Home
9:41
Predicting the Why · Brief
01 / 12
Applied Research Project · EMBA 26

Predicting
the why.

A behavioural intelligence framework for understanding human action in AI systems.

ResearcherArunamirtharaj C.
MentorHimmat Singh
IndustryAkshith Sharma · Brained
Swipe →
The hook

AI sees everything. It understands almost nothing.

When a student pauses for 30 seconds before a math problem, the system records the pause and moves on.

It cannot tell confusion from careful thinking, anxiety from strategic deliberation.

It tracks what. It misses why.

The scale

The data is there. The interpretation isn't.

500B+
Events logged daily by Netflix. The same scale at Spotify.
73%
of companies struggle to use behavioural data for anything beyond basic segmentation. (Forrester 2022)
Why the gap persists

Three structural reasons.

REASON 01

Actions are easy to measure. States are not.

Clicks capture automatically. Emotions don't show up in server logs.

REASON 02

The relevant research is scattered.

Psychologists and computer scientists rarely cite each other.

REASON 03

Incentives favour optimization.

Engagement KPIs reward time-on-platform over user goal achievement.

Literature gap

Four rigorous traditions. None talking to each other.

SDT — Psychology

Self-Determination Theory

Explains why: autonomy, competence, relatedness. Lacks real-time detection.

Affective Computing — HCI

Emotion detection

Decodes feelings from signals. Disconnected from motivation theory.

JITAI — Health

Just-in-Time Interventions

Detects vulnerability & receptivity windows. Limited beyond health.

Learning Analytics — EdTech

Clickstream patterns

Captures what users do. Skips why.

The framework

ABSD — four layers from action to motivation.

Each layer generates the one above; AI must work in reverse.

04
Drive
Motivational · Stable
03
State
Cognitive-emotional
02
Behaviour
Pattern · Temporal
01
Action
Observable · Discrete

Most AI sits at Layer 1. ABSD is the roadmap up.

Layer 01
01
Action
Observable · Discrete

What the user does. Discrete events tracked with high fidelity by almost every system today — and where most stop.

clickstimestampssubmissionsnavigationkeystrokes
Layer 02 — the bridge
02
Behaviour
Pattern · Temporal

How the user does it. Patterns emerging across actions — not single events. The bridge between what AI sees and what it needs to understand.

Hesitation
→ Uncertainty
Thrashing
→ Decision paralysis
Regression
→ Confusion
Acceleration
→ Boredom / overconfidence
Persistence
→ Growth orientation
Avoidance
→ Anxiety / helplessness
Flow
→ Optimal engagement
Layer 03
03
State
Cognitive-Emotional

What the user feels. Inferred — not directly observable. Confusion can resolve into engagement or spiral into frustration depending on system response inside that window.

confusionconfidencefrustrationflowanxietycuriosityboredom
Layer 04 — the deepest
04
Drive
Motivational · Stable

Why the user does it. The fundamental psychological needs that shape how users respond when states are challenged.

autonomycompetencerelatedness

A user with high competence drive persists through confusion. One with frustrated competence disengages.

Approach

Mixed methods. Anchored in CX.

METHOD 01

Literature synthesis

Across SDT, affective computing, JITAI, learning analytics.

METHOD 02

Expert interviews — 5

Semi-structured, 15–20 min. CX primary, EdTech secondary.

METHOD 03

Practitioner survey — 30–50

Descriptive breadth on signal recognition and framework utility.

METHOD 04

Case analysis — 2 systems

Illustrative ABSD applications. Public docs.

Timeline

Four weeks. 19 May → 14 June.

19 – 23 MAY
Lock instruments & outreach
Pilot survey, prospect 10–12 interviewees.
24 – 30 MAY
Interviews + survey live
5 interviews, one per day. Survey deployed.
26 MAY – 3 JUN
Case analysis (parallel)
Two AI systems through the ABSD lens.
1 – 8 JUN
Coding & synthesis
Cross-method refinement of the framework.
6 – 14 JUN
Draft, review, submit
Final on 14 June 2026.
Get involved

Open research.
Read. Take part. Share.

 Read the full paper at predictwhy.com
 Build AI products? Take the 5-min survey.
 Share with anyone working on AI & user understanding.