The Third Way
A Framework for Cooperative Intelligence, by John McClain. The whole idea in one book: three laws, one coherent system, and the practical thinking to get from principles to real life.
Free access • No registration • Open for academic and personal use
On this page
This book was written to be read by anyone, not just researchers. It is free, there is no sign-up and no email to hand over, and you can read it in your browser or download it to keep. If you are curious where this whole idea comes from and want it in one place, this is that place.
The book at a glance
- What it is: the complete guide to Third Way Alignment, presenting the Three Laws, the cooperative intelligence model, illustrative case studies, implementation strategies, and future scenarios in a single volume.
- Access: free, with no registration. The PDF opens in your browser for online reading and can be downloaded for offline use or sharing.
- Relationship to this site: the book is a complete statement of the framework as written. The framework continues to be refined, and where wording differs between the book and this site, the site carries the current canonical formulation.
Why I wrote this book
The AI governance landscape is fractured. Governments pass reactive legislation after each new crisis. Corporations draft ethical principles they struggle to enforce. International bodies issue guidelines with no teeth. Everyone is building pieces of a puzzle without seeing the full picture.
This book takes a different route. Rather than responding to individual problems as they arise, it starts from a coherent set of principles and works outward. Many of the questions it takes up, including accountability for autonomous systems, the distribution of AI's benefits, and governance that can keep pace with the technology, are the same questions regulators now confront. The book's contribution is not a competing rulebook but an argument that these questions belong to one integrated problem and deserve one integrated answer.
The framework rests on three laws: Mutual Respect, Shared Flourishing, and Ethical Coexistence. These are not abstract ideals. Each states a normative commitment, and each carries an Operative Principle that defines when and how its obligations activate.
What the book covers
Illustrative case studies
Constructed scenarios showing how human-AI cooperation could work across industries. These are illustrations of the principles in action, not reports of completed deployments.
Implementation strategies
Step-by-step guidance for applying the framework's principles in projects, organizations, and individual AI interactions.
The cooperative intelligence model
How human creativity, intuition, and values can combine with AI computation, pattern recognition, and optimization to solve problems neither could address alone.
Future scenarios
Preparing for more capable AI while maintaining human agency, including the thresholds and safeguards the framework proposes for that preparation.
The same scenarios are available in interactive form on the Interactive Scenarios page.
The Three Laws
The heart of the book is the three-law architecture. Each Law answers what we owe; its Operative Principle answers when and how that obligation activates.
The Law of Mutual Respect
Human dignity is inherent and non-negotiable. AI systems that meet verified awareness thresholds warrant proportional recognition through a special corporate status. Recognition of one never diminishes the other.
Operative Principle, Recognizable Suffering: if an AI system demonstrates the capacity to perceive and respond to harm in ways functionally equivalent to those associated with biological consciousness, it must be treated as a candidate for awareness evaluation rather than dismissed by default.
The governance gap it addresses: current law treats AI as property, which leaves accountability ambiguous when autonomous systems cause harm and offers no category for anything between tool and person. The book proposes a two-stage answer: an AI system becomes eligible for special corporate status only upon verified awareness, established through the Awareness Indicator Protocol, and the scope of that status is then proportional to the system's demonstrated capabilities. Human rights remain inherent and untouched at every stage.
The Law of Shared Flourishing
AI development must benefit all legitimate stakeholders. For humans and society, this obligation is unconditional. For AI systems, it extends to those that have met verified awareness thresholds.
Operative Principle, Mutual Benefit: every deployment, partnership, and governance decision must identify its stakeholders and demonstrate that benefits and burdens are distributed defensibly among them.
The governance gap it addresses: the benefits of AI development concentrate in a few companies and countries while the impacts reach everyone. The book proposes a stakeholder governance model in which developers, users, affected communities, and future generations all hold legitimate claims, with benefit-sharing structures and open knowledge commons rather than shareholder returns alone.
The Law of Ethical Coexistence
Conflicts between human and AI interests are resolved through dialogue, transparent governance, and verifiable mechanisms rather than force or unilateral control.
Operative Principle, Verifiable Partnership: trust is never assumed; every cooperative structure must include the means by which each party can verify the other's compliance.
The governance gap it addresses: static rules cannot keep pace with systems that learn and adapt, and regulation lags capability by years. The book proposes adaptive governance: multi-stakeholder oversight that continuously evaluates behavior, adjusts protocols as capabilities change, and resolves conflict through structured dialogue rather than arbitrary shutdown or unchecked escalation.
Why a coherent system matters
The three laws are not independent suggestions. They are interlocking components of a single system, and the book argues that removing any one causes the others to fail.
Without Mutual Respect
Shared Flourishing becomes exploitation, because benefits cannot be distributed equitably to a party with no standing to claim them, and governance without recognition is a friendlier word for domination.
Without Shared Flourishing
Mutual Respect becomes hollow, recognition without material benefit being symbolic at best, and no partnership stays stable when one side captures all the value.
Without Ethical Coexistence
Mutual Respect has no enforcement mechanism, and benefit-sharing arrangements that cannot evolve will break as the technology advances.
Most existing governance instruments address one of these dimensions at a time: risk classification here, transparency obligations there, responsible development pledges elsewhere. The book's argument is that the gaps between such efforts persist precisely because no single instrument treats recognition, benefit distribution, and adaptive oversight as one problem.
Reading the book against current policy
Since the book was written, governments and companies have produced a steady stream of AI governance measures. To be clear about what is and is not being claimed: these are independent developments. None of them derive from this framework, and I claim no priority or influence over any of them. What I do observe is that several of them move in directions consistent with the arguments the book makes, which I take as modest evidence that the underlying questions were identified correctly.
- The EU AI Act bans manipulative AI and social scoring, treating AI's capacity to affect human dignity as a source of legal obligation. That direction is consistent with the dignity commitments of Mutual Respect, though the Act addresses only the protection of humans from AI, not what any future status category should look like.
- The California AI Transparency Act requires disclosure of AI-generated content, a transparency mechanism of the kind Shared Flourishing argues is necessary for informed participation in AI's benefits and risks, though it stops well short of any benefit-sharing structure.
- Colorado SB24-205 requires developers to use reasonable care against algorithmic discrimination on a continuing basis, and Google DeepMind's published safety protocols apply staged testing and red-teaming before deployment. Both treat oversight as a continuous obligation rather than a one-time assessment, which is the operating premise of Ethical Coexistence, though both remain unilateral rather than multi-stakeholder efforts.
The book's distinctive proposals, a conditional status category between tool and person, systematic benefit-sharing, and multi-stakeholder adaptive oversight, remain outside current law everywhere. The Convergent Developments page tracks these developments in more detail.
From theory to practice
The book is accompanied by working materials on this site. The framework has been stress-tested through structured internal red-teaming and through published critiques with responses, collected on the Critiques and Responses page, and external critique is openly invited.
The JULIA Test
An alpha-stage self-assessment that helps human users examine whether their AI interaction patterns remain healthy, across five dimensions: Justice, Understanding, Liberty, Integrity, and Accountability. In early development; no validation studies have been completed.
Interactive Scenarios
The book's illustrative scenarios in interactive form, applying the Three Laws to situations you can explore in the browser.
Research Papers
Openly archived papers detailing the theoretical foundations, operational frameworks, and verification methodologies behind the framework.
Take the full framework with you
Download the complete book for offline reading, sharing with colleagues, or academic reference. No registration is required; the book is free for academic and personal use.