Geography Reference
In-Depth Information
in general (McGuiness, 1994) and it seems logical that one aspect of data visualization where
expert-novice differences would be greatest is in the handling of uncertainty.
However, domain expertise is just one of several dimensions of expertise that could be
considered in visualization design. For example, some individuals might have little expertise
in the particular problem domain but might have a great deal of experience making decisions
and solving problems under uncertain conditions. We might call these individuals decision
experts - they might include policy makers, corporate executives and administrators who are
frequently tasked with making difficult choices under less-than-ideal circumstances. While
they are not necessarily domain experts, it is hard to call them 'novices' in the typical sense of
the term. Another type of expertise that could play a role in determining appropriate design
strategies is experience with complex visualization displays. Individuals who are not dazzled
and overwhelmed by multiple detailed representations and sophisticated interaction will
likely respond to a presentation of uncertainty information differently than those who do
not encounter those types of displays regularly.
14.3.2 The influence of decision problem presentation
and decision frame
The presentation of a decision problem can influence the way a decision maker approaches
the problem and ultimately reaches a conclusion. For example, if model output is reported
to an analyst using a specific level of 'certainty' rather than a specific (equivalent) level of
'uncertainty', different decision outcomes could result. More generally, the simple inclusion
or omission of uncertainty information would also undoubtedly influence the decisions.
Similarly, if a decision is presented as choosing between specific alternatives (choose A, B
or C) compared with a ranking task (rank the alternatives from least to most favourable),
different decisions could result (Deitrick, 2006). As discussed above, some individuals are
able to consider statistical uncertainty more readily than others. Perhaps the design of
uncertainty representations should be in part governed by these factors.
For example, treating uncertainty as a discrete variable (a piece of data is either 'certain'
or 'not certain') is an approach that might be appropriate for one user group, while another
group might be more likely to require a representation that treats uncertainty as continuous,
with percentages and error ranges. Additionally, the representation of uncertainty to some
may be more sensible if the possible outcomes are represented as discrete scenarios ('either
this, this or this') rather than a statistical range of outcomes ('somewhere between this
and this'). These differences can be addressed in terms of the decision framework within
which an individual solves a problem. Interviews and cognitive walk-through methods with
potential users could allow researchers to identify those components of the representation
that are usable, the methods by which users reach decisions, and the most appropriate way
to communicate uncertainty to that user group.
14.3.3 The influence of differing levels of uncertainty information
control on decision making
The information-seeking literature suggests that individuals may vary (perhaps according
to their relative expertise) in how much information they can use, and how much control
Search WWH ::
Custom Search