Hubert Dreyfus

Quick Facts

• Full name: Hubert Lederer Dreyfus
• Lived: 1929-2017
• From: United States
• Main home institution: University of California, Berkeley
• Main fields: phenomenology, existentialism, philosophy of mind, philosophy of AI
• Best known for: arguing that human intelligence is embodied, practical, social, and hard to reduce to explicit rules
• Major works: What Computers Can't Do, Mind Over Machine, Being-in-the-World, On the Internet

The Big Question

Can intelligence be explained as symbol manipulation, or does human understanding depend on having a body, habits, skills, and a shared world?

Dreyfus answered that human-style intelligence begins in practical involvement. We are already walking, talking, using tools, reading moods, and dealing with what matters. Thought grows out of this skilled contact with the world.

In One Minute

Hubert Dreyfus was an American philosopher who brought Martin Heidegger and Maurice Merleau-Ponty into debates about artificial intelligence.

Early AI often treated intelligence as storing symbols and applying formal rules. Dreyfus thought that missed how people understand ordinary situations. A person can enter a classroom, catch a ball, or find the right tool because body, training, culture, and concern already guide what stands out.

His target was not every use of computers. His sharper claim was that rule-based, disembodied models give a poor account of everyday human understanding.

What They Taught

Dreyfus taught that human beings are agents already involved in a meaningful world. This is the Heideggerian idea of being-in-the-world: we do not first meet neutral objects and then add meaning. We usually meet things as useful, dangerous, urgent, familiar, broken, or out of place.

Take a hammer, a phone, or a keyboard. When you use it well, the tool mostly disappears into the activity. Heidegger called this ready-to-hand: a thing shows up through use. Dreyfus used this against theories that start with a detached mind looking at objects from the outside.

He also stressed background practices: shared habits and unspoken know-how. In a cafe, you know roughly where to stand, how to order, how close to stand to strangers, and when the exchange is over. Nobody gives you a complete rulebook. You learn the pattern by living in a community.

This matters for AI because old-fashioned symbolic AI modeled intelligence as explicit representations plus rules. A representation is an internal stand-in, like "cups are containers." A rule says what to do, like "if a cup is full, keep it upright." Dreyfus argued that human understanding cannot be built from rules alone, because real life is open-ended. Intelligence includes knowing which facts matter now.

Dreyfus did not deny that people use rules. Beginners often need them. A new driver thinks through mirrors, signals, lanes, and braking distance. An expert driver responds more directly. The road, traffic, and car form one field of action. For Dreyfus, that is trained, embodied skill.

Key Ideas With Examples

• Phenomenology: the study of experience as it is lived. Dreyfus used it to describe ordinary action before theory.
• Being-in-the-world: humans are always involved with tools, people, purposes, and situations. A door shows up as something to open, knock on, lock, or hold for someone.
• Ready-to-hand: a tool fades into the task when it works. A musician does not inspect the guitar while playing; the guitar becomes part of the playing.
• Background practices: shared know-how that is rarely stated. In a library, people lower their voice without calculating a rule.
• Embodiment: thinking depends on bodily skill and perception. You understand "too steep" or "within reach" through a body that climbs and reaches.
• Skillful coping: absorbed, flexible action. A cook adjusts heat and salt by feel, not by checking a rule for every movement.
• Symbolic AI: AI that treats intelligence as symbol manipulation by formal rules. Dreyfus thought it worked best in narrow domains.
• The relevance problem: the problem of knowing what matters. Asked to bring the chair by the window, you ignore its molecule count and notice its location and use.

Major Works

• "Alchemy and Artificial Intelligence" (1965): an early RAND report arguing that AI optimism rested on hidden assumptions about mind, knowledge, and the world.
• What Computers Can't Do (1972), revised as What Computers Still Can't Do (1992): Dreyfus's famous critique of "artificial reason." It argues that disembodied rule systems cannot reproduce human background understanding and practical skill.
• Mind Over Machine (1986, with Stuart Dreyfus): presents the Dreyfus model of skill acquisition: learners move from rule-guided novice behavior toward expert responsiveness.
• Being-in-the-World (1990): a commentary on Division I of Being and Time. It explains Heidegger's account of world, tools, action, and everyday understanding.
• On the Internet (2001; revised 2008): applies Dreyfus's ideas about embodiment and risk to online life, arguing that distance and low commitment can weaken trust and responsibility.
• All Things Shining (2011, with Sean Dorrance Kelly): asks how people can find meaning after the decline of older religious and cultural authorities.
• Skillful Coping (2014): collects essays on perception, action, mind, and cognitive science.

Why It Matters

Dreyfus made a hard philosophical point concrete: human understanding is not just having correct inner descriptions. It is knowing how to move in a world that already matters to us.

His criticism helped open space for embodied cognition, situated robotics, and other approaches that tie mind to action, perception, and environment. Even critics often accept his warning: a system can pass narrow tests without the worldly understanding humans rely on.

He also mattered as a teacher. At Berkeley, he made Heidegger, Merleau-Ponty, Kierkegaard, and other difficult thinkers usable.

Proponents, Critics, and Opponents

Dreyfus's main allies were the phenomenologists. From Martin Heidegger, he took the idea that practical involvement comes before detached theory. From Maurice Merleau-Ponty, he took the bodily character of perception and action. He also fits with Gilbert Ryle, who distinguished knowing how from knowing that.

His main opponents were defenders of symbolic AI and formal models of mind. Early AI researchers such as Allen Newell, Herbert Simon, Marvin Minsky, and John McCarthy thought computers could model intelligence by representing facts and applying procedures. Dreyfus argued that this left out background know-how and embodiment. His criticism also pushes back against a broad reading of Alan Turing, where intelligence is treated mainly as successful test performance.

Critics said Dreyfus overstated what computers could not do and underestimated nonhuman ways of solving problems. Chess programs, search systems, and later statistical methods complicated the debate. His lasting point is narrower: task success does not by itself show embodied, situated understanding.

Related Pages

• Martin Heidegger
• Maurice Merleau-Ponty
• Alan Turing
• Gilbert Ryle
• Phenomenology
• Philosophy of Technology and AI
• Existentialism
• Being and Time
• Paul Feyerabend