Information Technology Reference
In-Depth Information
but the study did not formally compare these two modes. Only Parikh et al. (2001) specifically
studied the mode of guidance, comparing predefined guidance with dynamic guidance and find-
ing that dynamic guidance significantly increased decision quality and user learning while signif-
icantly decreasing decision time. The study did not confirm the hypotheses that users of dynamic
guidance would be more satisfied or would take less total time to make their decisions. Given the
limited empirical testing of this dimension, refining it at this time has no basis.
Invocation Style: Automatic, On-Demand, or Hybrid
Two of the empirical studies required users to request guidance; the others provided it automati-
cally. This difference is likely to affect outcomes, both because guidance on demand may in many
cases not be demanded and because guidance given automatically might irritate more than it
guides. Invocation style therefore seems a desirable dimension to add to the typology. Indeed,
some guidance mechanisms may lend themselves to one invocation style more than the other. One
can also conceive a hybrid style that combines automatic and on-demand invocation. For instance,
invocation might be made “switchable,” allowing users to choose for themselves between auto-
matic and on-demand guidance. Or invocation might be primarily on-demand, with certain con-
ditions triggering the guidance automatically.
Timing: Concurrent, Prospective, or Retrospective
Perhaps the greatest surprise in the empirical studies is the number of cases where the guidance
was delivered as feedback after users had made their choices rather than as guidance prior
to choosing. Following the lead of Limayem and DeSanctis (1993) for GDSS and Jankowski
(1997) for CASE tools, distinguishing these two types of guidance makes sense. Dhaliwal and
Benbasat (1996) made a similar distinction in studying explanations for knowledge-based sys-
tems. Limayem and DeSanctis (1993), as well as Dhaliwal and Benbasat (1996), used the termi-
nology feedforward and feedback, but since feedforward and feedback have formal definitions
that not all guidance may satisfy, alternative phrases are preferred here. Guidance that applies to
a choice at the point of interaction is termed concurrent guidance, since it is provided concur-
rently with the choice point, whereas guidance that refers back to a choice already made is termed
retrospective guidance.
Why would a system designer build retrospective guidance into a system? If a user is per-
forming a task repeatedly (Huguenard and Ballou, 2001), feedback on what was just chosen can
guide subsequent choices. In other cases (Limayem and DeSanctis, 2000; Montazemi et al., 1996;
Antony et al., 2004), the guidance may be corrective, giving users a chance to change their
choices. These observations suggest that the timing of guidance merits further study and is worth
including in the typology.
Concurrent guidance would seem to be the most timely guidance. One can imagine situations
where the guidance might occur even earlier. Guidance for a series of inputs might be provided at
one time when a tool is first invoked. Guidance for planning a series of activities would guide
users in advance of actually invoking and using each tool. Situations such as these, where guid-
ance is provided significantly prior to the choice point, constitute a third case of timing: prospec-
tive guidance.
The original typology had a dimension distinguishing the scope of the guidance: short-range
versus long-range. This is the only dimension of the original typology that none of the decisional
guidance studies addressed. This dimension can now be dropped, since it is covered by the new
Search WWH ::




Custom Search