Netflix processes 500 billion events per day and still cannot tell why a viewer abandoned an episode. YouTube detects that you watched three political videos and floods your feed for weeks — because it reads behaviour without reading state. Spotify maps your mood from two billion listening sessions and adjusts in real time. Three platforms. Three different layers. One framework to explain the difference.
The data is there. The interpretation is not.
Three structural reasons explain why this gap persists. Actions are easy to measure. Internal states are not. The research that could help is scattered across fields that do not talk to each other. And business incentives favour optimisation over understanding.
Forrester found that 73% of companies struggle to use behavioural data for anything beyond basic segmentation. McKinsey found that companies excelling at personalisation generate 40% more revenue than average. The distance between those two numbers is the gap this framework addresses.
Read the full paperFour layers from action to motivation.
Most AI systems operate at Layer 1. A few reach Layer 2. Almost none attempt Layer 3 or 4. The framework gives you a map of where you are, and what moving one layer deeper looks like.
Built for the taxi, the desk, or the projector.
Open research. You can take part.
This work needs practitioner input. If you build AI products in customer experience, education, or anywhere AI meets users, your perspective sharpens the framework. The survey takes five minutes. The interview takes twenty.
19 May to 14 June 2026. Four weeks.
Activities run in parallel. This bar moves with today's date.
For when the conversation goes deeper.
Complete proposal
Long-form version with all sections, ethics, contributions, and references. Print to PDF supported.
Interview Protocol
15 semi-structured questions for expert practitioners. Mapped to RQ1, RQ2, and RQ3.
Survey Instrument
The formal instrument. 15 questions across four sections. Benchmarking, recognition, mapping, utility.
50 sources, connected
Every reference as an interactive node. Connections drawn between traditions and ABSD layers.
What we collect and why
Behavioural timing data, anonymity, participant rights, and how inference data is used and stored.