Information Technology Reference
In-Depth Information
preferences did not correlate with the actual
choices made when games were actually played.
Experiments conducted on the visual percep-
tion of graphics have discovered very approximate
power laws relating the numerical measure to
the reported perceived measure (Tufte, 2001).
Because perceptions change with experience and
perceptions are context-dependent (Macdonald-
Ross, 1977) different people see the same areas
somewhat differently. For example, the reported
perceived area of a circle grows exponentially
more slowly than the actual (physical) measured
area.
1
. Even for something as simple as the length
of a line Asch (1956) asserted that the reported
perception depends not only on the context, but
also on what other people have already said about
the lines. The above results beg the question of
whether we need designs that correct for the visual
transformations of the average perceiver partici-
pating in the average psychological experiment
(Tufte, 2001). While we know that a satisfactory
answer to the above question is to use a table of
actual physical data, it is the absence of such data
that represents the biggest challenge. Even when
actual data is available, it is poorly correlated
with self-reported (perceived data; reviewed in
Lawless et al., 2009). Lawless and his colleagues
(2000) found no association in multiple regres-
sions between winners--losers in air-to-air combat
and topic knowledge of air combat maneuvering.
Carley (2002) concluded that collecting data
directly from humans changes their responses.
And in a meta-analysis of thirty years of data,
Baumeister and his colleagues (2005) found only
a negligible association between self-esteem and
actual academics or work performance. Similarly,
a discrepancy exists between the views of manag-
ers and the decisions made by their organizations
(Bloom et al., 2007). To explain these findings,
we have concluded that observational data (self-
reports, surveys, and interviews) from agents can
be independent of action data (e.g., individuals
who are walking, driving a car, or flying an air-
plane; similarly, organizations during business
meetings, during merger and acquisition activities,
and during the search for technology partners).
States of interdependence are common in or-
ganizations (Smith & Tushman, 2005), probably
because individuals multitask poorly (Wickens,
1992), but multitasking is the fundamental rea-
son why organizations exist (Ambrose, 2001).
However, organizational theory has failed, per
Pfeffer and Fong (2005), possibly because social
scientists have not yet accounted for the existence
of illusions and incorporated them in social theory.
Adelson (2000) gives us tools to use against
the theoretical problem presented by illusions.
If human agents construct visual representations
of their world with but only a small contribution
from sensory data, the rest split between grouping
and experience shaping processes, then cognitive
and social dissonance processes operate on these
representations to form world views independently
of a data driven reality, producing the large discon-
nects between self-reports and actual behaviors.
But the result is a profusion of illusions that must
be contended with. For example, the illusions
known as “risk perceptions” have not only driven
the recent “domino-effect” collapse of the financial
system in 2008-09 (e.g., the collapse and merger
of Countrywide Financial Corporation and Merrill
Lynch by Bank of America, J.P. Morgan Chase
taking over Bear Sterns, IndyMac collapsing, Fan-
nie and Freddie going into receivership, Lehman
Brothers filing for bankruptcy, and American
International Group struggling to recapitalize),
but they are more likely to occur under certain
decision making structures (consensus rules) or
during states of high interdependence (e.g., market
bubbles, bank runs, and panics) than others which
pursue instead risk determinations (scientific
methods; majority rules; in Lawless et al., 2008).
Search WWH ::
Custom Search