These notes are a summary of concepts presented in “Affect-awareness framework for intelligent tutoring systems.”
Landowska, Agnieszka. (2013). Affect-awareness framework for intelligent tutoring systems. 2013 6th International Conference on Human System Interactions, HSI 2013. 540-547. 10.1109/HSI.2013.6577878.
- Definition and Purpose of Affect-Aware Systems
- Software programs that recognize user emotions
- Control mechanisms handle affect-related data to optimize interactions
- Handling Uncertainty in Emotional Recognition
- Sources of uncertainty
- Fuzzy nature of emotions
- Insufficient algorithmic accuracy
- Instability of emotions (frequent changes)
- Limitations of representation models
- Frameworks to manage uncertainty
- Probability
- Fuzzy set
- Evidence
- Possibilities
- Interval analysis
- Rough sets
- Certainty factor quantifies algorithm confidence in results
- Sources of uncertainty
- Components of Affect-Aware Systems
- Affect Recognition
- Video
- Voice
- Text
- Physiological measurements
- Multimodal input schemas ensure robustness
- Algorithms
- Gaussian Processes
- Support Vector Machines (SVMs)
- Affect Recognition
- Interpretation
- Six basic emotions (e.g., joy, sadness, fear, etc.) as combinations (Ekman’s model)
- Discrete Models
- Dimensional Models
- Whissel Wheel
- PAD model by Russel and Mehrabian
- Whissel Wheel
- Trustworthiness of recognition and classification
- Emotional stereotyping and user affect modeling
- Dimensional Models
- Affect-Aware Reaction:
- Affect-aware control mechanisms for interventions
- Multimodal responses based on behavior libraries
- Emotional states classified into action-triggering and non-triggering subsets
- Recognition and Representation Mechanisms
- Emotion Recognition Algorithms
- Multimodal approaches combining multiple data inputs
- Conditions
- Input collectability in the application environment
- Output compatibility with chosen representation models
- Certainty information
- Intervention Strategies and Mechanisms
- Affective Intervention
- Modifies standard control flow for counterproductive emotional states
- Intervention scenarios include emotional state and system reaction pair
- Affective Intervention
- Decision-Making:
- Considers hypothesis certainty to minimize disruptive interventions.
- Multimodal execution (text, audio, video)
- Patterns and Analysis
- Emotional reaction patterns influenced by
- Previous experiences
- Task type
- Neural system differences
- Analysis opportunities:
- Timeline trends (burn-down curves).
- Pattern recognition in emotional responses to stimuli and tasks.
- Emotional reaction patterns influenced by
- Recognition Enhancements:
- Lexical analysis: Keyword spotting, smiley detection, punctuation, capitalization heuristics.
- Keystroke dynamics
- Physiological response
- Decision Frameworks
- Dempster-Shafer theory of evidence
- Goal-Question-Metric method for system evaluation