Page 119 - Deep Learning

102 Creativity

To summarize, spread of activation through long-term memory serves
to activate at least three types of prior knowledge: mental representations of
actions, schemas for reasoning and processes that break down goals into sub-
goals. This list is not exhaustive but highlights important and well-documented
types of operations in analytical thinking. For the purpose of explaining insight,
it is essential that retrieval is selective. Due to the limited capacity of working
memory, not everything can be retrieved at once; only memory nodes that rise
above an activation threshold enter working memory.

Heuristic Search
The problem solver’s initial mental representation of the problem (the situation
and the goal) and the set of cognitive operations initially retrieved from mem-
ory define the space of solutions that can be reached via analytical thinking,
“the sector of exploratory activity” in Robert S. Woodworth’s phrase. A. Newell
35
and H. A. Simon called this a problem space, but their term suggests a space
of problems while what is meant is a space of potential solutions. I prefer the
term solution space. The solution space is a theoretical construct; it does not
exist in any explicit sense. It is a way of referring to all the solutions that are
possible, given the problem solver’s initial perception of the problem and the
repertoire of actions and operations initially retrieved.
A problem solution consists of a path through the solution space, a
sequence of cognitive operations that transforms the initial situation into a
situation in which the goal is satisfied. In a familiar task environment, the per-
son already knows which step is the right one at each successive choice point.
However, in unfamiliar environments, the person has to act tentatively and
explore alternatives. Analytical problem solving is difficult because the size
of a solution space is a function of the number of actions that are applicable
in each situation – the branching factor – and the number of actions along
the solution path – the path length. The number of problem states, S, is pro-
portional to b , where b is the branching factor and N the path length. S is
N
astronomical for even modest values of b and N, so solution spaces can only be
traversed selectively. By projecting prior experience onto the current situation,
both problem perception and memory retrieval help constrain the options to
be considered to the most promising ones.
The search through the solution space is a cyclic process. The problem
solver considers the current situation, thinks of one of more promising steps,
chooses one of them, performs the chosen one and evaluates its outcome; the
cycle repeats. Anticipating the outcomes of actions, drawing inferences and