Page 131 - Deep Learning

114 Creativity

According to the redistribution theory, this subjective experience is illusory.
The solution is always constructed piecemeal, but under some circumstances
the construction is so rapid and effortless that it leaves little or no trace
in consciousness. It is possible to specify those circumstances with some
precision.
If a problem solver could extend his look-ahead – his mental search
through the space of possible solutions – all the way to the goal, then there
would be no further uncertainty, no more false starts and no dead ends. If he
can see the entire path to the goal in his mind’s eye, he merely needs to carry
out the actions on that path in the physical solution space. Trial and error
thinking – look-ahead – preempts trial and error action.
But mental look-ahead is limited by working memory capacity.
46
Cognitive psychologists have not yet reached a consensus about the nature
of this capacity limit. The alternatives include a finite amount of activa-
47
tion, interference, rapid decay, small storage space and an inability to inhibit
distracting information. From the point of view of the theory of insight, the
important point is that capacity is limited (somehow). Mental look-ahead
cannot illuminate a large region of the solution space; it only reveals the
immediate vicinity of the current problem state. For novices working in an
unfamiliar solution space, the horizon of mental look-ahead is likely to be
only a few steps away.
What happens when one or more previously unheeded options come to
mind depends on the distance between the state in which the insight occurred
and the goal state, taken in relation to the horizon of mental look-ahead. The
variable of interest is the number of steps required to reach the goal from the
problem state in which the impasse is resolved. If working memory capacity
allows look-ahead to extend a maximum of n steps into the search space, and
the goal is more than n steps away, then mental look-ahead cannot reveal the
entire solution. In this case, the insight is followed by continued analytical
problem solving, complete with backups, errors, search, uncertainty and per-
haps further impasses. In this case, there is no subjective feeling of seeing the
entire solution in the mind’s eye, only relief at being able to think of a new
approach. This is partial insight.
If, on the other hand, the goal is fewer than n steps away from the state in
which the impasse was resolved, then the remaining path to the goal fits within
the horizon of mental look-ahead. Because n is small – at most half a dozen
steps – the construction of the new path in the mind’s eye can happen quickly.
The problem solver will experience the resolution of the impasse and the con-
struction of the path to the goal as a single event. This is full insight. The illusion