Skip to main content

The Multi-Model Mind: Meta-Rationality for Wardley Leaders

· 4 min read
Dave Hulbert
Dave Hulbert
Builder and maintainer of Wardley Leadership Strategies

Your AI safety team wants to pause deployment. Your product team sees competitive risk in waiting. The map shows a component that is custom-built and changing quickly. Which model should lead the decision? Usually, not just one. Wardley Doctrine already points in this direction with Use Appropriate Methods: avoid one fixed approach to delivery, governance, or mapping. This is where meta-rationality becomes useful. It is the habit of noticing when a method no longer fits and switching lenses without dropping discipline. Charlie Munger called this a "latticework of models." David Chapman describes it as meta-rationality: combine frames when reality refuses to fit one frame neatly.

Why meta-rationality matters now​

  • AI multiplies frames. Product, legal, ethics, and safety teams often interpret the same capability differently. Meta-rational leadership keeps those views in play long enough to make a better call.
  • Wardley Maps can ossify. Maps are situational, not sacred. Meta-rational practice helps leaders avoid treating a clean map as final truth.
  • Doctrine is context-sensitive. Principles like Use Appropriate Methods and Think Small Teams ask teams to pick methods for the current landscape, not repeat a ritual from last quarter.

A meta-rational stack for mappers​

1) Name the frame​

Before mapping, declare which lens you’re using as the primary lens (e.g., evolution and user need). Admit which perspectives you’re temporarily ignoring (e.g., power dynamics, regulatory posture).

2) Carry a latticework​

Pair the map with at least two other models that stress different dimensions. Examples: Cynefin for complexity class, OODA for tempo, Viable System Model for governance recursion, Panarchy for cross-scale shocks.

3) Switch modes deliberately​

Use experiments or incidents as prompts to change frame: if the map stalls, view the situation as a Panarchy adaptive cycle; if coordination fails, run a VSM audit; if risk spooks delivery, move to Cynefin’s exploratory probes.

4) Tolerate overlap​

Accept that models will disagree. Capture the contradictions in the map notes instead of hiding them. When models contradict each other, you’ve found leverage, not confusion.

5) Retire brittle models​

Chapman’s meta-rationality stresses knowing when a model no longer fits. If a component’s behaviour routinely violates your chosen frame, pivot the model instead of forcing the data.

Applying the latticework to doctrine​

  • Use Appropriate Methods → Frame checks. In cadence reviews, ask: Which model dominated our last decision? What would change if we applied a different frame? Rotate the dominant model based on signals, not habit.
  • Optimise Flow → Tempo-aware switching. Use OODA loops to decide when to tighten or loosen control on AI components, while Wardley Maps show which components can safely accelerate.
  • Be Transparent → Annotated maps. Record which models shaped each strategic play. Transparency makes it easier to revisit assumptions when signals shift.
  • Design for Evolution → Model sunsets. As components commoditise, retire bespoke models and move to simpler governance. Genesis work may need rich narrative models; utilities can rely on checklists.

Leadership moves​

  • Teach model pluralism. Run short sessions where teams map the same landscape through different frames. Highlight how each reveals different failure modes.
  • Attach models to risks. When declaring a risk, specify the model that surfaces it (e.g., VSM highlights missing recursion; Cynefin spots premature standardisation). This keeps risk debates concrete.
  • Use AI to broker frames. Fine-tuned agents can summarise how each model interprets new signals, such as drift, regulation, and user behaviour, then suggest a frame shift before teams harden into camps.
  • Measure meta-skill, not model count. Track how often teams switch frames when evidence demands it, not how many frameworks they can name.

What goes wrong when you worship the map​

  • Regulatory whiplash. Single-frame thinking on compliance can blindside deployment when the map says “fast” but lawmakers say “halt”.
  • Safety incidents. Over-trusting the map’s evolution guesses can ignore ethical signals that don’t fit the neat component boundaries.
  • Competitive myopia. Optimising for tempo alone can miss the cultural or political moves competitors are making off-map.

Connections and further reading​