Computational Reflections

A groundbreaking critique of the philosophical foundations of computing.

Computational Reflections is a wholly original investigation into the philosophical foundations of computing. The author's lifetime of work is distilled into this volume, which explores what it means to compute.

The standard theoretical foundations of computer science address the fundamental concept of “mechanism” but almost completely ignore the crucial role of “meaning” in any computational practice. Cantwell Smith takes the reader through these missing foundational gaps, including a historical analysis of why the field has reached its current state.

Despite its lack of treatment of semantics (“meaning”), computer science has borrowed and adopted semantic vocabulary to refer to mechanistic concepts, thus aggravating confusion within and without the field, especially so in philosophy, cognitive science, and contemporary artificial intelligence. This book’s arguments help to illustrate why computer science theory has almost nothing to say about “computation in the wild”—or the real-world practice of programmers and engineers who design the software and devices that we all use. The author argues that a true account of computation must do justice to the incredible complexity juggled by programmers in creating software that works, and he offers not only criticism but also directions for a successor account of computing.ENDORSEMENTS

“Brian Cantwell Smith’s provocative, carefully reasoned reflections on computing are indispensable for computer scientists, philosophers, and any others aspiring to think seriously about the issues raised by the increasingly consequential uses of digital-computational systems.”
—Joseph Rouse, Hedding Professor of Moral Science, Wesleyan University; author of Articulating the World and Social Practices as Biological Niche Construction

“Cantwell Smith exposes a core problem of the computer-saturated, AI-driven world: confusing symbols with reality. Generative AI does not represent lived meaning but produces coherent hallucinations from tokens, which limits its suitability for roles requiring genuine understanding, care, and responsibility.”
—Terry Winograd, Professor of Computer Science, Emeritus, Stanford University; Codirector of the Stanford Human–Computer Interaction GroupBrian Cantwell Smith was Professor of Information and of Philosophy as well as the Reid Hoffman Professor of Artificial Intelligence and the Human at the University of Toronto. His books include On the Origin of Objects and The Promise of Artificial Intelligence (both MIT Press).