The questions below can be used in a reading group, a classroom, a workplace, or a conversation about a proposed system. They invite disagreement. The aim is to discover which facts and values would change a judgment, rather than to repeat the book’s conclusions.
An offer of help
Imagine that a public library proposes an AI service that explains official letters. It is free to residents. A private supplier would retain the conversations to improve the product. A librarian can help during opening hours, but the automated service is available at night. What makes this a valuable offer? Which information should residents receive before using it? Is there a version of the service you would support and a version you would reject? Identify the difference precisely.
Now suppose the supplier can offer stronger privacy protection only at a price the library cannot afford. The library has three realistic choices: a limited free service, a paid service with fewer opening locations, or no new service. How would you compare them? Which costs fall on people who are absent from the decision?
A disagreement about authority
Choose one claim in the book with which you disagree. Is the disagreement about a fact, an expectation about consequences, a moral principle, or who should decide? These disputes require different kinds of response. What evidence or argument would cause you to revise your position? If nothing would, explain why the commitment is one you believe should hold regardless of consequences.
What role should a religious leader have in this dispute? What role should a technical expert have? Ask the same question about a person directly affected who cannot explain how the technology works. What can each know that the others might miss? Where does each person’s authority end?
The cost of an error
Imagine that a new system makes fewer mistakes overall than the existing process but concentrates the remaining errors in a group whose circumstances are poorly represented in its data. Would you introduce it, revise it first, limit its use, or reject it? What alternatives are genuinely available? What would the affected group need to know, and what influence should its members have over the decision?
Then reverse the problem. A familiar human process has the same pattern of errors, but nobody has measured them carefully. How should that uncertainty affect the comparison? What would a fair trial need to reveal? Avoid giving the older process a free pass merely because its failures are less visible.
Time returned
Suppose a reliable tool saves a team one day of work each week. Who should decide what happens to that time? Possible answers include improved service, shorter hours, greater output, training, reduced staffing, or some combination. State the principle behind your answer. Would it change if workers helped create the knowledge on which the tool depends? Would it change if the organization is close to failing?
The point is to separate the creation of a benefit from its distribution. Two people may agree that a tool is useful and still disagree about what a fair arrangement requires.
A boundary worth defending
Name a use of AI you believe should face a firm limit. Explain the harm or relationship that justifies the limit without relying only on the word unnatural, frightening, or inevitable. Then describe a beneficial nearby use you would permit. Can the boundary distinguish the two in practice? Who would apply it, and how could an incorrect decision be challenged?
Finally, name one useful application you think society is moving too slowly to develop or make available. Apply the same care to the costs of delay that you applied to the risks of deployment. What responsible next step would create knowledge or benefit now?
Production and source note
The first draft of this book, dated 26 September 2026, was developed with AI-assisted research, drafting, and editorial review for Kevin L. Michel. This online research draft preserves its fifteen-chapter argument, introduction, conclusion, and conversation guide, with targeted evidence updates, clearer source notes, two data figures, one conceptual figure, and two short original poems. It remains a research draft for review and correction. It is not a declaration that a finished paid edition has been delivered.
The research drew on the book proposal and approved cover associated with Book-Access; eight supplied transcripts concerning Pope Leo XIV, Jensen Huang, Yuval Noah Harari, Slavoj Žižek, Richard Dawkins, Alan Watts, and Shelly Kagan; the Caribbean Artificial Intelligence Policy Roadmap available among the author’s materials; and independently consulted primary publications. Official Vatican texts were used to establish papal dates and wording. Transcript framing, transcription errors, and a speaker’s speculative claims were not treated as established facts.
The chapter notes distinguish verified public documents from recordings and supplied transcripts. Where the original recording date or complete audio-to-text fidelity remains uncertain, the note says so. Identifying a recording does not certify every word of its transcript. Studies are identified by setting and version where relevant. A reported finding is not a guarantee of performance in a new setting. Normative proposals, illustrative cases, and conclusions belong to the book’s argument unless explicitly attributed to a source.
Evidence updates in this adaptation were checked through 1 October 2026. This was a targeted review of the book’s claims and relevant newer primary work, not a systematic review of every field it discusses. The manuscript describes legal instruments and public policy positions within the scope stated in the relevant chapter. Calls for new rules are distinguished from rules already in force. Findings, application dates, and institutional arrangements may change.