Paul Cooper.AI

Paul Cooper · Published writing

Papers &
responses.

Read the argument, follow the sources, and see what has been checked.

Two linked cobalt and copper frames
Conceptual paper · Published

A Group’s Success Does Not Tell Us Who Has a Say

Human authority in shared decisions, roles, and goals.

A response to Michael Levin’s Ingressing Minds. It asks whether people can question their roles, correct the record and reconsider a shared plan, even when a group performs well.

Evidence and review status

Arguments, cited sources and elementary mathematical proofs. No new empirical study or proof-assistant verification is claimed.

No independent external expert validation is recorded for this edition.

Essay · Published

A More Capable Group Should Still Hear Its Members

The companion essay.

A plain-English introduction to the same question: does becoming more capable together also leave each person with a meaningful say?

Evidence and review status

An author argument and a proposed empirical question; the paper provides the fuller reasoning and sources.

Publication on the website or Substack does not imply independent expert validation.

Letters

Personal responses, with the original message kept distinct.

Draft response · Public silent edition

The Kággaba statement on AI & our response ↗

Read the original statement first, then Paul’s response. The current public edition uses text and visuals; it has no narration. The response is AI-assisted and remains a draft.

In preparation

Mathematical papers

Further manuscripts are being prepared for publication. Released papers will appear here with their assumptions, source materials and verification records. Drafts are not listed as published work.

Follow new writing on Substack ↗

What supports a claim?

Publication status and evidence status answer different questions. Each status belongs to a particular claim and version.

Author claim
An assessment or proposal attributed to its author. Read its argument, sources and stated limits; publication alone does not establish independent validation.
Checkable computational evidence
A linked software-generated record, with the exact checked scope, inputs, code or checker, and run instructions. A numerical test and a formal proof certificate are identified separately.
External expert validation
A named expert with relevant domain expertise supports a specified claim in an attributable review. The reviewer, version, conclusion and limits must be linked.

An AI review, a submitted manuscript or a published webpage does not by itself qualify as external expert validation. An independent rerun is recorded separately from the original computation.