God and the Correlation Trap
You pray before an exam. You pass. You thank God.
This is one of the oldest and most human things there is. But there is a quieter question underneath it, one that doesn’t require you to be an atheist to ask: did the prayer cause the outcome or did it just precede it?
I’m not trying to argue God doesn’t exist1. That’s a different conversation and honestly not one I’m qualified to have. What I want to argue is narrower: that the everyday evidence people cite for God’s intervention, answered prayers, miraculous timing, inexplicable recoveries, is almost entirely correlational. And that we, as pattern-seeking creatures, are spectacularly bad at noticing the difference.
The Setup
In statistics, correlation means two things move together. Causation means one thing produces the other. The gap between them is where most of our intuitions go wrong. The classic example: ice cream sales and drowning rates are correlated. They rise together every summer. That doesn’t mean ice cream causes drowning since both are downstream of a third variable, heat. The correlation is real. The causal story is wrong.
Now apply that to prayer. Person prays, person succeeds, person attributes success to prayer. The sequence is real. But the causal arrow is assumed and not demonstrated. For causation to hold you’d need something harder: you’d need to rule out every other explanation.
Survivorship Bias Does the Heavy Lifting
Here’s the uncomfortable statistical truth: we only hear from people whose prayers were “answered.” The student who prayed and passed tells the story. The one who prayed just as fervently and failed doesn’t frame it as evidence against God, they reinterpret it (“it wasn’t meant to be,” “God had a different plan”). The outcomes that could falsify the claim are quietly reclassified. And nobody keeps score. There’s no ledger of unanswered prayers sitting somewhere because the whole framework isn’t set up to be falsified in the first place.
This is survivorship bias at scale, running continuously across billions of people2. The hits get remembered and the misses get reframed. And what’s left looks like a stunning pattern of divine responsiveness, even if the underlying distribution is entirely random.
A fund manager who beats the market three years in a row looks like a genius. Across thousands of fund managers, some will do this by pure chance. Prayer outcomes work the same way. Enough people praying for enough things and some remarkable coincidences are not just possible, they’re guaranteed. We just don’t naturally think in populations. We think in stories and the stories we hear are always the ones that survived.
The Placebo Is Real (And That’s Not Nothing)
Here’s where I want to be careful because the psychological angle is genuinely interesting and I think people dismiss it too quickly. Prayer probably does help people. Not because it moves the external world but because of what it does internally. It focuses attention. It reduces anxiety. It provides a sense of agency in situations where you otherwise have none. People who pray before high stakes events often report feeling calmer, more grounded, more deliberate and those psychological states have measurable downstream effects on performance.
So there’s a real mechanism. It is just not the one people think it is. The causation runs like this: prayer → psychological state → behavior → outcome. God is not a node in that graph. The comfort is real and the clarity is real and the results are real. But the causal route goes entirely through the person’s own mind.
This matters because it means the benefit of prayer doesn’t require the theology to be true. Which is exactly why it is such a clean example of a confounded correlation, the practice works, the experience feels meaningful and so the metaphysical claim gets a free ride.
What Would Causation Actually Require?
This is the philosophical question worth sitting with. To establish that prayer causes outcomes you’d need something like a randomized controlled trial. Randomly assign people to pray or not pray, hold everything else constant and measure outcomes. Some researchers have actually tried this, intercessory prayer studies and the results are charitably inconclusive3. The better controlled the study, the smaller the effect.4
But more fundamentally: what would a causal account even look like? It would require God to be a variable, something that responds differentially based on whether prayer occurred, whose effect can be isolated from effort, luck and psychology. That’s a strange thing to ask of an omniscient, omnipotent being whose plans presumably don’t hinge on whether someone remembered to fold their hands before bed.
Most theological frameworks, if you press them, quietly dissolve the causal claim. God works in mysterious ways. Everything is part of a plan. The prayer doesn’t change the outcome, it changes you. Which when translated back into secular language is just the placebo mechanism again.
The Takeaway
None of this is an argument for nihilism or even against prayer. A practice that genuinely reduces anxiety, focuses attention and provides meaning is worth having regardless of its metaphysical underpinnings.
But there is an intellectual honesty cost to conflating the two. When we insist that correlation is causation, we stop asking the harder questions about effort, about luck, about the structural conditions that actually determine outcomes. We hand credit to an invisible variable when it belongs to the person who showed up and worked and happened to also pray. That person deserves to own their outcome. I think that’s actually the more meaningful thing to believe.
I don’t think that’s a cynical point. I think it is just an honest one.
[1] This post slightly deviates from my focus on AI and represents my thoughts on the philosophical implications of everyday beliefs, something I personally find fascinating and worth understanding better.
[2] Survivorship bias is the logical error of focusing on outcomes that made it through a selection process while overlooking those that did not. It is strangely relevant to my analysis (on a statistical level) between “answered prayers” and “unanswered prayers.”
[3] The popular belief that prayer ‘causally’ affects patients’ outcomes was debunked in this study where there was no ‘unsurpisingly’ significant difference between patients who were prayed for and those who were not. Interestingly, the patients who were prayed for had a slightly higher rate of complications, though this was not statistically significant.
[4] Further studying the prayer effects, the Cochrane systematic review of intercessory prayer trials concluded that the evidence about the prayer and patient outcomes was inconclusive of being in favor of or against prayer.