Game Development Reference
In-Depth Information
are perceived as situation-outcome expectan-
cies when the outcome is attributed to external
causes and are perceived as action-control and
action-outcome expectancies when the outcome
is attributed to internal causes. Internal causes are
defined as the learner's actions, circumstances,
skills and abilities. External causes are related
to another's actions, circumstances, skills and
abilities. Intrinsic-values and extrinsic-values are
relevant when determining the subjective value
of actions and outcomes. These values are related
to evaluating the activity and the outcome per se,
i.e. intrinsic, or evaluating them for other goals,
i.e. extrinsic.
Control-value theory is an integrative frame-
work of assumptions corresponding to other emo-
tional theories, such as expectancy-value theories
of emotion, theories of perceived control and
attributional theories of achievement emotions.
Therefore, this theory, rather than challenging the
OCC model (Ortony et al., 1990), incorporates
some of its assumptions. The theory classifies
'achievement emotions' into three types according
to their focus and time frame: prospective outcome,
retrospective outcome and activity emotions. The
basic assumptions of the control and value apprais-
als are summarized in Table 1. The prospective
outcome emotions arise from the expectations of
succeeding or failing. The retrospective outcome
emotions focus on the attribution of success and
failure and the activity emotions are determined by
the activity demands and the person's disposition to
participate, which are mapped to perceived control
over the activity and its value. It is important to
signal that 'achievement emotions' are the result
of the product of appraisals of control and value
and that if one of these is lacking no-emotion is
induced. In addition, achievement emotions are
domain dependent, e.g. the emotions arising in a
History domain will be different from the ones
arising in a Physics domain. The Control-value
theory has proven effective at recognizing the
learners' emotions in the English, German, Math
and Physics domains (Goetz, Frenzel, Pekrun,
Hall, & Lüdtke, 2007). The method used by
Goetz et al. (2007) is the self-report of emotions
using the Achievement Emotions Questionnaire
(AEQ), which was designed and validated through
Structural Equation Modeling (SEM) (Pekrun,
Goetz, & Perry, 2005).
Student Modeling and Probabilistic
Relational Models (PRMs)
Modeling the user's behavior and related data is a
task that involves uncertainty (Sucar & Noguez,
2008). In the learning context, there are still ques-
tions about how the learner achieves knowledge
and understanding. In addition, the personal dif-
ferences between learners, e.g. diversity between
attitudes, standards, beliefs and goals, make it more
difficult to ensure that the learner is experiencing
a determined emotion in a specific situation or
domain. Also, deciding which information has to
be taken into account to update the student model
represents a challenge. Therefore, as a solution to
this problem, Bayesian Networks (BNs) have been
applied to modeling. BNs are AI tools applied to
domains with inherent uncertainty. BNs are graphs
that represent causal relations between random
variables or events (Jensen & Nielsen, 2007) and
have been applied to building cognitive (Noguez
& Sucar, 2006, Muñoz, Noguez, Mc Kevitt, Neri,
Robledo-Rella, & Lunney, 2009) and affective
student models (Conati & Maclaren, 2009). The
main effort in designing BNs lies in the time that
must be invested in selecting meaningful informa-
tion from the knowledge domain and the resultant
complexity of the Bayesian Network (BN), i.e.
the number of random variables is large and each
node has more than three parents with several
states. To facilitate the design of student models
with BNs, probabilistic relational models (PRMs)
are used (Sucar & Noguez, 2008). PRMs are an
object-oriented representation of the knowledge
domain that can be easily transformed into a BN
or several BNs. Their representation facilitates
applying the model to several domains and the
Search Nedrilad ::




Custom Search