The Machine Question is an independent research project exploring artificial intelligence through philosophy, technology and human cognition. Essays on intelligence, creativity and the evolving relationship between humans and machines.
Method · Transparency · Responsibility
AI Use
Policy
How artificial intelligence is integrated into my research and editorial work.
Artificial intelligence can expand the ability to explore, organise, compare and test ideas. It does not replace authorship, critical judgement or responsibility. This page explains how AI tools are used within my research, writing, translation and editorial workflow, and which parts of that process remain exclusively human.
Core principle
AI may assist the process. It is never treated as the author of the work.
Updated 21 August 2026
02
Core principles
Human judgement remains the governing layer.
AI tools may contribute to parts of the workflow, but they do not determine the argument, the final wording or the decision to publish. Those remain human editorial responsibilities.
Human authorship
Signed work is presented as human-authored. AI may assist the process, but it is not credited with authorship or treated as the intellectual origin of the published argument.
Human control
AI-generated suggestions, drafts, comparisons or transformations are subject to editorial selection, revision and rejection before they become part of published material.
Claims must be checked
AI output is not treated as a source of authority. Material facts, quotations, references and claims requiring verification should be checked against appropriate underlying sources.
Responsibility stays human
Responsibility for what is published, including errors, omissions, interpretation and editorial judgement, remains with the human author and cannot be transferred to the tool used during production.
Governing principle
AI is used to extend analytical capacity, not to outsource judgement.
03
Operational use
Where AI enters the workflow.
AI may be used at specific stages of research and production where it improves comparison, structure, linguistic precision or editorial efficiency. Its role varies according to the work and is never assumed to be necessary.
Exploration and orientation
AI may assist with preliminary exploration, identifying concepts, questions, possible connections or areas requiring further research. It is not treated as a substitute for primary or authoritative sources.
Testing relationships
Models may be used to compare arguments, terminology, interpretations or alternative structures, helping expose inconsistencies or relationships that merit closer human examination.
Organisation of material
AI may help organise notes, map conceptual sequences or test possible architectures for an article. The final argument and its hierarchy remain editorial decisions.
Revision and language
AI may assist with editing, linguistic refinement, consistency checks, compression or expansion. Suggested wording is accepted only where it preserves the intended meaning and authorial voice.
Cross-language work
AI may support translation or comparative language review. Final translated text is subject to human revision for accuracy, cultural equivalence, tone and conceptual precision.
Visual development
Generative tools may contribute to visual concepts, editorial imagery or design exploration. Their use does not alter the responsibility for the final selection, framing or publication of the visual material.
Practical rule
AI can accelerate a stage of the process. It does not determine the intellectual destination.
04
Human domain
What is not delegated to a machine.
The distinction between assistance and authorship is not determined by whether AI has touched a text. It is determined by who originates the intellectual position, exercises judgement and accepts responsibility for the finished work.
Human responsibility
The argument must remain attributable to a person.
The choice of subject, underlying thesis, interpretation of evidence, intellectual position and final decision to publish remain mine. Where a work carries my name, I accept responsibility for it irrespective of the technologies used during its preparation.
Machine role
Assistance without authority.
An AI system may propose, reorganise, compare, translate, challenge or reformulate material. None of those operations gives the system editorial authority over the work or makes its output inherently reliable.
Intellectual position
The central thesis and the perspective from which a subject is examined are human decisions rather than conclusions delegated to a generative system.
Editorial judgement
Relevance, proportion, interpretation, emphasis and the distinction between what should be included and excluded remain matters of human editorial judgement.
Authorial coherence
AI-assisted language is revised where necessary so that published work remains consistent with the conceptual vocabulary, tone and analytical method of the author.
Final accountability
Publication is a human act. The decision that a work is sufficiently accurate, coherent and complete to appear under my name cannot be delegated to an AI system.
The boundary
The machine may participate in the making of the work. It does not inherit the authority of the author.
05
Evidence & disclosure
Sources, data and synthetic material.
The use of AI does not change the standards applied to evidence, attribution or factual accuracy. Where synthetic or AI-assisted material is relevant to understanding how a work was produced, that use should be disclosed in an appropriate way.
Authority remains external
AI-generated statements are not cited as evidence merely because a model produced them. Factual claims requiring support should be grounded in identifiable primary, institutional, scholarly or otherwise appropriate sources.
Numbers are checked
Data, dates, statistics and quantitative claims generated or reformulated with AI assistance are subject to verification before publication where they materially affect the argument.
Meaning before fluency
AI-assisted translation is reviewed for conceptual accuracy, terminology, register and cultural equivalence. Fluency alone is not treated as evidence that a translation is faithful.
Synthetic imagery
AI-generated or AI-assisted visuals may be used for conceptual or editorial purposes. Where a visual could reasonably be mistaken for documentary evidence of a real event, person or situation, appropriate disclosure should accompany it.
Transparency standard
Disclosure should be proportionate to the role AI actually played, not performed as a ritual detached from the substance of the work.
06
Responsibility & governance
Responsibility does not move to the machine.
AI tools, practices and legal requirements continue to evolve. This policy therefore describes a method rather than endorsing any particular model or technology, and may be updated as the relevant technical and regulatory environment develops.
Published work remains my responsibility
The use of an AI system does not reduce or transfer responsibility for published content. Errors, interpretations, omissions and editorial decisions remain attributable to the human author who chooses to publish the work.
Applicable rules are respected
AI-assisted activities are conducted with regard to applicable rules concerning data protection, copyright, transparency and the use of artificial-intelligence systems, including relevant European Union and United Kingdom requirements where they apply.
The method can change
Tools and workflows may evolve as capabilities, risks and standards change. This policy may therefore be revised, but the underlying principles of human authorship, verification, transparency and responsibility remain the reference point.
Final position
I use artificial intelligence because it can extend what a human researcher and writer is able to examine. I do not use it to remove the human from the act of thinking.
Questions & transparency
Questions about this policy?
Questions concerning the role of AI in material published on paolocalvi.co.uk, or requests for clarification about this policy, may be sent directly to Paolo Calvi.
Contact
privacy@paolocalvi.com