The systems that interest me
I am interested in systems where the true state cannot be directly observed, where available signals are incomplete, and where waiting for perfect knowledge would prevent action entirely.
- A learning platform attempting to estimate educational value.
- A student trying to determine whether learning occurred.
- An engineer deciding whether a model is sufficient to build.
- An immigrant evaluating fragmented community information.
- A physical system whose internal interactions are inferred through outputs.
Uncertainty is not permission for carelessness
Acknowledging uncertainty does not mean every claim is equally reasonable. Evidence can be weak or strong. Assumptions can be visible or hidden. Decisions can be reversible or irreversible. Consequences can be small or catastrophic.
Confidence should be earned
A system should not become confident merely because it has a number. A model can produce precise outputs while depending on weak assumptions. Precision is not the same as justification.
This is why my work repeatedly returns to evidence, context, system boundaries, latent variables, and the question of sufficient understanding.
Decrease one uncertainty, open two upgraded uncertainties.