinforce the behavior they have shaped. In this way, the individual is being served by a machine even as she is being made into one. We are becoming simply the result of our past desires, cast into the future, with no way out of circling the algorithmic drain.
Because they are based entirely on past data, machines are thought to lack “human error”- the dangerous traits like sexism and racism that cloud ordinary human judgment.
But in vast and interesting ways, algorithms have themselves been prone to such forms of error. An automated guess is only as good as whatever data the machine has been trained on. When the data is itself limited or skewed, the guesses necessarily will be also.
But such errors, as damaging as they may be, are not the only danger of algorithmic learning.
Machines can only learn what can be taught. The limits of machine “knowledge” are the limits of our own. Though a LLM may surpass the human mind in scale, it cannot do what a mind cannot do. And it is precisely things that the “mind” alone cannot do- or cannot do without a body- that is the domain of Religion. Faith, hope, charity, repentance, confession, absolution- all of these are practices of persons, not Machines. Accordingly, all of these are realities that will never be placed before our eyes by an algorithm. They will not be entertained at all by a mind that is entrapped by digital media.
Algorithms impose their power in places where it does not belong-- increasingly in the realm of human relations- but they also reduce human engagement with the material world to the form of engagement a machine would have.
They take this narrowness and apply it with great power, with the power of giant institutions and corporations behind them.
They do so opaquely and persistently. By shaping human behavior, they reinforce those qualities that can be reinforced by machines.
So what does any of this have to do with “religious deconstruction”?
It is my evaluation that “religious deconstruction”, is increasingly more machine-led and less human-led.
Consider this. The actual goods of reconsidering one’s faith might include humility, submission, charity, community, repentance and confession, as I’ve just named.
Insofar as none of these goods can be had by a machine, nor understood or seen in a data set, I’d argue that none of these goods can be provoked by a machine. So an individual who is engaging in their journey of religious deconstruction primarily led by those who are “online” is having their own preferences, beliefs, and imaginaries shaped by a machine who itself can have no preferences, beliefs, or imaginaries. In this way, as the epigraph notes, we are dangerously close to a place where our systems “enforce the limit of their own understanding.” We are dangerously close to a place where Machines are dictating what the Church can be.
Insofar as the current religious landscape is almost irretrievably mapped by the forms of technology that rely on opaque and reductive algorithms, it is rocky soil for a true engagement with a historic faith. Much of what goes under the label of “religious deconstruction” has been predicted and shaped by a machine. By yielding to it, we are becoming less human by the hour. We are becoming more like machines, those entities who have power but no values and whose desires and behaviors are shaped not by something outside themselves, but by a vast, corporate controlled landscape of desire.
My intervention was addressed to those who were wrestling with deep questions. These questions were instigated by personal loss but also by tragic leadership failures that sometimes made the church feel like a liar and a thief. By expanding the view of what the church was and is for, I hoped I might provide a net of sorts to catch those who might otherwise fall out of the Christian community entirely.
The danger at hand is different than the one I feared. It seemed to me that religious deconstruction held two paths- stay religious or leave the faith. I did not anticipate a third option, which was to radically revise the faith and make it into something much more bespoke that would reflect the algorithmic narrowing of our desires.
Doctrinal revisions are of course nothing new, and are not primarily what concerns me here. Rather, the revisions I observe concern what the church itself is.
Religious deconstruction might seem at first to be an expanding. But it is actually a narrowing. When our religious practices derive from our engagement with algorithmically mediated technologies, our vision for what religion is is characterized by a self-focused narrowing. We come to see more of what we desire, and so desire more of what we see. What a religion is, or what a practice of personal faith might be, is slotted into the very same holes that might be assigned to interior decorating or weeknight recipes. If you like fast family suppers, you will see more of them. If you love moody bookcases, there are thousands to feast your eyes on. And if you feel embarrassed or frustrated or isolated from the faith of your youth, then there are dozens, hundreds, even thousands of individuals who will help you curate that animus. The algorithm will keep you right where you are.
There is a promise buried in the Christian faith, right in your baptism. If you were dipped in water by a pastor and your well-meaning parents, you were baptized into the covenant between God and the world. This covenant was mediated between God and Abraham and then revealed in Christ, who in water and blood spoke God’s own word to the world- that God is making all things new. In baptism, in the splashing of water on skin, the most basic of theological equations is revealed. In Christ God makes matter matter. He moves on the earth and sets it aflame and invites us in our tunics of skin to be made God’s own, now and into eternity, where our fading life will once again be remade. He calls us out of ourselves, out of our incessant, self-absorbed narrowing, to expand our visions. Instead of being fed over and over again with what we already desire, we are invited into a new life that we could not imagine for ourselves. No one, seeking to have their ideas reinforced, would imagine forgiveness, or chastity, or confession. But these are the very goods that the Church has on offer- that you might look up, and live.
Some will of course fall away and renounce their baptism. But many more, it seems, are undertaking a radical revision of religion itself. Instead of staying or leaving, they are remaking the Church in the image of the Machine, revising it so that it fits their own narrowed desires, values, and demands. When religious deconstruction takes such a form, it should not be coddled or encouraged. Let the dead bury their own dead. There is only one way that leads to life, and it is narrow, yes, but that narrowing leads us beyond ourselves, to the only One who is not made by human hands.
See Brian Christian, The Alignment Problem, which is my favorite treatment of this question. The frame of “power and values” comes from Christian, though I’ve put it in another key.
Interesting point - the alignment problem mention caught my attention at the beginning.
I did want to add a few nuances/caveats to some technical descriptions in your piece, and a comment about the ecclesial and theopraxy dimension:
- An algorithm is not necessarily a "model that predicts future outcomes based on past ones".
At its base, an algorithm is a set of steps or process based in mathematical rules, that achieves a specific goal, solves a problem, or produces an output from a given set of inputs. This can be as simple as a recipe or a basic algebra function, or complex as in neural reinforcement learning or the concept of "learning" from data occurrences to estimate future "events".
- Advertising optimizations do not necessarily need large language computing models; statistical machine learning in advertising has been used for years before the recent advances and market debuts that have pushed LLMs into mainstream use.
Regarding deconstruction and the church, I think you're getting at important realities of mimesis, social constructions of reality, and bourgeoisie discourses being mediated and reinforced by social dynamics and network effects vs. being completely organic. However, I don't think it's best to understand such a phenomenon by merely reducing it to one specific explanation or facet.
I'm concerned by any conclusion, including yours, that generalizes the problem with an explanation that places the onus on individuals in an individualistic ( one could invoke the "buffered self") society, because it seems like an easy out for churches and ministries to blame broader cultural and society "trends" vs. looking retrospectively at how ministry practices don't work, fail to form people or in some cases even simply get them on the path to a real understanding of Jesus, and lack a firm charism that should organically develop people into a Christ follower.
One of the issues that I observe particularly in a lot of reformed, "cultural engagement" focused, Tim Keller-adjacent churches and ministry leaders, as well as, frankly, perhaps more academically inclined "pastor theologians", and many swathes of churches such as the ACNA, is they seem to think that it's important to analyze, understand and respond (or react) to every such trend such as this, vs. spending the energy and organizational leadership (not merely "management") to simply build the foundations of a robust charism, spiritual practices, and the church as a means of grace.
I don't actually think it's super helpful to constantly analyze and try to understand or conclusively attribute things such as deconstruction; there will continue to be other phenomenon and similar "syncretist" or de-centering movements like this, but it's better for church and ministry leaders to build a robust foundation and focus on core of the faith + nurturing a robust "charism", and then let the chips fall where they may. There will likely continue to be a shuffling and people will "leave" the faith and/or involvement in a church (whether permanently or for a season), but ministry leaders shouldn't exert all their energy reacting to or trying to analyze these moves and trends, as if you have to completely or properly understand them in order to have a response. There is already a robust foundation found in scripture, the tradition of multiple different church bodies and expressions, and the movement of the Holy Spirit in the church and the world at large.
So articulately reasoned and argued. Much food for thought. That eschewing this norm of pursuing “wisdom/life” in online “communities/spaces” would necessitate actual human community and shared pursuit of wisdom in the real world is reason enough to pull the plug on the machine that is arguably the biggest contributing factor to the explosion of mental health issues at present. Can we do it? Will our children? There is much to think about here, and your view of what faith could/should be is a hopeful anecdote to the, may I say, dirty word, “religion.” Thank you.