iScale: Studying Long-Term Experiences through Memory - Modeling Users' Experiences with Interactive Systems - page 109

Study 2 tested the effectiveness of graphing with the two different versions of
iScale against a control condition, which allowed for the direct reporting of experi-
ences, without any form of graphical representation. Overall, we found that the con-
structive iScale provided better assistance than the the value-account iScale tool in
the reconstruction process. When using the constructive iScale, participants elicited
approximately 50% more experience reports than in the control (no-graphing) con-
dition, were more likely to recall cues referring to a discrete event, or ones referring
to temporal information such as when the experience took place, and were more
consistent across multiple recalls in estimating the time when each experience took
place. Moreover, contrary to our expectations, participants graphed patterns were
more consistent in the constructive than in the value-account condition. These find-
ings support the idea that graphing, through imposing a chronological order in the
recall process (Anderson and Conway, 1993), supports the reconstruction of the
context in which experiences took place, thus forming stronger temporal and se-
mantic links across the distinct experiences (Kahneman et al., 2004).
While iScale is promising, one crucial aspect must be kept in mind. There is a
likely discrepancy between experiences elicited through longitudinal field studies
composed of records of moments (e.g., through experience sampling, Hektner et al.,
2007) and retrospective data elicited from memory through iScale. Retrospective
reconstructions cover long periods of time and, thus, systematic biases, such as the
overemphasis on especially salient moments, are likely to occur (Kahneman, 1999;
Bartlett, 1932). The current studies provide no insights into how these retrospec-
tions differ from the actual experiences and future work should inquire in those
differences with actual longitudinal studies.
In this chapter we argued that veridicality may not be as important as the test-
retest consistency of the recall process, because people actually communicate and
act upon their own biased memory and not on an unbiased objective summary of
what actually happened. In supporting design, understanding what users remem-
ber may be more important than what they actually experienced. Still, designers
are not always interested in users memories. Often, the actual, and not the remem-
bered, experiences should be at the forefront. Consider, for instance, the case where
we might want to know the reasons that underly non-responsible driving behavior.
Memories offer little understanding as to what motivates such behaviors. Retro-
spective techniques are not aimed to replace longitudinal field studies and in-situ
methods. Instead, we propose that retrospective techniques may be a viable alter-
native to longitudinal studies when memories are placed at higher importance than
actuality.
Next, in our study, we assessed users' test-retest consistency in recall using two
different pieces of information: the exact time that an experience took place, and the
overall graphed pattern. These two served to indicate the two different kinds of re-
called information: episodic and value-charged (i.e., affect). The former, we thought,
would signify the effective recall of a substantial amount of contextual cues from
episodic memory. We expected that estimating this temporal information would be
an error-prone activity. Thus, a more effective recall would have a strong impact on
users' consistency in time estimation, through the presence of more contextual cues.