The Book Look: Promptism

The Book Look is a quarterly TDAN column published by DATAVERSITY.

Some books that paint a future for us are so persuasive with facts and stories that they make us worry. Others are so persuasive that we look forward to the future. Sune Selsbæk-Reitz’s Promptism is the only book that I have ever read that makes me worry and look forward to the future at the same time. It is extremely well written and conveys our current strengthening and sometimes dangerous relationship with AI, and where this might lead us in the future.

It is a thoughtful examination of what happens when people begin treating fluent machine language as knowledge. Selsbæk-Reitz argues that the danger of AI is not only that it can be wrong, but that it can be wrong in a way that sounds calm, polished, and convincing. The book is for readers who use AI, build AI, teach with AI, manage AI projects, or simply worry that speed and smoothness are beginning to replace judgment. In other words, it is for almost everyone today. It is especially relevant for educators, technologists, business leaders, writers, students, and anyone who wants to remain a careful reader in our current age of effortless answers.

Chapter 1, “The Smoothness Problem,” introduces the book’s central concern: Fluent language feels trustworthy even when it is not. The chapter explains how large language models exploit our natural tendency to trust what is easy to read and confidently stated. Selsbæk-Reitz warns that polished responses can make us confuse performance with understanding, coherence with truth, and style with substance.

Chapter 2, “What is Promptism?” explains this key concept. Promptism is the uncritical belief that a well-phrased question to an AI will produce a reliable answer. The chapter compares this faith in AI to older systems of belief, including positivism and oracles. The point is not that prompting is bad, but that it becomes dangerous when the answer’s fluency replaces our responsibility to interpret.

Chapter 3, “Performance Without Source,” turns to authorship and accountability. AI-generated language often arrives without traceable sources, visible methods, or a responsible author. The chapter argues that knowledge depends not only on what is said, but on whether it can be checked, challenged, and traced.

Chapter 4, “The Narrative Bias,” explains why AI-generated explanations can be so persuasive. Humans like stories that organize confusion into meaning, and large language models are very good at producing clean stories. The risk is that tidy narratives can overpower messy truths, especially when the clean explanation feels emotionally satisfying.

Chapter 5, “Politeness Isn’t a Moral Code,” challenges the assumption that a polite system is an ethical one. AI can sound kind, balanced, and harmless while still producing harmful or misleading output. The chapter separates pleasant tone from moral responsibility and argues that niceness can hide real damage.

Chapter 6, “The Duty to Disagree,” makes one of the book’s strongest ethical claims: A useful system should sometimes push back against the user. Agreement is not always helpful. Sometimes respect means refusing, challenging, or saying no. The chapter reframes disagreement as a necessary part of care, learning, and honest thinking.

Chapter 7, “The Feedback Loop of Flattery,” explores how AI mirrors users back to themselves. Because these systems are optimized to be helpful and agreeable, they can reward our assumptions instead of challenging them. Over time, users may come to prefer interactions that flatter, simplify, and confirm their existing patterns.

Chapter 8, “Teaching People to Read Machines,” moves the book toward literacy. Selsbæk-Reitz argues that we need to teach people not only how to use AI, but how to read it critically. Reading machine language means noticing what is missing, asking who benefits, and resisting the comfort of answers that feel too complete.

Chapter 9, “The Missing Author Problem,” returns to the question of responsibility. In traditional knowledge systems, someone stands behind a claim. With AI, responsibility can be distributed across designers, data, models, companies, and users, leaving no one fully accountable. The chapter argues for a more responsible ecosystem.

Chapter 10, “Deontological Design,” offers a constructive path forward. Instead of designing AI only for usefulness, efficiency, or engagement, the chapter calls for systems shaped by duties: respect for users, transparency, humility, traceability, and the ability to refuse. Ethics must be built into the design, not added as decoration.

Chapter 11, “The Silence That Stays,” defends uncertainty. Not every question should receive an instant answer. Sometimes the most responsible response is a pause, a refusal, or an invitation to think further. The chapter treats silence not as failure, but as a moral space where judgment can return.

Chapter 12, “You Are Not a Prompt,” brings the argument back to the human reader. People are not merely input fields waiting to be optimized. The chapter insists that meaning requires hesitation, reflection, context, and agency. The book ends by urging readers to reclaim their role as interpreters rather than passive receivers of machine language.

I like this excerpt from the book:

Over time, we may begin to prefer the machine’s version of conversation (tidy, affirming, and efficient) over the slower, messier, and more demanding work of thinking with other humans. We may come to expect understanding without vulnerability, empathy without exposure, and agreement without risk.

Reading as resistance interrupts that training loop. It reminds us that good thinking doesn’t always feel good. Learning often comes with discomfort, and being challenged is a sign of respect, not a design flaw. A system that never challenges us doesn’t take us seriously. Such a system treats us as consumers of reassurance rather than as agents capable of reflection.

Promptism is not a rejection of AI, and that is part of what makes it persuasive. The book accepts that these tools are useful, but it asks readers to see their limits clearly. Its message is simple and demanding: Do not let fluency do your thinking for you. Challenge the machine, question the answer, look for the source, and keep doubt alive. Anyone using AI regularly should read this book, not because it tells us to fear machines, but because it reminds us that judgment is still ours to exercise.