The problem worth investigating
Flat documents retrieve related words but miss prerequisite and dependency structure.
This innovation is treated as a falsifiable engineering hypothesis. The goal is not to prove that an advanced technique is impressive; it is to determine whether it produces a measurable improvement over a simpler control.
Experimental architecture
- 01Entity registry
Instrumented independently so the experiment can reveal which stage contributes value. - 02Typed edges
Instrumented independently so the experiment can reveal which stage contributes value. - 03Prerequisite graph
Instrumented independently so the experiment can reveal which stage contributes value. - 04Article/topic mapping
Instrumented independently so the experiment can reveal which stage contributes value. - 05Project concept mapping
Instrumented independently so the experiment can reveal which stage contributes value. - 06Graph traversal
Instrumented independently so the experiment can reveal which stage contributes value.
Algorithms under study
Graph traversal
Hypothesis. Graph traversal is included because it addresses a specific measurable part of the system rather than being added as decoration.
Risk. The component must be compared with a simpler baseline and removed if it adds complexity without measurable value.
Evidence. Task-specific quality metric, latency, reliability and failure-case analysis.
Centrality signals
Hypothesis. Centrality signals is included because it addresses a specific measurable part of the system rather than being added as decoration.
Risk. The component must be compared with a simpler baseline and removed if it adds complexity without measurable value.
Evidence. Task-specific quality metric, latency, reliability and failure-case analysis.
Similarity edges
Hypothesis. Similarity edges is included because it addresses a specific measurable part of the system rather than being added as decoration.
Risk. The component must be compared with a simpler baseline and removed if it adds complexity without measurable value.
Evidence. Task-specific quality metric, latency, reliability and failure-case analysis.
Prerequisite path search
Hypothesis. Prerequisite path search is included because it addresses a specific measurable part of the system rather than being added as decoration.
Risk. The component must be compared with a simpler baseline and removed if it adds complexity without measurable value.
Evidence. Task-specific quality metric, latency, reliability and failure-case analysis.
What data the experiment needs
- Topics — recorded with enough context to reproduce and audit the result.
- Concepts — recorded with enough context to reproduce and audit the result.
- Articles — recorded with enough context to reproduce and audit the result.
- Projects — recorded with enough context to reproduce and audit the result.
- Questions — recorded with enough context to reproduce and audit the result.
- Relations — recorded with enough context to reproduce and audit the result.
How CortexLab decides whether the idea survives
- Path relevance
- Internal-link quality
- Prerequisite accuracy
Results should be compared against a control, segmented for failure cases and repeated across enough observations to avoid promoting noise into product behavior.
What can go wrong
Wrong reward
Optimization can improve the metric while making the actual experience worse.
Overconfidence
Small or biased samples can make experimental gains look more certain than they are.
Distribution shift
A policy that works on past users may degrade as topics, traffic and behavior change.
Next research milestones
- 01Graph embeddings
- 02Community detection
- 03Automated relation proposals
- 04Human graph editor
Questions still open
- Which simpler baseline must this beat before deployment?
- How should uncertainty be calibrated and communicated?
- What evidence would make us reject the idea?
- How do we prevent reward hacking or accidental optimization of engagement alone?
- Which decisions must remain human-reviewed?