Should We Still Make Decisions? The Human Decision in the Age of Artificial Intelligence by Éric Hazan and Olivier Sibony, published by Flammarion in February 2026, marks a turning point in French debate on decisional autonomy. This book arrives at a moment when Europe is refining its rules for AI regulation and when algorithmic transparency is becoming a democratic requirement.
The Authors
Éric Hazan, former Senior Partner at McKinsey & Company until 2025, now directs the impact fund Ardabelle Capital. He teaches digital strategy and AI at HEC and Sciences Po. Olivier Sibony has been a professor at HEC Paris since 2015 and Associate Fellow at Oxford. Co-author of the bestseller “Noise” with Daniel Kahneman and Cass Sunstein, he contributed to formalizing the field of “behavioral strategy,” exploring the impact of cognitive biases on strategic decisions.
The Essentials
- The central finding: we have entered a world where machines can make better decisions than we do in everyday situations — medical diagnosis, recruitment, risk of recidivism assessment
- The historical reference: Paul Meehl’s pioneering 1954 work on the superiority of statistical analysis over expert judgment, validated by 70 years of research
- Three decisional territories: delegation (AI decides), co-decision (AI helps structure), and conscious refusal to delegate even when the machine is better
- The political stakes: who decides, on behalf of whom, where, and according to what rules?
The Central Thesis: Mapping Decisional Autonomy
Hazan and Sibony analyze “the fields and conditions in which decision-making can be effectively delegated to artificially intelligent tools.” Their approach is neither technophile nor technophobe: “It analyzes, without naive optimism or reflexive technophobia, what we can delegate to machines — and what we must imperatively preserve as a human prerogative.”
Hazan proposes “a genuine mapping of decisions. Rather than being for or against AI, it proposes to identify different territories: those where delegation is already widely accepted, those where uncertainty dominates, and those that society may judge as non-delegable.” This “decisional geography” constitutes the book’s major conceptual innovation.
The Intellectual Lineage of Meehl
The authors build on the foundational work of Paul Meehl, who in 1954 “demonstrated how statistical analysis of objective data could match the judgment of experts when deciding on medical treatment, granting conditional release to a prisoner, or admitting students to a selective university.”
Meehl’s “Clinical versus Statistical Prediction” established that algorithmic methods systematically surpass clinical judgment, a conclusion that “remarkably well withstood the test of time for half a century.” This 1954 thesis constitutes “one of the first theoretical works laying the foundations for the use of statistics and computational modeling in research in psychiatry and clinical psychology.”
The Three Registers of Augmented Decision-Making
The book develops an operational typology: “The first is delegation: when AI decides in our place, provided that trust is built through processes of evaluation, testing, and auditing. The second is co-decision: AI does not decide, but helps to frame a problem, generate options, structure reasoning. The third concerns cases where we consciously choose not to delegate, even if the machine could do better.”
This framework materializes in specific examples: “algorithmic trading: algorithms make millions of decisions per second, where humans previously intervened. The speed is such that a computer located a few meters closer to a stock exchange server can offer an advantage in nanoseconds.” Conversely, in recruitment, “AI should not decide in your place. On the other hand, it can become a thinking partner. A good LLM tool will challenge you, ask questions you hadn’t thought of, push you to be more rigorous — as a good colleague would. Co-decision is this: an AI acting as a sparring partner in the decision-making process.”
The Paradox of Autonomy in Delegation
The book explores a fundamental tension: “deciding is a prerogative we associate with our autonomy, our dignity, our very role, particularly professional. This capacity to decide, as an individual or manager, is central to our identity. So when AI decides in our place, it touches something fundamental, what defines us.”
Yet, “the question is not whether to be for or against AI, but to determine where to delegate, where to co-decide, and where to refuse to delegate. And in this shift, one thing becomes clear: humans do not stop deciding, they decide differently.” This reformulation of autonomy constitutes the work’s most stimulating conceptual contribution.
The Blind Spots: The Anthropological Dimension
The book remains marked by its authors, experts in business strategy and behavioral science. While they mention “the growing risk of an algorithmic technocracy built on the use of tools whose power and functioning remain opaque” and call for “a collaborative reinvention of democracy and AI,” their analysis remains centered on decision-making efficiency rather than on profound anthropological transformations.
The question of trust, which is central in the philosophy of Paul Ricœur (“trust does not rest on blind faith, but on a shared intelligibility of actions and their reasons”), deserves more thorough development in the context of non-certifiable AI systems. Similarly, the impact on the formation of critical judgment, a major issue raised by other French thinkers on AI, is merely touched upon.
Why Read It
“Should We Still Make Decisions?” addresses all those seeking an operational analytical framework for decisional AI. “This is not a book about AI. This is a book about decision-making,” the authors warn. This pragmatic approach constitutes its strength: rather than philosophizing in the abstract, Hazan and Sibony offer conceptual tools to navigate a world where “the frontier of AI’s proven effectiveness constantly shifts.”
The book enriches French debate on algorithmic ethics by proposing an alternative to the prevailing moralism. Its typology of decisional territories could influence European regulation of decision-making algorithms, particularly in the high-risk sectors identified by the European AI Act. In a country where “France may be the only country teaching philosophy in high school,” this work builds a bridge between critical reflection and technological pragmatism.
Bibliographic Information: - Title: Should We Still Make Decisions? The Human Decision in the Age of Artificial Intelligence - Authors: Éric Hazan and Olivier Sibony - Publisher: Flammarion - Publication Date: February 25, 2026 - Pages: 240 pages