Cognitive Science: The Study of Mind, Brain and Intelligence

What Is Cognitive Science?

Cognitive science is the interdisciplinary study of mind and intelligence. It brings together psychology, neuroscience, artificial intelligence, linguistics, philosophy and anthropology to understand how we perceive, remember, think, learn, communicate and act. Modern cognitive science investigates everything from the neural circuits underlying vision to the algorithms behind machine learning and the cultural influences that shape human reasoning.

The field took shape during the cognitive revolution of the 1950s, when psychologists, linguists and early computer scientists rejected the behaviourist idea that the mind was off-limits to science. The term “cognitive science” itself dates from the 1970s, and the Cognitive Science Society was founded in 1979. Our history of cognitive science traces the full story.

Core Questions Cognitive Science Asks

Cognitive science is defined less by a single method than by a shared set of questions. Among the most important:

• Perception: how does the brain turn light, sound and touch into a stable world of objects? (See attention and perception.)
• Memory and learning: how is experience stored, why do we forget, and why do some study methods work far better than others? (See memory and learning.)
• Language: how do children learn a language so quickly, and how do we understand a sentence in a fraction of a second? (See language processing.)
• Reasoning and decisions: how do people solve problems, and why do sensible minds make systematic errors? (See problem solving and decision making and biases.)
• Development: what is present at birth and what is built through experience? (See cognitive development.)
• Consciousness: why is there something it is like to have experiences at all? (See consciousness.)
• Minds and machines: can a machine think, and what do AI systems reveal about human intelligence? (See AI and cognitive science.)

Levels of Analysis: Marr’s Three Levels

One reason cognitive science needs several disciplines is that a mental process can be explained at different levels. The vision scientist David Marr, in his book Vision (1982), proposed three levels that are still the standard way to organise explanations:

• Computational level – what and why: what problem is the system solving, and why is that the right thing to compute? For vision, one problem is recovering the 3D shape of objects from a flat retinal image.
• Algorithmic (representational) level – how: what representations and step-by-step procedures carry out the computation? Psychologists and AI researchers usually work here, for example by modelling how edges are detected and grouped.
• Implementational level – with what hardware: how are those procedures physically realised in neurons, synapses or silicon? This is where neuroscience comes in.

The levels constrain each other but are not interchangeable. Knowing which neurons fire during a memory task does not by itself tell you what the memory system is for, and a good algorithm is useless as a theory of the mind if the brain could not possibly implement it. Much of the field’s progress comes from connecting findings across levels.

Key Ideas in Cognitive Science

Representation and computation

The founding idea of cognitive science is that thinking involves mental representations (internal states that stand for things in the world) and computations that transform them. Early versions, inspired by digital computers and by Allen Newell and Herbert Simon’s work on problem solving, treated the mind as a manipulator of symbols and rules. From the 1980s, connectionism and neural networks offered an alternative in which knowledge is stored as patterns of connection strengths across many simple units — the ancestor of today’s deep learning.

Embodiment

Classical cognitive science sometimes treated the body as a mere input and output device. Embodied cognition argues instead that thinking is shaped by having a body that moves and acts in an environment: perception is tied to action, abstract concepts often borrow from bodily experience, and some cognitive work is offloaded onto tools and surroundings. How far this changes the classical picture is still debated.

Prediction

A more recent unifying idea is that the brain is fundamentally a prediction machine. According to predictive processing, the brain continually generates expectations about incoming sensory signals and updates its internal model when those expectations are wrong. The framework aims to explain perception, attention, action and learning with one principle, and links naturally to Bayesian models of cognition and to machine learning.