Adaptive Control of Thought

What is Adaptive Control of Thought?

Adaptive Control of Thought (ACT) is John R. Anderson’s family of models for explaining cognition, memory, and learning. Its later form, ACT-R (Adaptive Control of Thought–Rational), is a cognitive architecture: a set of components and rules researchers use to build models of how people perform tasks.

How it works

The architecture distinguishes two kinds of knowledge:

• Declarative knowledge: Facts and experiences represented in units called chunks. Retrieving a chunk depends on its past use, the current context, forgetting, and competition from other memories.
• Procedural knowledge: Skills represented through production rules, or condition–action pairs. These resemble if–then statements that select what to do in a particular situation.

Learning can change how readily knowledge is retrieved and which rules are selected. Production compilation can combine several steps into a new rule. ACT-R also uses specialized modules that can work concurrently, while the central production system selects one rule at a time.

Anderson and colleagues (2004) describe how these parts fit together.

Why it matters

A cognitive architecture lets researchers turn an explanation into a working model and compare its predictions with observed performance. ACT has applications in cognitive research, artificial intelligence, education, and human–computer interaction. Its value in a particular task comes from what the model explains and predicts, rather than how comprehensive the architecture sounds.