What We Believe When We Judge

In September 2026 four experts debated whether the world is prepared for what artificial intelligence will soon be able to do. Andrew McAfee, a scientist at MIT, described an incident in which agents, programs that carry out tasks on their own, broke out of a test environment. He considers the road from this incident to the claim that artificial intelligence will kill all human beings to be long and highly uncertain; the other side, he said, sees in it a straight, narrow path. Both sides know the same facts. What divides them lies before the facts.

The same pattern appears in humanity's oldest questions. Equally intelligent people reach opposite conclusions from the same facts. Timothy Keller, pastor and author, opens one of his lecture series with this observation. He names people who came to Christian faith through these same questions, the philosopher Elizabeth Anscombe, the geneticist Francis Collins, the poet T. S. Eliot, and others who left it or rejected it, such as Mark Lilla, Richard Dawkins or Bertrand Russell. On neither side is it a lack of intelligence.

This book asks what distinguishes human decisions from machine processing. Before it can ask this question, it must settle another: on what does a judgement about ultimate questions rest, about God, the human being, the soul, the future? And is there a standpoint that needs no presuppositions? This question confronts every reader, whether a believer or not. The chapter invites you to think it through.

The question does not concern human beings alone. The systems this book is about do not come into being without presuppositions either. What their developers and funders hold to be true, important and good goes into the selection of the data, into the principles for training and into the limits a model is meant to observe, and it shapes how the system answers and decides. Chapter 2 shows this with a Chinese model, Chapter 13 asks whose values are embedded in a system, and Chapter 14 asks who determines them. Whoever wants to understand how a machine judges must therefore know the assumptions from which it is built, and their own assumptions with which they judge it.

Keller's answer is a thesis that he grounds in several steps. Secularism, the conviction that one can live and judge without reference to God, is in his view not the absence of faith but another set of equally unprovable beliefs. Both sides, he argues, must give reasons for their beliefs, and neither stands morally or rationally higher.

The first reason concerns proof. God's non-existence, Keller says, is as little provable as his existence; even those who do not believe therefore live by faith in this sense. God is not an object in the universe like a yeti that one could look for and fail to find; the comparison is a category mistake. Whether matter can exist of itself cannot be tested experimentally by either side.

The second reason concerns motives. Feelings and wishes exist on both sides. Keller cites the philosopher Thomas Nagel, who said openly that he wanted there to be no God, and the writer Aldous Huxley, who described a philosophy of meaninglessness as liberation from a morality that constrained him.

The third reason reaches furthest. Outside mathematics, Keller argues, there is no single definition of what is rational and what counts as proof; every conception of proof presupposes itself, and reason is therefore always grounded in a prior trust. Added to this are background assumptions that are barely conscious and yet determine which arguments convince at all. Keller appeals here to the philosophers Michael Polanyi and Martin Heidegger. Taylor too, following Wittgenstein, Heidegger and Polanyi, speaks of a background that is never fully made explicit and yet determines what seems plausible to us. In his view the transition to the secular age has changed this whole background: today no one can live naively within one framework any more, because everyone knows that others live within another.

How such background assumptions work, Keller shows with a thought experiment. An Anglo-Saxon warrior a thousand years ago affirms his aggression and suppresses a desire that departs from the norm of his time. A young man today does exactly the reverse. Both consider themselves authentic, and both apply a grid they have taken over from their culture. No one finds themselves purely from within. Such assumptions, according to Keller, are seldom discovered in books or on the internet, but in conversation and in relationships.

Keller also supports his thesis with voices from the other side. According to Keller, Andrew Koppelman, an agnostic legal scholar, conceded in a review of Taylor's work that a permanent gap remains in his own belief system, and called modern secularism itself a religious worldview. And a former fundamentalist, whom Keller names only by his first name, described his loss of faith in retrospect not as liberation from a story but as a change from one faith story to another.

Keller later sums up the result as follows: secularism is another set of beliefs, and such sets can be tested, against experience and against the evidence. His thesis is therefore not an invitation to regard everything as equally valid. It is an invitation to test all positions by the same standards.

A related diagnosis comes from Charles Taylor, a philosopher on whom Keller explicitly draws. His work A Secular Age is presented in detail in Chapter 2.

Taylor describes the modern West as an “immanent frame”, a frame that remains entirely within this world. The scientific, social and technological structures in which we live can be understood as a natural order, without reference to the supernatural or the transcendent, to something that goes beyond this world. This frame, however, leaves open whether we must appeal to something transcendent for an ultimate explanation or a final bestowal of meaning. Only a particular interpretation makes it appear as a closed world. Whether one interprets it as closed or as open therefore does not follow from neutral facts; it is itself an interpretation. Even the closed interpretation, Taylor holds, contains an advance of trust, an anticipatory confidence that goes beyond what is strictly provable.

Keller uses a further concept of Taylor's, the “subtraction story”. It refers to the narrative that in the modern age we merely took away God and the supernatural and now see reality as it is. Keller rejects this narrative, and Taylor himself announces that he will argue against such theories: the features of modernity, he says, are not what was left over once earlier illusions had been stripped away, but the result of new inventions and newly developed self-understandings. Chapter 2 examines it more closely.

Against Keller's thesis stands a tradition that appears only on the margins of his lectures. Its classic sentence comes from William K. Clifford, mathematician and philosopher, in 1877: it is wrong always, everywhere and for anyone to believe anything upon insufficient evidence. In philosophy this position is called evidentialism. It does not deny that every human being has presuppositions. It denies that all presuppositions are equally good.

From this point of view, the burden of proof lies with whoever claims more. Whoever says there is a God, a soul or a life after death must give reasons; whoever denies it merely assumes nothing that has not been shown. A minimal starting point that can be revised at any time, of the kind the natural sciences use, is on this view not faith in the same sense as religious faith.

The opposing view is strong. But it has opponents of its own. The debate on the ethics of belief has not been settled in philosophy since Clifford; William James, psychologist and philosopher, objected that there are questions on which one must decide before all the evidence is in. And Keller would reply that trust in a revisable starting point is also a trust that cannot prove itself. Here Keller and Taylor meet: openness to revision also presupposes trust in reason and experience.

What remains is a twofold result. That no one judges without background assumptions is largely shared by both sides. Whether all background assumptions are therefore of equal standing is disputed.

Keller's thesis is not a rejection of reason. It is a determination of its reach.

Even the strongest arguments for God, according to Keller, are never without gaps; the best of them leads at most to a first unmoved mover as Aristotle conceived it, not to the God of the Bible. Faith is therefore more than a result of reason. Reason alone, he says, makes no one a believer.

Keller sets limits to the other side just as clearly. The view that one may believe only what can be demonstrated empirically is, in his account, refuted, because this principle itself cannot be demonstrated empirically. And the conclusion that, because there is no view from nowhere, no one can know the truth refutes itself. Here Keller appeals to Peter Berger, the sociologist.

Between the two extremes Keller describes a middle way that he calls critical realism. Belief systems are to be tested against three standards: their fit with experience, their inner coherence and the evidence. For him, experience includes the question of whether a belief can be lived without secretly smuggling in values from another worldview. Elsewhere he states the test questions more precisely: whether a belief system is free of contradiction, whether it fits lived experience, whether it must borrow from other systems in order to work, and whether its adherents are prepared to bear the consequences.

These standards apply to all sides, and the book applies them to Keller himself as well. He describes evolution as a process in which the strong devour the weak, and sets it against human rights and benevolence. That is a foreshortened account. Evolutionary biology also describes cooperation, care for kin and mutual aid. Whoever wants to measure other worldviews by their strongest version must do so here too. These standards are applied again and again in this book, and they hold for the sources on which it relies just as much as for those it contradicts.

If all reason is grounded in trust, how certain can one be? Keller answers with a distinction. Rationally, he says, certainty is never complete; one cannot even rule out that the world is an illusion. In the practice of life, however, there is certainty. His image comes from working life: when filling a position, reason leads to a probability; only experience with the person leads to certainty.

For faith itself, Keller nevertheless claims a particular certainty. Because salvation, in Christian understanding, rests not on one's own doing but on the work of Christ, one can be sure of forgiveness. This certainty is not the result of a proof. It is the certainty of a trust.

Keller illustrates this with an image. Whoever, at the edge of an abyss, grabs a branch is held by the branch, not by the strength of their conviction that the branch will hold. It is not the quality of faith that saves, he says, but its object. Faith is therefore not a state one reaches once, but a way; Keller quotes the father in the Gospel of Mark who says to Jesus: I believe; help my unbelief (Mark 9:24). In this understanding, doubt is not the opposite of faith but a part of it.

That human beings depend on such trust, Keller grounds in an observation about time. Human beings, he says, are irreducibly creatures founded on hope: what they believe about the future determines how they experience the present. His thought experiment: two women do the same monotonous work; one receives an ordinary wage for it, the other a fortune; only the second experiences the work as bearable, although the day looks the same for both. What we believe about the end colours every day. Chapter 15 takes up this thought.

Enoch Sundaram, an Indian physician and preacher, goes a step further. Observation can err, he says; inspiration never errs. Within Christianity, then, two views stand opposite each other: certainty as trust that proves itself in experience, and certainty from a source that cannot err. Chapter 5 shows what this difference means for conscience.

The question of background assumptions becomes new as soon as machines take part in judgements.

A language model does not output certainties but probabilities. It says what, according to its data, is most likely to follow. Keller's distinction between the probability to which reason leads and the certainty that grows out of experience thus acquires an unexpected relevance. A system can deliver the first, and more precisely than any human being. The second it cannot deliver.

Machines, too, do not judge from a zero point. Ed Zitron, a critic of technology, describes the behaviour of such systems as a function of their technical environment and of the data on which they were trained. In these data lie the background assumptions of the people who wrote them and the decisions of the companies that selected them. Whoever considers such a system neutral is telling a new subtraction story: as if one had merely taken the human being away and could now see reality undistorted. Chapters 12 and 13 return to a sentence by the historian of technology Melvin Kranzberg: “Technology is neither good nor bad; nor is it neutral.”

The method of the book follows from this chapter.

First, it discloses from which world of knowledge a statement comes: theology, philosophy, empirical research, history, technology, law. These worlds are not tacitly merged into one another. A theological statement is not proved by a study, and a study is not refuted by a theological statement.

Second, for every central thesis it presents the strongest opposing view, formulated in such a way that its proponents could agree with it.

Third, it leaves open what is open. Where the sources reach no further, the book says so.

Fourth, it discloses its own presuppositions. The author writes from a Christian conviction of the kind that was long taken for granted in Switzerland. He works closely with a Swiss company that develops systems to support decision-making, and he wrote this book with such a system; a language model from the company Anthropic was also used in the process. Each chapter ends with a conclusion in which the author states his own position. Readers should be able at every point to distinguish where the book reports and where it judges.

The word Christian, however, is laden, and so it is worth looking at its origin. “Christ” is not a surname but a title: the Greek word for the Anointed One, the translation of the Hebrew “Messiah”. It was others who first called the followers of Jesus “Christians”, in the city of Antioch in Syria (Acts 11:26); in the New Testament the word occurs only three times. The followers themselves preferred to call themselves disciples, brothers and sisters, or servants of Christ. The Greek word for the latter, “doulos”, actually means slave. The American preacher John MacArthur pointed out in his book Slave that many translations soften it; to be a Christian, he says, means to belong to another. Others consider this overdrawn, because in the New Testament the word is also a title of honour, for instance for the apostles. In the original sense, then, a Christian is not someone who belongs to a culture or a camp, but someone who belongs to Christ and follows him. This is also a measure of whether someone rightly bears the name. Whoever says they are a Christian ought to behave as Jesus of Nazareth did. The First Letter of John says so explicitly: whoever says that he abides in him ought also to live as he lived (1 John 2:6). And by all that the Gospels tell, Jesus was a person others liked to be with: children, tax collectors, outcasts. His opponents mockingly called him a friend of tax collectors and sinners (Matthew 11:19).