Thesis. The B(O, C) formula is not a metaphor for "how confident you are." It is an explicit, multiplicative composition of four operationalizable quantities — attentional focus F, emotional coherence E, internal contradiction σ, and empirical reinforcement Λ — and it predicts the stability of an observer's configurations against interaction with environment. If any single component collapses to zero, B collapses to zero. That multiplicative structure is the whole point.
The four components
F — attentional focus. How concentrated the observer's attention is on the configuration. High F means cognitive resources are directed at tracking Ψ without dispersion. Measured via attention stability and processing depth.
E — emotional coherence. Alignment of the observer's emotional state with the intention regarding configuration C. High E means the emotional system supports and reinforces the cognitive commitment; low E means emotional dissonance, conflicting feelings, or fragmented engagement.
σ — internal contradiction. The observer's entropy of doubt regarding configuration C. The formula uses (1 − σ) so that low contradiction (high internal consistency) pushes B up; σ = 1 (fully contradicted, maximum doubt) collapses B to zero regardless of F or E.
Λ — empirical reinforcement. Accumulated confirmatory experience within configuration C. How much past experience supports and reinforces the current configuration. High Λ means the observer has extensive experiential backing; low Λ means the configuration is novel or unsupported by prior experience.
Why multiplicative
The natural objection: why not B = w1·F + w2·E + w3·(1−σ) + w4·Λ ? Because any single component going to zero is a death blow. A perfect signal (F = 1) in a perfectly chaotic context (Λ = 0) gives you nothing — your representation has nowhere to attach. Multiplicative form encodes this AND-gate structure. Additive form would let one strong component compensate for a missing one, which empirically does not happen.
This is the same reason the geometric mean shows up in healthy growth formulas: when a process requires all inputs jointly, you cannot trade one off against another.
The weights w1..w4
The weights specify the kind of observer you are. A scientist on a clean experiment weights F highest; a chess player in time-trouble weights E (emotional alignment under pressure) highest; a journalist verifying a source weights Λ (experiential backing) highest. The weights sum to 1 by convention, but the framework does not force this — what matters is their ratio.
The measuring B parameter article gives several elicitation protocols: paired comparison, regression against known outcomes, and a Bayesian update procedure for refining weights as evidence accumulates.
Where this beats Bayesian probability
A Bayesian credence is one-dimensional: a number in [0, 1]. B(O, C) is structurally four-dimensional, and the four dimensions are not interchangeable. You can have two observers with identical Bayesian credences but very different B profiles, and they will behave very differently under perturbation. Specifically, an observer with high F but low E will flip beliefs under social pressure; an observer with high E but low F will defend wrong beliefs against new evidence. Bayesian probability cannot see this distinction.
For the formal proof that B is non-reducible to a single probability, see Cognitive coherence measurability.
Practical estimation, in 30 seconds
For a working estimate of B(O, C) on a real claim:
- Score F from 0 to 1 by asking: how well does this match what reliable sources independently report?
- Score E from 0 to 1 by asking: does the observer hold other beliefs that contradict this one?
- Score σ from 0 to 1 by asking: how much of the context is noise versus signal?
- Score Λ from 0 to 1 by asking: how rich is the contextual data backing the claim?
- Apply equal weights (w1 = w2 = w3 = w4 = 0.25) as a first pass.
This is not the formal procedure — it is the napkin version. For decisions of consequence, use the full elicitation in the belief paper.