predictwhy · com
Home Read Paper Take the survey →
At a glance · 90-second read

The paper, at a glance.

If you only have a minute — this is the proposal in four questions and one framework, in plain language.

What

A framework for AI that reads why, not just what.

We propose ABSD — a four-layer model that helps AI systems move from tracking what people do to understanding why they do it.

Four layers: Action, Behaviour, State, Drive. Surface to depth. Each layer is a clearer view of the user.

Why

Today's AI sees everything, understands almost nothing.

Netflix logs 500 billion events daily. Yet when a customer abandons a cart, no system can tell sticker shock from a baby crying in the background.

That gap is the lost business — silent churn, missed upsells, students who quietly disengage long before the dropout shows.

How

Stack four lenses. Read up the stack.

Behaviour patterns (hesitation, thrashing, regression) sit between the raw click and the human emotion driving it.

Read those patterns — and the system can infer confusion, frustration, flow — and respond with the right move, in real time.

Where

Any product where AI meets people.

Empirically anchored in customer experience and education — the most validated domains.

Designed to transfer: finance (anxiety-driven trades), healthcare (right-moment intervention), SaaS (silent churn).

The framework in business terms

ABSD, without the academia.

Each layer is just a question your product should be able to answer. The deeper you go, the more your AI is actually working for the customer rather than at them.

Layer
What it is
What you ask
What you do with it
04
Drive
The underlying need
"What do they actually want?" Their motivation. Why they're using the product at all.
Design loyalty programs and onboarding around the real reason people come back — competence, control, connection. Not just price.
e.g. a learner who needs to feel capable, not just entertained.
03
State
The mood right now
"How are they feeling?" Their current emotional state. Confused, confident, frustrated, in flow.
Trigger the right intervention at the right moment — a tooltip, a human handoff, a simpler interface — before frustration becomes a churn event.
e.g. detect rising frustration in a support chat → route to a human.
02
Behaviour
The pattern in their actions
"How are they doing it?" The rhythm and shape of their actions over time. Not what they clicked — how they're clicking.
Catch the warning signs early. Hesitation, re-checking, rapid switching — these patterns predict confusion days or weeks before performance drops.
e.g. spot a user repeatedly going back to onboarding → they're stuck.
01
Action
The clicks & taps
"What did they just do?" The raw events. What every analytics dashboard already shows you.
Standard analytics. Funnel reports. Conversion tracking. This is the floor, not the ceiling — most teams stop here.
e.g. "page views were down 8% week-over-week."

Most products today read Layer 1. The opportunity — and the next decade of AI — is in reading up the stack.

The whole paper, in one sentence

Build AI that reads why people do what they do — and you build products that don't just react to customers, they understand them.

Where to next

Now you've got the shape of it.

If you'd like to read deeper, take part, or share with someone who'd find this useful — pick a door.

5 min read

On-the-go brief

Same content with more substance — phone-first and scrollable. The proposal in five minutes.

Open brief →
~5 minutes · live

Take the survey

If you build AI products, your view on these signals shapes the framework. Anonymous, instant save.

Take survey →
~25 min · academic

The full paper

The complete proposal with literature review, ethics, contributions, references. Print-ready.

Read paper →