1. Definition
Evidence & Friction is the bridge between what the Pyragogy Syllabus currently proposes and what the wider Pyragogy research ecosystem can challenge, qualify, or revise.
The roles stay distinct:
- Syllabus — the current map of concepts, risks, practices, and protocols;
- Cognitive Interview Protocol (CIP-KGE) — a human-gated route for turning expert knowledge into reviewable change proposals;
- UnPeeragogy — a friction lens for finding failure modes, counterevidence, boundary conditions, and gaps between declared practice and lived practice;
- Working Patterns — an evidence-aware map of organisational interventions, including a separate set of human–AI research candidates.
No layer is allowed to certify itself simply by pointing at another Pyragogy project.
2. Use Case
Use this layer when a Syllabus node makes a claim about a risk, intervention, mechanism, or desirable practice and a reader needs to know its current research state.
The question is not only “is there a source?” but:
- what exactly is being claimed;
- whether the evidence concerns the problem or the proposed solution;
- whether the intervention was actually implemented;
- what outcomes were observed;
- what it costs and who bears that cost;
- what evidence complicates the claim;
- what remains unknown.
3. Human Role
The human decides whether a source, incident, pattern, or interview actually bears on the node.
They remain responsible for distinguishing evidence from analogy, deciding whether a proposed Knowledge Patch should enter the graph, and preserving disagreement when the material does not justify a single conclusion.
4. AI Role
AI can retrieve related material, compare claims, identify possible tensions, separate observation from interpretation, and propose bounded changes.
It must not silently strengthen a claim, treat a research candidate as a validated practice, invent a missing source, or write a contested interpretation into the map as fact.
For interview-derived changes, the AI produces a proposal. A human reviews the diff.
5. Friction
Before a node is strengthened, the research loop should force a few uncomfortable questions:
What would make this claim wrong?
Do we have evidence for the intervention, or only evidence that the problem exists?
Is “human oversight” real authority or ceremonial approval?
Did the practice happen as described?
Which costs disappear from the success story?
What do we still not know?
The friction is not there to block revision. It is there to make revision more local, inspectable, and reversible.
6. Risk
The main failure mode is false synthesis: making several internal Pyragogy projects appear to corroborate one another when they are actually sharing assumptions, analogies, or research lineage.
Two safeguards matter in this pilot:
- UnPeeragogy is not direct evidence for an AI-learning claim unless a traceable case or source actually supports that relation. It may instead provide a method for looking for failure and counterevidence.
- Working Patterns AI records are research candidates. They may sharpen a question or suggest an intervention to test, but they are not validated answers.
7. Observable Markers
This layer is working when:
- a reader can tell whether a node is a supported claim, theoretical synthesis, design hypothesis, or unresolved question;
- cross-project links say why they matter rather than merely listing related pages;
- counterevidence and implementation failures remain visible;
- research candidates are visibly labelled as candidates;
- a proposed graph change can be reviewed before it is accepted;
- an interview can end with a clear account of what changed, what did not, and why.
The research loop
Syllabus
what do we currently propose?
↓
CIP-KGE
what new knowledge is being proposed?
↓
UnPeeragogy
where does the theory break or need qualification?
↓
Working Patterns
what evidence, alternatives, conditions, and costs are relevant?
↓
Human review
accept · revise · contest · reject
↓
Syllabus
change the map
↺The loop does not require every project to contribute to every node. An honest “no direct evidence identified here yet” is more useful than a decorative link.
Pilot nodes
The first small evidence-layer pilot is attached to three nodes:
- adult_reflective_practice — whether accountability checkpoints create meaningful review or ritual approval;
- automation_bias — separating the documented risk from unvalidated Pyragogy interventions intended to mitigate it;
- ai_over_reliance — distinguishing the broad over-reliance hypothesis from proposed high-friction remedies.
The pilot is intentionally small. If these links make the map easier to understand and challenge, the same pattern can later be extended to other nodes.
Pyragogy Interviews
The interview format is one way this loop can receive new human experience.
An AI Research Teammate can enter a conversation with a relevant Syllabus subgraph, ask for concrete incidents rather than generic opinions, preserve evidence and interpretation separately, and finish by proposing a reviewable Knowledge Patch.
The public output is not just a video. It can include a final question:
What changed in the map because of this conversation?
A proposed change remains a proposal until a human validates it.