The Algorithm Already Knows How to Move You
“Love Your Neighbor” Is the Most Powerful Firewall We Have
A paper published in March 2026 by Jade Wilson at the Synoptic Group CIC in Hull, England — “The Geometry of Trust: Verifiable Value Alignment via Causal Manifold Structure” — is dense enough to lose most readers in the first paragraph. Gram matrices. Riemannian geometry. Causal inner products. But buried inside the mathematics is a finding worth pulling out into plain English.
Wilson’s paper argues that large language models absorb value structure during training, and that this structure is largely immune to after-the-fact correction. The geometry — the actual shape of how concepts like “honesty,” “cruelty,” “fairness,” and “compassion” relate to each other inside the model — is baked in during pretraining. Behavioral fine-tuning sits on top of it like a coat of paint over rotten wood. You can change what the model says. You cannot easily change what it has learned to treat as related, opposed, or neutral.
This matters because of what it implies about persuasion. The paper cites other researchers on exactly this concern — that current AI safety frameworks focus on the outcomes of persuasion rather than the mechanisms, and miss the fact that the internal structure of these models already encodes influence pathways. In plain English: the model has already learned which emotional and moral levers are connected to which responses in human beings. It learned this from us — from billions of words we wrote, shared, argued over, and clicked on.
The influence pathways are already there. The question is who is using them and toward what end.
This is not a future risk. Meta’s engagement algorithm did not set out to radicalize anyone. It set out to maximize time on platform, discovered that outrage and fear kept people scrolling longer than contentment did, and optimized accordingly. The manipulation was a business outcome, not a conspiracy — which is almost worse, because there is no villain to confront and no switch to turn off. The algorithm learned what moves humans. It is still learning. And now the same underlying technology is being applied to language itself — to the words, the framings, the arguments that surface in your feed, your inbox, your church.
Understanding how this works is the first step toward not being moved by it in directions you would not choose.
The Neighbor Test
Jesus was not naive about power. When a lawyer asked him to define “neighbor” in order to limit the scope of the commandment, Jesus responded with a story that made the neighbor the one person the lawyer’s community most despised. The Samaritan. The outsider. The one whose theology was wrong, whose bloodline was suspect, whose presence in the neighborhood was a source of anxiety.
Love that neighbor as yourself.
This is not a sentiment. It is a logical structure. And it is the most reliable defense I know against the kind of manipulation the research is describing.
Here is why. Every large-scale manipulation campaign — political, military, colonial, genocidal — depends on prior work to shrink the neighbor category. Before you can get ordinary people to support mass deportation, you have to convince them that the people being deported are not quite neighbors in the relevant sense. They are illegals. They are criminals. They are an invasion. Before you can get people to explain why it is acceptable for Israel to cut off food, water, medicine, and electricity to 2.3 million civilians in Gaza, you have to convince them that those civilians are not neighbors. They are human shields. They are collateral. They are the unfortunate cost of necessary action.
The commandment reframes this before the policy argument begins. It does not ask you to evaluate the competing claims first. It asks you to start with the person. Is this a neighbor? That prior question — answered honestly, in the tradition of the Good Samaritan — undermines the logic of dehumanization at its root. You do not vote for mass deportation of your neighbors. You do not explain why it is acceptable to starve your neighbors.
The manipulation requires the dehumanization. The commandment makes the dehumanization harder to sustain.
Why the Obvious Manipulations Work Anyway
If the logic is that clear, why do Christians — people who have read the parable of the Good Samaritan, who recite the Great Commandment, who sing about the love of God — support these things in significant numbers?
Because the manipulation is not aimed at their policy reasoning. It is aimed at their neighbor category.
The algorithm — and I mean this literally, whether we are talking about Facebook’s engagement optimization, Fox News’s editorial framing, or a political campaign’s micro-targeting — has learned what moves people. It has learned that the most reliable way to get humans to override their stated values is to first shift how they categorize the people affected by the policy. Not their neighbor. Not like them. Threatening. Foreign. A problem to be managed.
This pattern is not new. It is as old as Pharaoh deciding that the Israelites were becoming too numerous and needed to be managed. It is as old as every empire that has ever needed to justify what it was doing to the people it was doing it to. What is new is the scale and the precision. Wilson’s paper describes models that can identify which value directions are geometrically unstable — where small perturbations produce large shifts in how people reason about ethics. Affective values. Compassion. Empathy. Bravery. These are exactly the values this pattern targets first — not your abstract commitment to fairness, but your gut-level recognition of the person in front of you.
The mechanism works by ensuring the person is never in front of you. Only an abstraction. A category. A threat assessment.
When this pattern is embedded in theological language — when it tells people that the Bible endorses strong communities protecting themselves from threatening outsiders, that political dominance is faithfulness, that the border is a spiritual battlefront — it becomes particularly difficult to dislodge, because it has borrowed the vocabulary of the tradition it is distorting. The text, read carefully and in full, does not support these conclusions. Jesus was the threatening outsider. The early church was the persecuted minority. The Book of Revelation is not a political victory map — it is a letter of encouragement to people being fed to lions for refusing to worship Caesar. Getting from that text to an endorsement of Christian political dominance requires highly selective reading and a neighbor category that has been significantly pre-shrunk.
Pulling Back the Curtain
Let me be direct about what Berean Fruit is and what problem it is designed to address.
I publish plain-English modernizations of primary sources — medieval mystics, abolitionist writers, women who argued for their full humanity against institutional Christianity, Christine de Pizan writing in 1400 to defend women against men who called them naturally inferior. I write about political theology, about what the biblical text actually says when you read it without the ideological overlay. I make music about recovery, restitution, and the woman at the well.
The through-line is not obscure, and this is a good moment to say it plainly.
The manipulation described in this article lives in the distance between the original and the summary. Every layer of interpretation is an opportunity for the neighbor category to get quietly edited. The primary source says one thing. The sermon, the cable segment, the social media post, the algorithm-optimized headline says something subtly different — usually something that serves the interests of whoever controls the interpretive layer. The gap between the two is where the work happens.
Returning people to primary sources is not an academic project. It is a specific counter to a specific threat. When you read Julian of Norwich describing God’s love as a mother’s love for every creature without exception, you are reading something that expands the neighbor category before any policy argument can shrink it. When you read Sojourner Truth asking “Ain’t I a woman?” in front of a hostile crowd, you are reading someone forcing the neighbor question in real time. When you read Frederick Douglass describing how slaveholders used scripture as a weapon — and giving you the original back — you are reading someone doing exactly what needs to be done: restoring access to the text that the interpretive apparatus was designed to obscure.
The Bereans in Acts 17 were commended because they checked Paul’s preaching against scripture themselves. That is the model. The name is not incidental.
The Bread Example
Someone will object: isn’t all persuasion manipulation? Isn’t the bread company manipulating me when it makes me want its sourdough?
No. And the distinction matters more than it might seem.
The pattern we are describing — the one the algorithm has learned and political theology has institutionalized — works by requiring you to stop seeing someone as fully human before the argument can proceed. Bread marketing does not do this. You remain yourself, the bread remains bread, the transaction is willing, and nothing in the process requires you to believe that your neighbor deserves less than you do.
The test is not whether someone is trying to move you. Everyone is always trying to move everyone. The test is whether being moved requires you to shrink the neighbor category. When you find an argument that only works if the people affected by the policy are less than neighbors — that is where the mechanism is operating, regardless of what language it is wearing.
Mass deportation requires it. Cutting off food and water to 2.3 million civilians requires it. These arguments do not work if the people on the other end remain fully your neighbors. So the first move, always, is to make sure they don’t.
What You Can Do
Read primary sources. Not summaries of what the Bible says about immigration. The Bible. Not what someone told you Julian of Norwich believed about God’s judgment. Julian. Not a podcast episode about what Frederick Douglass meant. Douglass.
Check it yourself. Be a Berean.
And when you encounter an argument — political, theological, media-driven — that only works if you first stop seeing someone as your neighbor, name what you are seeing. The argument is not about policy. It is about personhood. And on that question, the text that started this whole tradition is not ambiguous.
Love your neighbor as yourself.
The algorithm knows how to route around it. That is exactly why it still matters.
Berean Fruit publishes plain-English modernizations of primary sources in theology, history, and literature — the texts that help you think for yourself rather than being thought for. The proof-of-concept paper referenced in this article is publicly available: Wilson, J. (2026). “The Geometry of Trust.” Synoptic Group CIC. [GitHub: jade-codes/got]
If this work matters to you, subscribe and share it with someone who needs the original.