Cognition: Mental Representations, Problem Solving, and Decision Making - Foundations for Designing User-Centered Systems - page 187

marketing and for maintenance, although you are still likely to value the mug in
hand more than the mug next to you.
6.4.4.6 Learning and Feedback
Dawes ( 1994 ) argues that decision making cannot improve without feedback. In
his book, House of cards, he presents numerous examples of professionals some of
whom do and some who do not get feedback about their decisions. Those that
receive feedback have a chance of improving and usually do. Those that do not get
feedback do not improve, although they do become more confident with practice,
as they presumably get faster at it. He examines a wide range of tasks, but par-
ticularly takes psychotherapists to task, as very often in the past they provided
treatment without receiving feedback on the outcomes of their treatments.
So, if you would like your users to improve their performance and to learn, you
need to provide feedback. A failure to do so may only allow them to increase their
confidence in their ability to do the task without increasing performance.
6.4.5 Larger Scale Decision Making Process: Expertise
and RPDM
Decision making in the real world is also influenced by learning, mental models,
and expertise. Work by Simon and his colleagues looked at how expertise influ-
enced decision making. They showed that, in some cases, expertise led to faster
decisions, but essentially the same decisions; for example, young business majors
would make the same decisions for a business case, but would take longer (Simon
1997 ).
They also studied how people played chess as an analogue for other decision
making tasks (Chase and Simon 1973 ; Gobet and Simon 1996b ; Simon and Chase
1973 ). They found good chess players would consider more reorganization of the
positions and better possible moves than would novices, who would have to do
more problem solving. This expertise would also allow good players to remember
positions more readily because they were not remembering all the pieces indi-
vidually, but recognizing groups of pieces as patterns in a way similar to how
ABC, 747, and HFE (and others noted in Table 5.1 ) might be considered as single
objects instead of three separate characters (Gobet and Simon 1996a ).
This approach has been implemented in computational models, including
programs that solve problems (Ritter and Bibby 2008 ) and play chess (Gobet et al.
1997 ). These models start out with effortful problem solving or problem solving-
like behavior that, with practice, becomes faster and uses recognition to drive its
behavior. Sometimes the strategy does not change with improved learning (it is
just faster), and sometimes the ability to recognize what to do changes how the