CONVERSATION
Understand how algorithmic power governs society and how to hold it accountable
While much of the public debate remains focused on whether AI systems are truly intelligent, when they may surpass human intelligence, or how useful or dangerous they may become, the deeper issue is being overlooked.
Today’s most powerful algorithmic systems increasingly reassemble powers that political thought once deliberately separated under the doctrine of Trias Politica.
The book argues that compute, energy, training operate as an algorithmic executive; that code, model design, optimization targets, and embedded constraints function as an algorithmic legislature; and that evaluation, auditing, and benchmarking function as an algorithmic judiciary. What would be intolerable for many, the fusion of rulemaking, execution, and judgment, is becoming the standard of our society enabled by algorithmic systems.
They now shape access, opportunity, and justice at machine speed – yet they operate largely beyond democratic scrutiny.
Trias Algorithmica: What Code Rules, by Hamilton Mann, a Thinkers50 honouree, originator of the concept of Artificial Integrity, and leading authority in AI transformation, traces where this power originates, how it sustains itself through manufactured dependence, and what a constitutional framework for algorithmic governance must look like.
The book introduces original concepts including aggregogenous monopoly, Technological Stockholm Syndrome, dependautonomy, and unpacks the revaluing mechanism of work. It also presents a practical framework of nine critical questions for interrogating algorithmic power deployment.
“A fascinating assessment for the age of AI. Who decides, on what basis, and backed by what kind of legitimacy? Everyone who cares about democracy should pay attention.”
— SIMON JOHNSON, Ronald A. Kurtz Professor of Entrepreneurship, head of the Global Economics and Management Group at MIT Sloan; 2024 Nobel Laureate in Economic Sciences; co-author of Power and Progress
"Algorithms do not merely push out the production possibility frontier. They reallocate control rights among citizens, silently and at scale. That an algorithm can serve as rule-maker, rule-executor, and referee does not mean it should. But absent deliberate democratic resistance, this is where we are heading. If you doubt it is already underway, consider the surveillance state you enter each morning when you open your web browser. Trias Algorithmica anatomizes this pernicious algorithmic creep and offers tools to address it."
— DAVID AUTOR, Economics Department Head, Daniel (1972) and Gail Rubinfeld Professor, Google Technology and Society Visiting Fellow, Margaret MacVicar Faculty Fellow at MIT Department of Economics; Co-author of The Work of the Future
"I build the systems this book is about. Perception stacks that decide what a car sees. Robots that act in physical space. Foundation models that increasingly mediate how people work and think. From inside that work, Mann's argument lands with uncomfortable accuracy. The temptation in our field is to treat governance as a layer added late — a compliance review, a red team, a policy document downstream of the model. Mann's claim is that by then it is already too late. The constitutional choices have been made: in the objective function, in the deployment surface, in who gets to approve a release and who gets to overturn it. Engineering decisions are constitutional decisions wearing technical clothing. This is not a comfortable book for builders. It should not be. The most consequential systems of the coming decade will be judged less by what they can do than by whether their power can be answered for. Mann gives us a vocabulary, and an architecture, for that answering. I do not agree with every step of the argument. I am glad the argument exists."
— AMNON SHASHUA, Founder, President and CEO at Mobileye; Sachs Professor of Computer Science at the Hebrew University of Jerusalem’s Rachel and Selim Benin School of Computer Science and Engineering
"Trias Algorithmica: What Code Rules makes an engaging argument that AI systems possess tremendous power without the legitimacy traditionally conferred by separation of powers. The metaphor that the same AI company typically carries out the rule-making (legislative), deployment (executive), and monitoring (judicial) provokes one to question how much the behavior of modern neural networks results from legislative design by AI engineers as opposed to the human behavior the AIs learn from."
— DE KAI, AI Professor; Author of Raising AI
"In this fascinating book, Mann argues that opaque, unaccountable algorithmic systems — like social media platforms, AI chatbots, or predictive policing — play three governance roles at once: they make policy, they execute the laws they've created, and they distribute justice according to their own notions of fairness. In the name of technological efficiency, we have private systems of quasi-government gaining increasing control over our lives. Mann argues for taking back control and adding much needed friction and oversight to these systems. A compelling approach to restoring accountability in algorithmic governance."
— CATHY O’NEIL, Mathematician, Founder of ORCAA; Author of Weapons of Math Destruction and The Shame Machine
"Mann comprehensively applies a separation of powers framework to critique current AI and prescribe its better development and use. Trias Algorithmica demonstrates that no AI system should consolidate the power to set goals, to execute them, and to judge whether they have been executed well. This is remarkably creative work that sheds new light on AI governance."
— FRANK PASQUALE, Professor of Law at Cornell University; Author of The Black Box Society
"Trias Algorithmica reframes AI governance around a simple democratic principle: people must retain the right to question, contest, and reshape the systems that shape them. Hamilton Mann's call for algorithmic separation of powers is a timely contribution to a more accountable digital future."
— AUDREY TANG, Taiwan's Cyber Ambassador; Co-author of Civic.AI and Plurality.net
“Trias Algorithmica is a necessary book for anyone trying to understand how AI is changing power itself. Hamilton Mann brings a rare combination of technological fluency and moral urgency to one of the defining questions of our time: how societies can preserve accountability, human agency, and legitimacy as decision making shifts from human to machine intelligence.”
— SINAN ARAL, David Austin Professor of Management, Marketing, IT and Data Science at MIT; Author of The Hype Machine
"Mann offers a critical warning: As algorithms learn to govern, humanity must relearn how to constrain power. For seventy years we have asked whether machines can think. Hamilton Mann argues we've been asking the wrong question. The one that matters now is whether machine decisions can in fact be legitimate. Trias Algorithmica reframes AI as a constitutional problem. When platforms write policy as a unitary legislature, judiciary, and executive they fuse power into a single, frictionless stack. Their AI algorithms exercise this power at superhuman speeds. Mann then proposes a key separation of powers as a pushback toward individual autonomy."
— MARSHALL VAN ALSTYNE, Allen and Kelli Questrom Professor in Information Economics at Boston University Questrom School of Business; Co-author of Platform Revolution
"As AI systems increasingly shape what we see, choose, and become, the right to question machine-mediated decisions should become central to democratic life. Trias Algorithmica is an ambitious call to build transparency, appeal, and human judgment into the architecture of AI itself.”
— NITA FARAHANY, Robinson O. Everett Professor of Law and Philosophy at Duke University; Author of The Battle for Your Brain
"Hamilton Mann’s Trias Algorithmica is an ambitious and thought-provoking contribution to the growing conversation about governance, accountability, and power in the age of artificial intelligence. At a time when leaders are rethinking not only technology strategy but also the institutional logic underlying intelligence-driven systems, Mann challenges readers to consider how algorithmic systems increasingly shape decision-making, oversight, and human agency. The book raises important strategic and philosophical questions that business and policy leaders can no longer afford to ignore."
— GEOFFREY PARKER, Charles E. Hutchinson ’68A Professor of Engineering Innovation at Dartmouth; Co-author of Platform Revolution
"As artificial intelligence increasingly decides who gets a job, a loan, or a fair trial, we are drifting into a society where our civil rights are dictated by opaque, proprietary code. For years, we have distracted ourselves with the question of whether machines can think, missing the more urgent reality: machines are now deciding. They are fusing the rule-maker, executioner, and judge into a single algorithm. Trias Algorithmica makes the danger unmistakable. Algorithmic governance is not a technological shift but a constitutional crisis, and preserving our autonomy in the AI age will require what silent optimization has stripped away: friction, due process, and a new separation of powers. An essential and timely read."
— VIVIENNE MING, CEO and Founder at Socos Labs; Visiting Professor at UCL Global Business School for Health; Author of Robot-Proof
"Hamilton Mann’s Trias Algorithmica is a bracing intervention into the crisis of legitimacy stalking our automated age. Mann identifies that our current technological acceleration represents a silent fusion of legislative, executive, and judicial powers within the black box of the code. By elegantly resurrecting Montesquieu for the era of generative AI, he exposes how the disappearance of procedural friction that is marketed as an efficiency feature but alas embeds bug of tyranny. Essential reading for anyone who believes that liberty cannot survive without the right to interrupt the machine."
— NILS GILMAN, Deputy Editor of Noema Magazine and Editor-in-Chief of Berggruen Press; Former Associate Chancellor at UC Berkeley