ABSD Framework Research
Predicting
the Why
AI systems see everything. They understand almost nothing. A framework for closing the gap.
01 / 12
predictwhy.com
The problem
73% of companies. Stuck at basic segmentation.
The data exists. Netflix processes 500 billion events per day. The problem is interpretation - not volume.
73%
Stuck at basic segmentation · Forrester 2022
40%
More revenue from good personalisation · McKinsey 2021
500B
Events per day · Netflix · still no state inference
02 / 12
predictwhy.com
Core claim
AI systems are optimised to predict what users will do. Not why they are doing it.
The gap has a structure. That structure can be named. ABSD names it.
03 / 12
predictwhy.com
The framework
Four layers. Each one beneath the last.
L4
Drive
Why. Competence, autonomy, relatedness.
L3
State
What users feel. Confusion, flow, frustration.
L2
Behaviour
Patterns over time. Hesitation, regression.
L1
Action
Observable. Where most AI operates today.
04 / 12
predictwhy.com
The key insight
Drive generates State. State produces Behaviour. Behaviour manifests as Action.
AI systems observe Action and must infer upward. Most never try. Confidence degrades: Action 96% → Behaviour 82% → State 70–79% → Drive longitudinal.
05 / 12
predictwhy.com
Theoretical foundations
Four traditions. Never previously connected.
Self-Determination Theory
Grounds the Drive layer - competence, autonomy, relatedness
Affective Computing
Grounds the State layer - 70–79% accuracy from interaction data
JITAI Research
Grounds the Behaviour layer - vulnerability and receptivity windows
Learning Analytics
Grounds the Action layer - 78–84% dropout prediction from patterns
06 / 12
predictwhy.com
The bridge between layers
Seven signals. All already in your data.
Hesitation
Pause before action · uncertainty
Thrashing
Rapid switching · overwhelm
Regression
Returning to earlier content · consolidation
Acceleration
Speeding through · boredom or overconfidence
Persistence
Continuing after failure · high drive
Avoidance
Early exit, minimal engagement · anxiety
Flow
Steady progress · optimal state
07 / 12
predictwhy.com
Critical qualification
The same signal means different things in different contexts.
A 30-second pause before a complex financial trade is prudence. Before a simple form field it is confusion. Three modifiers determine which: domain context, user context, and temporal context. Apply all three before acting on any signal.
08 / 12
predictwhy.com
Case study · CX
Netflix. 500B events per day. Layer 1 only.
When you pause a film, Netflix logs the event. It cannot determine whether the pause was anticipation, confusion, or discomfort. Action: 95%. Behaviour: partial. State: minimal. Drive: absent. The signals that would enable Layer 3 inference are in the data. The framework to read them is not.
09 / 12
predictwhy.com/cases
Case study · EdTech
Khan Academy. Mastery tracking. Without mastery understanding.
A student who answers correctly after 45 seconds of hesitation and one who answers in 4 seconds both receive the same credit. VanLehn (2011): AI tutors underperform human tutors precisely because they lack this psychological understanding. That gap remains accurate today.
10 / 12
predictwhy.com/cases
Open research · live now
53 references. Live survey. Expert interviews.
50
References across 4 traditions
14+
Survey responses - practitioners
5
Expert interviews - CX and EdTech
11 / 12
predictwhy.com/findings
Take the next step
Does this resonate with your experience?
The survey takes 5 minutes. The voice interview 20. Your participation shapes this research directly.
predictwhy.com/survey
12 / 12
predictwhy.com