Expectancy-Calibrated Feedback Form¶
Feedback artifact — instantiates Prediction-Error Learning Calibration
A feedback template that records what a person expected before it records what happened, so praise and correction land on the surprise rather than the raw result.
Human feedback usually rewards outcomes, not surprises — the rep who closed the sure-thing deal gets the same applause as the one who pulled off a long shot, and the person whose well-judged bet went wrong gets punished for bad luck. Expectancy-Calibrated Feedback Form is the artifact that fixes the ordering: it asks the recipient to state what they expected — their own prior, ideally before the outcome is known — and only then records what happened, so the feedback conversation can attach to the gap. Its defining move is establishing a shared reference frame of expectation first, which is what lets a manager praise the judgment that beat the odds rather than the result that merely met them, and withhold blame from a loss that was honestly forecast. It is the human, social counterpart to the algorithmic delta log: same signed-surprise logic, delivered through a person, in a conversation, close to the event.
Example¶
A sales manager debriefs a rep after a hard-fought enterprise deal. The form goes in order. Going in, what did you think? — "Maybe 40% — the incumbent was entrenched." What happened? — Closed. Because the expectation was captured first, the praise now targets the specific thing that beat the rep's own read — the reframing that neutralized the incumbent — rather than the raw win. On the next line is a deal the rep had flagged at 20% that fell through: no reproach, because that outcome was expected and priced in. The form deliberately does the same for both signs of surprise, and it is filled within a day of each event, while the rep can still connect the feedback to what they actually did. What the rep learns is not "closing is good, losing is bad" but "your judgment was sharp here and your read was off there" — the only feedback that improves calibration.
How it works¶
- Elicit the expectation first. The recipient states their prior — a rough probability or a qualitative read — before the outcome is discussed, establishing the reference the feedback will be measured against.
- Then record the outcome, and locate the gap. Feedback attaches to the distance between expected and received, not to the result's raw size.
- Deliver through the person, close to the event. The form is a script for a conversation held soon enough that the recipient can still trace the feedback to their own actions.
The distinguishing feature is the expectation-first ordering delivered socially — it is a conversation instrument, not a data store.
Tuning parameters¶
- Whose expectation anchors it — the recipient's own prior, the manager's, or a jointly agreed one. The recipient's prior teaches self-calibration but can be gamed; a shared prior is fairer but slower.
- Reference explicitness — numeric odds versus a qualitative read. Numbers sharpen the gap and enable scoring; words lower the barrier and suit softer domains.
- Timing window — how soon after the event feedback is delivered. Tighter is more attributable but less reflective; looser risks the trail going cold.
- Channel formality — a logged form versus a light verbal ritual. Formality builds a durable record and even-handedness; informality preserves trust and speed.
When it helps, and when it misleads¶
Its strength is that it disarms the two feedback pathologies at once: rewarding expected wins as if they were achievements, and punishing well-judged bets for unlucky results. Over time it teaches people to hold and state honest priors, which is the foundation of calibration — and by insisting feedback be delivered close to the event through the actual observer, it keeps the signal attributable instead of stale and secondhand.
Its central vulnerability is that it is gameable through the expectation itself: a recipient who learns the game can sandbag — state a deflated prior so any ordinary result registers as a triumph.[n1] Elicited after the fact, the "expectation" is also contaminated by knowing the answer, collapsing the whole exercise back into outcome praise. And over-formalized, it can make ordinary feedback feel like an audit. The discipline that keeps it honest is to commit expectations before outcomes — ideally drawing the prior from a committed record rather than asking for it in hindsight — and to watch for priors that keep drifting suspiciously low.
How it implements the components¶
Expectancy-Calibrated Feedback Form realizes the human-feedback side of the archetype — the components that route surprise through a person rather than a computation:
value_reference_frame— it establishes the recipient's stated expectation as the baseline the feedback is measured against, per person and per event.feedback_timing_alignment— it prescribes delivery close enough to the event that the recipient can still attribute the feedback to their own actions.social_observation_channel— it is the conduit through which an observer's judgment enters the loop, turning a private outcome into shared, calibrated feedback.
It does not compute or store the numeric signed error (that is Prediction–Outcome Delta Log), set how much the model should update on a surprise (Learning Rate Schedule), or run the sign-specific follow-through — harvesting the upside is Positive Surprise Capture, reviewing the downside is Negative Prediction Error Review.
Related¶
- Instantiates: Prediction-Error Learning Calibration — this artifact delivers the archetype's surprise-over-outcome logic in a human feedback conversation.
- Sibling mechanisms: Prediction–Outcome Delta Log · Negative Prediction Error Review · Positive Surprise Capture · Learning Rate Schedule · Credit Assignment Trace · Calibration Curve Review · Reward Baseline Dashboard · Reward Signal Red Team · Shortcut Probe Holdout Set · Surprise Threshold Alert · Temporal-Difference Update Rule
Editorial Notes¶
Form Classification¶
Form family: Interface, Display & Cue
Rationale: Expectancy-Calibrated Feedback Form operates as a user-facing prompt, display, template, or perceptual cue that shapes attention and action at the point of use because it a feedback template that records what a person expected before it records what happened, so praise and correction land on the surprise rather than the raw result.
Independent corroboration: The frozen evidence defines Expectancy-Calibrated Feedback Form as 'A feedback template that records what a person expected before it records what happened, so praise and correction land on the surprise rather than the raw result', so its operative form is Interface, Display & Cue.
Nearest alternative: Communication, Facilitation & Learning — The feedback form shapes expectation and outcome entry at the moment of conversation, while the surrounding exchange delivers learning content.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Psychology
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Recording predictions before outcomes to center learning on surprise derives from expectancy and prediction-error psychology.
Related originating lineages:
- Education & Pedagogy — Formative-feedback practice materially shapes the structured feedback artifact. Formative feedback practice materially shapes the person-facing conversation and learning artifact.
- Statistics & Experimental Design — Precommitted probabilities and calibration against outcomes materially shape the expectation-first measurement discipline.
Review resolution: Both reviewers agree that psychology is primary. I retain education_pedagogy, statistics_experimental_design only as formative origin lineages; cross_disciplinary_synthesis is appropriate because the final form materially combines the agreed primary with the retained formative lineages. Reach is multi_domain because the structure transfers across several fields but is not a near-universal human pattern, an applicability judgment kept separate from provenance. Encyclopedia synthesis is true because the exact generalized artifact is an encyclopedia-authored combination or refinement. No unresolved historical ambiguity remains after reconciling the secondary fields.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
The form calibrates the delivery of feedback, not the magnitude of any model update — it changes what a person is told and when, not how far a policy or belief moves in response. That sizing belongs to Learning Rate Schedule. The form is also only as trustworthy as the priors it captures; wherever committed predictions already exist in a Prediction–Outcome Delta Log, drawing the expectation from there rather than re-eliciting it in hindsight is what closes the sandbagging loophole.
[n1] Sandbagging — deliberately stating a low expectation so that an ordinary result registers as a positive surprise. Because the form scales feedback to the gap from the stated prior, it rewards honest priors and is exploited by dishonest ones; anchoring on a pre-committed prediction is the guard. ↩