Learning and Declarative Knowledge
John Anderson, a Psychology professor at Carnegie-Mellon University, wrote about the two distinctions between declarative and procedural knowledge in his ACT production theory of the unitary theory of cognition. It borrows ideas from Newell's symbolic framework (1972). The ACT production system proposed a distinction between procedural knowledge and declarative knowledge.
In 1983, Anderson provided a fuller description of the ACT and developed a theory called ACT*. This, in turn, evolved into the ACT-R (Atomic Components of Thought) theory (1993), in which an architecture of cognition is modeled to explain how the process of acquisition can be tuned to the statistical structure of the environment.
Defining
Declarative knowledge is knowing “that” (e.g., that Washington D.C. is the capital of America), as opposed to procedural knowledge is knowing “how” (e.g., how to drive a car).
Declarative knowledge is further divided into:
- Episodic knowledge: memory for “episodes” (i.e., the context of where, when, who with etc); usually measured by accuracy measures, has autobiographical reference.
- Semantic knowledge: Memory for knowledge of the world, facts, meaning of words, etc. (e.g., knowing that the first month of the year is April (alphabetically) but January (chronologically).
Procedural Overlap
When training complex cognitive skills, what we are normally saying is that there will be “procedural overlap” — the skills and knowledge that have been compiled while practicing in the learning environment are applicable back on the job. That is, we are assuming that a transfer from the learned task to the new task will be positive in that the underlying set of “productions” overlap. A production is a set of conditions-action pairs (if-then) and are the building blocks of procedural knowledge.
The trouble with this approach is that experts use declarative knowledge when problem solving — proposition models in which highly complex schemata are built from:
- Plans (simple): How goals, which are distinguished by an artifact are related to time and space.
- Concepts (simple): The representation of a class of objects, events, or other entities by their characteristic feature or mental image.
- Principles (simple): How one change is related to another change.
- Causal Network (complex): A combination of principles and concepts that are linked to each other with cause and effect and/or natural-process relationships.
- Goal-Plan Hierarchies (complex): Goal and/or plan structures that are linked to each other with non-arbitrary relationships.
- Conceptual model (complex): Concepts that are linked to each other with non-arbitrary relationships, such as a concept map.
So in reality, we expect the learners to transfer a simple production system into a highly complex cognitive schemata system. But they simply have nothing to build upon. We know that learning, innovation, creativity, and all the other things that go into creating complex knowledge are best accomplished by constructing a scaffold that allows one to connect various contexts such as relationships, physical tools, and mental tools. Yet we give them a very short step ladder (the production) that can be used to change a light-bulb, however, what they really need is a scaffold that will allow them to reach several stories (declarative knowledge).
The main reasons are resources and expectations. In college, it is downright difficult, if not impossible, to train complex cognitive skills in a semester; yet look what most problem solving courses in the corporate training world are -- a couple of hours, eight hours top. And then we expect the learners to transfer what they have learned in the classroom to the job. Yet, all they have are a very few simple if/then statements to take back to the job. Such training is totally inadequate.
Part of the problem is that we too often view the training process as cybernetics — a communication theory that treats organisms and organizations as being very much alike in that both display behavior. And we expect this cybernetics training process to behave in such a way that it steers us to a desired destination by treating “ways of behaving.” Now this would not be too bad except for one minor point. . . we look for no feedback. Thus we steer this training ship of ours towards a destination, yet on the way we take no compass or sextant readings. We simple do not where we are at any point during our trip, yet we expect this cybernetics organization of ours to get us to where we are going.
And the reason we blindly steer this ship of ours is that our expectation is not a result that will fix or improve a process, but rather a “completion” that tells us we have finished a specific cycle of training. Thus what we should be doing is looking for results and taking measurements along the way to ensure we are indeed getting the result or impact that we desire; rather than simply performing training for the sake of doing it.
Notes
Schemata (Schema): A mental model of a person, object or situation. Schema include cognitive maps (mental representations of familiar parts of one's world), images, concept schema (categories of objects, events, or ideas with common properties), event scripts (schema about familiar sequences of events or activities) and mental models (clusters of relationships between objects or processes).
Also see Artifacts.
Reference
Anderson , J. R. (1976). Language, memory, and thought. Hillsdale, NJ: Erlbaum.
Merrienboer, Jeroen (1997). Training Complex Cognitive Skills: A Four-Component Instructional Design Model for Technical Training. Englewood Cliffs, NJ: Educational Technology Publications.
Newell, A. (1972). Human Problem Solving. Englewood Cliffs, NJ: Prentice-Hall.