Shen, L., Wang, M.,
& Shen, R. (2009)
Affective e-Learning
using "emotional" data to improve learning in pervasive learning
environment
Educational technology
& Society, 12, 2, 176-189
p 176 'Pervasive leaning is
supported by wireless communication and wearable computing'
p 176, 177 Affective computing ' computer methods that are related to, derived from or deliberately
designed to influence emotions. It involves two areas: emotion synthesis
used to artificially imitate some of the physical or behavioural explosions
associated with affective states , and emotion analysis which is often employed in decision making for
interactive systems' e.g. synthesis (robots) analysis (monitoring).
Theoretical
background
' according to Fowler (1997) study, the relationship between
learning performance and the arousal is a type of inverted -U curve, and people
learn best when their emotions are at a moderate optimal arousal level'
Circumflex model (arousal, valence)
OCC model - cognitive appraisal model for emotions. 22 emotion
categories to goals, relevant events, actions of an agent, attitudes of
attractive or unattractive objects..........
Kort's Learning spiral model a state of evolution model Kort,
Reilly and Picard (2001). p 180 proposed a four quadrant learning spiral model
in which emotions change while the learner moves through quadrants and up the
spiral. In quadrant 1 the learner
is experiencing positive affect and constructing knowledge. At this
point, the learner is working through the material with ease. Once
discrepancies arise between the learner's knowledge he moves to Q2 which
consists of constructive knowledge and negative effect. Affect as confusion. If this is not resolved
learner may move to Q3 , frustration. If any misconceptions get discarded then
moves onto Q4 'unlearning and positive affect' then propelled back to Q1
looking for new learning.
In Affective computing detection rates for non verbal cues is
about 80% therefore
' because physiological signals are more difficult to conceal or
manipulate ....... And potentially less intrusive to text and measure, they are
a more reliable representation of inner feelings '
Affective e-Learning model
Use two sets of data
1.
Collected
Biophysical data HR, Skin conductance, BP, EEG. Incorporates elements from
Circumflex - to describe biophysical in terms of learner's emotions ( interest, engagement, confusion, frustration, boredom, hopefulness,
satisfaction, disappointment)
2. OCC cognitive
appraisal model for emotions.
considers contextual information of the learner and the learning
setting. Infers learner emotions from this data
So in this model 1 represents the emotional states and 2 the
appraisal mechanism . Assumes appraisal leads to emotions which in turn are
reflected biologically. A Baysian method was used to model the relations
between the emotional states and their causal variables.
Method
One subject, 20 trials, 40 minute biophysical recordings while
student studies for examination in otherwise ,normal context. Self paced
learning
EEG 3 electrodes, PCz, A1,A2.
Proposed Outcome of the system: generates responses to the learner p181
' based on his/her emotional states, cognitive abilities, and learning goals.'
This model can also
be used to customise the interaction between the learner and learning system,
to predict learner responses to system behaviour, and to predict learner's
future interaction with the learning system'
Results
Tested user satisfaction by comparing the number of user adjustments required for the
'emotion aware' system v a 'non
emotion aware system' latter required more users interventions
Adding EEG moved best classification rate ( for emotional
involvement) from 68.1 to 86.3%. P 187 ' this suggests the close relationship
between brain waves and emotions during learning'
