![](bg3.png)
state;
• Motor expressions that communicate reactions
and emotional and behavioral tendencies; and
• Physiological reactions that act to regulate the
system, determining the activation of
neuroendocrine processes (related to the nervous and
endocrine influences) such as heart rate, skin
conductance, blood pressure, respiration, and pupil
dilation (Shami, 2008).
Our approach considers a set of methods that allows
designers to identify the user’s subjective feelings as
reported by the user as well as cognitive appraisals
and motor expressions derived from the participation
of evaluators.
3 EMOTIONAL EVALUATION
In the literature, it is possible to identify methods,
techniques and tools with which to assess emotions
in humans. Each of them has features characterizing
them more appropriate for certain aspects of
emotions. Generally, the instruments applied in the
methods, techniques and tools can be classified as
verbal or non-verbal. In this research, we consider
an instrument as verbal when the user explicitly
verbalizes what s/he is feeling.
According to Desmet (2003), verbal instruments
enable users to express their emotion in scales (when
a user says “I am very happy” or “I am not anxious
at all”) and to report “mixed” emotions as tension.
However, they are difficult to apply across cultures
because it may not be easy to translate emotions into
words. On the other hand, non-verbal instruments
can be considered discreet and independent of
culture and language. However, they can be
subjective as they generally use universal symbols
such as pictograms.
Figure 2 presents a taxonomy, which classifies
emotional assessment metrics as verbal or non-
verbal. Each final node in the taxonomy represents
one of the five components of Scherer’s model, and
each parent node represents a set of available
methods, techniques, tools and instruments that can
be used to measure a component.
Cognitive assessments are linked to the
interpretation of a situation and further development
of emotions. This component can be measured by
the Geneva Appraisal Questionnaire (GAF) (Geneva
Emotion Research Group, 2010), the Think-aloud
method (Someren et al., 1994) and the Subjective
Discourse Analysis (Lefevre and Lefevre, 2005).
Although the Subjective Discourse Analysis
considers spoken statements, this technique was
classified as non-verbal because users do not
explicitly say what they are feeling. The evaluators
should interpret the statements spoken during the
user’s interaction and classify the related emotion.
According to Scherer (2005) and Desmet (2003),
there is no objective method capable of measuring
subjective feelings. It is necessary to query the user,
and thus, the methods involve self-assessment. Most
of the methods for evaluating subjective feelings are
non-verbal, such as the SAM (Self-Assessment
Manikin) (Lang, 1985), Emocards (Reijneveld et al.,
2003), and Preemo (Desmet, 2003). A verbal
instrument is the Affect Grid (Russell, 1989).
The motor expressions are related to facial
movements, body gestures as well as to some
characteristics of speech as speed, intensity, melody
and sound. Methods that can be applied include the
Facial Action Coding System (FACS) (Ekman et al.,
2002), the Ten Heuristics of Emotion (Lera and
Domingo, 2007) and electromyography.
Physiological reactions are non-verbal and can
be measured by electrocardiogram, respiration rate,
electrodermal activity, electromyography,
pupillometry, etc. Physiological reactions allow
designers to evaluate the user’s emotional responses
in an experimental context once the users
spontaneously and unconsciously reveal their
emotions (Cristescu, 2008); (Yusoff and Salim,
2010). However, most of these evaluations require
expensive instruments and are intrusive and complex
(Axelrod and Hone, 2008); (Cristescu, 2008).
Finally, behavioral tendencies are also non-
verbal and generally are evaluated by performance
indicators, such as the time required to complete a
task, the accuracy of reaching a goal, the number of
errors and the number of creative’ ideas during the
interaction (Mahlke and Mingue, 2008).
4 A HYBRID APPROACH
Aiming to minimize the detection of false positives
in the emotional evaluation of information systems,
this work proposes a hybrid approach based on the
emotion model described by Scherer (1984).
Considering the evaluation methods and instruments
presented in the literature, we have selected a subset
that matches the five components of the Scherer’s
model. This selection considered the main
stakeholder (user or specialist) responsible for the
final result of each method or instrument, aiming to
balance the final emotional assessment of the
information system.
AHybridEvaluationApproachfortheEmotionalStateofInformationSystemsUsers
47