Skip to content

Expectation Audit

A read-only diagnostic review — instantiates Self-Fulfilling Prophecy Interruption

Traces where an expectation, label, or score enters a process and maps the behavior channels through which it may be quietly manufacturing its own confirmation.

Before you can interrupt a self-confirming loop you have to prove one is running and see where it lives. Expectation Audit walks a process and finds the point where a belief, label, reputation, forecast, or score first enters, then follows every path by which that expectation changes what people do — attention, tone, opportunity, scrutiny, interpretation, feedback — and closes the trace into a loop that ends back at the "confirming" outcome. Its defining move is that it is read-only: it changes nothing and prescribes nothing. Its entire product is a map — the expectation named and located, its behavior channels enumerated, and the loop drawn — that the interrupting mechanisms then act on. If a page could be summarized as "and then we fixed it," it is not this mechanism.

Example

A professional-services firm keeps a staffing note that some junior consultants are "not client-ready." An expectation audit traces the note's life. It originates in a single early review; from there it is visible to every future staffing partner; it changes which engagements those juniors are offered (back-office, low client contact), which in turn thins their client-facing track record, which is exactly what the next promotion committee reads as evidence they are "not client-ready." The audit draws that as a closed loop and marks the two links where the outcome is partly manufactured by the label rather than by the person. It hands over a diagram and a channel list — and stops. Deciding whether to cut the loop at staffing, at review, or at the note itself is left to the mechanisms that follow.

How it works

The distinguishing discipline is that it only reads the system:

  • Locate the expectation's entry point — where the label, score, or belief first attaches, and to whom.
  • Inventory who holds it and where it is visible — every downstream reader and decision it touches.
  • Follow each behavior channel to an outcome — for each holder, what action does the expectation change, and what does that action produce.
  • Close the loop — connect expectation → behavior → response → outcome → interpretation, and flag each link where confirmation could be produced by treatment rather than by the target.

It deliberately declines to choose an interruption point or to intervene; naming those channels precisely is the whole job.

Tuning parameters

  • Scope — a single suspected label versus every expectation trigger in a process. Wider scope finds more loops but dilutes depth.
  • Tracing depth — stop at "this expectation changes staffing," or follow the chain to the confirming outcome. Deeper tracing is what turns a bias suspicion into a loop.
  • Quantify vs. name — merely list each channel, or estimate how strongly each one bends behavior. Quantifying prioritizes the interruption but costs instrumentation.
  • Retrospective vs. live — reconstruct a past loop from records, or watch one operate in real time.

When it helps, and when it misleads

Its strength is making an invisible loop legible: it separates "someone is biased" from "here is the specific channel through which the belief becomes true," which is the difference between a grievance and a fixable system. It is the input every downstream interruption depends on.

Its failure mode is that an audit that stops at mapping produces insight and no change — and its classic misuse is exactly that: run to exonerate ("we looked, we found nothing"), or treated as if the map were the fix. Merton's original self-fulfilling prophecy[1] is a loop, not a static bias, so a map that names no interruptible link has not finished the job. The discipline that guards against this is to require the audit to hand off at least one named, cuttable channel to an interrupting mechanism, and to treat the map as a starting diagnosis, not a verdict.

How it implements the components

Expectation Audit realizes the mapping side of the archetype — the diagnosis, not the cure:

  • expectation_map — names the triggering expectation and its provenance: who holds it, about whom, predicting what, visible where.
  • behavior_channel — enumerates each path by which the expectation changes action.
  • expectation_feedback_loop — assembles those channels into the closed, potentially self-confirming loop.

It does not check whether the expectation was actually warranted (base_rate_and_confound_check, Expectation Calibration Review), nor cut, redirect, or monitor any channel — those interruption and monitoring components belong to siblings such as Blind Review and Outcome Monitoring Review.

Notes

Like a diagnostic estimate, this audit is deliberately an input, not a decision: it sizes and locates the loop but says nothing about where to cut it. Keeping mapping separate from intervention is what lets a team improve the map — a deeper trace, live instrumentation — without re-arguing the whole remedy. It overlaps with Prediction Impact Audit (scored/automated predictions) and Teacher / Manager Expectation Check (a first-person self-review); this one is the organization-level, system-wide map.

References

[1] Robert K. Merton's self-fulfilling prophecy — a false definition of a situation that evokes behavior making the originally false conception come true. Named here because the audit's loop view is exactly Merton's point: the target is a causal circuit, not a mistaken belief in isolation. ↩