Second-Order Expectation Survey¶
Measurement instrument — instantiates Higher-Order Expectation Anchoring
Elicits not just what each person believes but what they think others believe — and what they think others think others believe — then tags every answer by its belief order.
A Second-Order Expectation Survey pulls apart the layers of belief that a single question usually smears together. Rather than asking one question, it asks the same choice at several registers: what you privately believe is best, what you expect the group to choose, and what you expect the group expects the group to choose. Its defining move is that every response is stored with a belief-order tag, so a first-order conviction is never silently mistaken for a forecast of the crowd. The instrument does not try to change minds or shield answers from influence; it only makes the recursion measurable — turning "everyone seems to agree" into data that can reveal whether the agreement is shared conviction or a stack of mutual guesses.
Example¶
A film magazine polls its forty critics ahead of the Best Picture announcement. A naive poll would ask one question — "which film wins?" — and report a favorite. This survey asks three. First, privately, which film do you think is the best work? Second, which film do you think will actually win? Third, which film do you think most critics expect to win? The answers diverge sharply. A quiet drama collects the most first-order votes for "best," but a splashy epic dominates the second and third registers: critics expect the epic to win because they believe other critics expect it to win. The survey's tag on each answer makes the gap legible — roughly two-thirds privately prefer the drama, yet nearly the same fraction predict the epic. That gap is the fingerprint of a beauty-contest dynamic, and it is invisible to any instrument that asks only one question.
How it works¶
The survey runs the same decision through parallel, order-labeled prompts and never collapses them. First-order items capture private judgment ("what is true / best / preferred"). Second-order items capture the expected group choice. Third-order items capture the expected expectation. Responses are aggregated within each order, and the headline output is not a winner but a set of per-order distributions plus the divergence between them. A large first-to-second-order gap flags that visible consensus is a forecast, not a preference. Optionally, respondents are scored for accuracy against the eventual outcome, so higher-order answers can be calibrated rather than taken at face value.
Tuning parameters¶
- Belief-order depth — how many recursion levels to probe (stop at second, push to third or fourth). Deeper probing exposes more of the loop, but the answers get noisier fast because people cannot reliably introspect past a couple of levels.
- Incentive scheme — unincentivized opinion versus an accuracy-scored prediction. Scoring the higher-order questions sharpens them but recruits gaming; leaving them unscored keeps them candid but soft.
- Reference group — whether "others" means all critics, a named elite panel, or the eventual voters. This quietly sets which expectation is being measured, and swapping it can flip the result.
- Anonymity — whether responses are collected privately. More anonymity reduces the pull to answer strategically, but this survey measures the field rather than protecting it, so it stops short of full blinding.
When it helps, and when it misleads¶
Its strength is diagnostic: it separates genuine agreement from preference falsification, showing when a group is publicly converging on something few privately endorse — the exact condition the surrounding archetype exists to treat. Read correctly, the first-to-second-order gap tells a facilitator whether to trust a consensus or to protect independent judgment before acting on it.
Its failure mode is that higher-order answers are intrinsically unreliable: people flatten out at roughly two levels of "I think that you think," so a third- or fourth-order number is often noise wearing a decimal point.[n1] The classic misuse is to read a second-order forecast ("what will win") as if it were a first-order fact ("what is best") and then anchor a decision to it — which quietly launders the crowd's guess into evidence. The guarding discipline is to always report each figure with its order attached and to treat divergence as a question to investigate, not a verdict to act on.
How it implements the components¶
higher_order_belief_map— the tiered questions are the map: they name and separately measure the first-order, second-order, and third-order layers instead of letting them collapse into one number.belief_order_depth_marker— every response carries the order it was elicited at, so downstream consumers can never confuse a preference with an expectation-of-an-expectation.
It does not build a facilitated structural map or set a belief_order_depth_limit — that is the Expectation-Ladder Mapping Workshop, its nearest twin, which constructs the layers qualitatively and caps how deep the group will reason. Nor does it protect a private_preference_baseline from contamination; measuring an answer is not the same as shielding it, which is the Blind Independent Forecast Round's job.
Related¶
- Instantiates: Higher-Order Expectation Anchoring — supplies the measured belief-layer map the rest of the intervention reasons over.
- Sibling mechanisms: Blind Independent Forecast Round · Common-Belief Dashboard · Common-Knowledge Brief · Delphi Round with Visibility Lag · Expectation-Ladder Mapping Workshop · Expectation Reset Announcement · Focal-Point Charter · Narrative Momentum Stress Test · Public-Signal Lag or Blinding
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: Second Order Expectation Survey operates by externalizes parallel first-, second-, and third-order expectation prompts in a stable survey instrument. That concrete deployed or enacted form is Representation, Specification & Plan under the frozen taxonomy.
Nearest alternative: Interface, Display & Cue — Although Interface, Display & Cue can support this mechanism, the frozen evidence makes its operative form the act that externalizes parallel first-, second-, and third-order expectation prompts in a stable survey instrument; the alternative is therefore secondary rather than defining.
Review outcome: Adjudicated after independent review; medium confidence.
Origin Attribution¶
Primary origin: Psychology
Origin pattern: Convergent development
Present-day reach: Multi-domain
Rationale: Eliciting beliefs about others' beliefs is rooted in social psychology and pluralistic-ignorance research.
Related originating lineages:
- Cognitive Science — Cognitive-science research on representation, learning, and recall supplies a parallel or contributing lineage for the mechanism's defining operation: elicits not just what each person believes but what they think others believe — and what they think others think others believe — then tags every answer by its belief order.
- Economics & Finance — Higher-order beliefs independently shape games and market coordination.
- Political Science — Political science and institutional power analysis supplies a parallel or contributing lineage for the mechanism's defining operation: elicits not just what each person believes but what they think others believe — and what they think others think others believe — then tags every answer by its belief order.
- Sociology & Anthropology — Collective expectations materially structure norms.
Review resolution: The blind reviewers agree that psychology is the primary origin and differ only on alternate origin disagreement, origin mode disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain convergent because the combined record shows independent disciplinary development. The broader reach of multi_domain records portability separately from historical provenance, and encyclopedia_synthesis=false preserves the affirmative synthesis judgment where either reviewer identified one.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] In the level-k / cognitive hierarchy models of strategic reasoning (Stahl & Wilson; Camerer, Ho & Chong), most people reason only one or two steps of "I think that you think…" before stopping, which is why empirically elicited third- and higher-order beliefs are dominated by noise. ↩