Self-Determination Theory explains why people engage. Affective Computing shows how to detect emotional states. JITAI research defines when to intervene. Learning Analytics maps what users actually do. Each tradition is rigorous within its domain. None of them talk to each other. The ABSD framework synthesises what they were pointing toward without knowing it.
What each tradition contributes. What each one misses.
| Tradition | What it explains | What it misses | ABSD contribution |
|---|---|---|---|
|
SDT
Self-Determination Theory
|
Why people engage. Autonomy, competence, and relatedness needs across thousands of validated studies in education, work, and digital systems. | How to detect need states from behaviour in real time. SDT is observational and longitudinal - not built for live inference from interaction data. | Drive layer with behavioural indicators. SDT provides the validated motivational constructs that make Drive inference principled rather than arbitrary. Layer 4 · Drive |
|
Affective Computing
Picard, 1997 onward
|
How to detect emotions from multimodal signals. State detection from text achieves 70–79% accuracy. Speech-based recognition now exceeds human baseline for certain categories. | Connection to motivational constructs. Cross-domain application. The field excels at detection but lacks principled frameworks for what states mean in terms of human drives. | State layer mapped to behavioural patterns. Affective Computing enables the Behaviour-to-State transition - the inference engine that ABSD depends on for Layer 3. Layer 3 · State |
|
JITAI
Just-in-Time Adaptive Intervention
|
When to intervene. Vulnerability windows and receptivity states. Meta-analyses confirm JITAIs outperform static interventions when timing matches user states. | Theoretical grounding in psychology. Signals beyond health behaviour. JITAI has a rich signal taxonomy but limited integration with motivational theory. | Behavioural signal taxonomy with state mappings. JITAI signals appear across domains - hesitation, avoidance, persistence - providing the cross-domain vocabulary Layer 2 needs. Layer 2 · Behaviour |
|
Learning Analytics
Siemens and Baker, 2012 onward
|
What users do. Clickstream data, temporal patterns, engagement trajectories. 78–84% accuracy predicting student dropout from behavioural patterns alone. | Why users do it. Wise and Shaffer noted that without theoretical frameworks connecting behaviours to underlying processes, analytics remains descriptive rather than explanatory. | Action-to-Behaviour pattern detection layer. Learning Analytics provides the proven methodology for temporal aggregation - the transformation that turns events into trajectories. Layer 1 → 2 · Transition |
What each field built. Where each one stopped.
Tap any tradition to expand its core claims, key works, limitations, and the specific gap that ABSD addresses.
Self-Determination Theory · Deci and Ryan, 1985 onward
Why people engage
Four decades of empirical work on the three basic psychological needs that drive human behaviour across every context where motivation matters.
Core claims
Three basic psychological needs drive human behaviour: autonomy (the need to feel volitional control), competence (the need to feel effective), and relatedness (the need to feel connected). When these needs are satisfied, intrinsic motivation and sustained engagement follow. When frustrated, people disengage or shift to controlled, externally driven motivation that produces lower quality outcomes.
Key works
Deci and Ryan (2000) - foundational statement of the three needs framework
Ryan and Deci (2017) - full theoretical integration across domains
Ng et al. (2012) - meta-analysis: need satisfaction predicts engagement across health contexts
De Vreede et al. (2021) - SDT applied to AI-assisted decision making
Peters, Calvo and Ryan (2018) - motivational design in digital experience
The gap ABSD closes
SDT limitation
SDT explains motivation rigorously. It does not tell you how to detect need states from live interaction data. The theory was built for longitudinal observation and self-report. ABSD operationalises SDT's constructs as the Drive layer - giving product teams a principled vocabulary for the motivational layer they have always designed around but never measured.
Affective Computing · Picard, MIT, 1997 onward
How to detect emotional states
Three decades of building computational systems that can recognise, interpret, and respond to human emotions from text, speech, and physiological signals.
Core claims
Computational systems can recognise and respond to emotion. Emotion detection from text achieves 70–79% accuracy in controlled settings. Speech-based recognition exceeds human baseline for certain emotion categories. D'Mello and Graesser showed that transitions between affective states during learning follow predictable dynamics - confusion can resolve into engagement or devolve into frustration, and the system's response during the confusion window determines which trajectory unfolds.
Key works
Picard (1997) - the founding text of the field
Calvo and D'Mello (2010) - interdisciplinary review of affect detection methods
D'Mello and Graesser (2012) - affective dynamics during complex learning
Ocumpaugh et al. (2014) - detecting states from behavioural indicators without physiological sensors
Poria et al. (2017) - multimodal fusion approaches
The gap ABSD closes
AC limitation
Affective Computing detects states. It does not connect them to motivational constructs or provide a cross-domain framework for what detected states mean in terms of human drives. ABSD takes the detection capabilities of Affective Computing and situates them within a principled motivational model - so that knowing a user is frustrated becomes actionable in terms of what they need, not just what they feel.
Just-in-Time Adaptive Intervention · Nahum-Shani et al., 2015 onward
When to intervene
A framework from mobile health research for detecting the right moment to deliver support - based on dynamically assessed vulnerability and receptivity states.
Core claims
Optimal intervention requires detecting two distinct states: vulnerability (when a person is at risk of negative outcomes) and receptivity (when a person is able to receive and process intervention). These are not the same. A user who is frustrated but still actively problem-solving may reject support they would have welcomed five minutes later. Meta-analyses confirm JITAIs outperform static interventions when timing matches user states.
Key works
Nahum-Shani et al. (2015) - pragmatic framework for building health behaviour models
Nahum-Shani et al. (2018) - key components and design principles for mobile JITAI
Klasnja et al. (2015) - microrandomized trials for developing JITAIs
Wang and Miller (2020) - meta-analytical review of JITAI effectiveness
Spruijt-Metz et al. (2015) - computational models for behaviour change
The gap ABSD closes
JITAI limitation
JITAI has a rich vocabulary for intervention timing and a proven signal taxonomy that predicts vulnerability and receptivity across domains. What it lacks is theoretical grounding in motivational psychology. A hesitation pattern means the same thing in a student and a patient - but the intervention each needs is shaped by their Drive, not just their State. ABSD provides that grounding.
Learning Analytics · Siemens and Baker, 2012 onward
What users actually do
A field dedicated to measuring, collecting, and reporting data about learners and their contexts - and demonstrating that behavioural patterns contain predictive information.
Core claims
Behavioural patterns contain predictive information that action counts alone do not. Studies using the Open University Learning Analytics Dataset achieve 78–84% accuracy predicting at-risk students from clickstream data. Predictive features include not just action counts but temporal patterns: spacing of study sessions, time-of-day variations, navigation sequences, and help-seeking behaviours. The data exists. The interpretation framework does not.
Key works
Siemens and Baker (2012) - defining the field at the intersection of analytics and learning
Baker et al. (2004) - gaming-the-system behaviours predict poor outcomes
Kuzilek et al. (2017) - Open University Learning Analytics Dataset
Li, Baker and Warschauer (2020) - clickstream data and self-regulated learning
Wise and Shaffer (2015) - why theory matters more than ever in big data
The gap ABSD closes
LA limitation
Learning Analytics proves that behavioural signals matter. Ferguson noted the field risks optimising for measurable proxies rather than actual learning. The field has called for integration with learning sciences and psychological theory. ABSD answers that call - not just for education, but for any domain where AI mediates human decision-making. The Action layer and Action-to-Behaviour transition are built directly on what Learning Analytics has proven.
Each tradition provides a specific capability at a specific layer transition.
The integration gap is not accidental.
The literature synthesis for this study found minimal cross-citation between motivational psychology and affective computing, between JITAI and learning analytics, and between Self-Determination Theory and AI system design. Zawacki-Richter et al. (2019) found that 62% of AI in higher education research is led by computer science and STEM researchers, with only 8.9% from education departments. Researchers in each field cite within their tradition but rarely build on adjacent work. This creates redundant discovery of similar phenomena using different vocabulary, and missed opportunities for theoretical synthesis.
ABSD addresses exactly this. Each tradition contributes a specific methodological capability to a specific layer transition. Together, they form a complete inferential architecture.
ABSD addresses exactly this. Each tradition contributes a specific methodological capability to a specific layer transition. Together, they form a complete inferential architecture.
01
Action
Bridged by Learning Analytics
Temporal aggregation. Individual events become trajectories. Proven methodology for the first transition.
02
Behaviour
Bridged by Affective Computing
Probabilistic inference. Behavioural patterns weighted against contextual modifiers to produce state distributions.
03
State
Bridged by SDT
Longitudinal pattern analysis. How users characteristically respond to states reveals their underlying drives.
04
Timing
Calibrated by JITAI
Vulnerability and receptivity windows. The same accurate inference, at the wrong moment, produces a negative outcome.