Cognition: Memory, Attention, and Learning - Foundations for Designing User-Centered Systems - page 143

In these types of problems the problem solvers (and computer models) start with
domain knowledge and the goals for which they are trying to find answers. Novices
work backward from the goal. In the case of physics problem solving, the task is to
derive a value of a variable (like final speed) given some known variables (like
mass, initial speed, and force) and some equations (like Force = Mass 9 Accel-
eration). Novices tend to work back from what they are trying to solve for final
speed, chaining inference rules together until they find known variables. If they
need to compute the final speed then they look for an equation that computes speed.
Then they look for more equations to find the variables in the first equation, until
they bottom out with known variables. This is known as backward chaining or
bottom-up reasoning.
As the reasoning terminates in the answer, how to apply the rules in a forward
sense, without search, is learned. More expert behavior is a mix of these
approaches. With even more practice experts can reason in a forward sense. ''If I
have speed and time then I have acceleration, then if I have acceleration I
have…..'' The answer in this case is found more quickly and with less effort. This
is known as forward chaining or top-down reasoning.
This type of reasoning difference is seen in computer repair and trouble-
shooting, software usage, and other domains where formal reasoning can be
applied. Better interfaces will support the novice by providing the appropriate
domain knowledge needed to learn the inferences, and support the novice and
expert by providing the state information to reason from.
Learning can also occur in the opposite direction, from (implicit) skills back to
knowledge about these skills, but this learning is less well understood. Implicit
learning can occur when the user is working at the skill level, with knowledge
eventually being derived on the cognitive level through the observation of one's
own behavior. This type of learning, which is an active area of research in psy-
chology, often leads to or arises from strategy changes.
Whichever way the user learns, some key phenomena survive:
• The ability to recognize correct/incorrect items comes before the ability to
generate correct items.
• Knowledge is not acquired in an all-or-nothing way. Novices go through a stage
of fragile knowledge, trusting that what has been acquired is correct, whereas
sometimes it may be incorrect.
• Experts acquire a rich repertoire of representations of their knowledge. It is not
only that experts know more than novices; what they know is much better
organized, understood to a greater depth, and more readily available.
5.4.2 Improvements from Learning
Perhaps the biggest regularity of human behavior in general and users in particular
is that the changes due to learning leads to them getting faster at performing a task
the more they do it. Similar curves showing decreases in performance times have