Cross-Impact Expert Elicitation¶
Elicitation method — instantiates Cross-Impact Interaction Mapping
Sources judgments about how drivers influence each other from domain experts — capturing the relation, its confidence, and any category-changing interaction — where evidence is thin but expertise is deep.
When the drivers are geopolitical, technological, or a decade out, there is no dataset to regress. What exists instead lives in the heads of people who have watched the system for years. Cross-Impact Expert Elicitation is the method for getting those tacit judgments out cleanly and onto the map. Its defining discipline is that it is a sourcing procedure, not a scoring scale and not a group therapy session: it puts a specific relation question to specific experts, records each judgment with its stated confidence, and — crucially — asks not only "how strongly does A affect B?" but "could A and B together produce something categorically new that neither implies alone?" It is the mechanism that populates the map's relation records when the only available evidence is disciplined human judgment, and its whole art is extracting that judgment without contaminating it.
Example¶
A national foresight cell is assessing how a bounded set of drivers — a rival's export controls, an allied bloc's re-armament, a semiconductor shortage, and a shift toward energy autarky — might interact over ten years. There is no historical base rate for the specific combination. So the cell runs a structured elicitation: fifteen analysts from trade, defense, and industry receive the driver pairs individually, judge each relation's direction and mechanism, and mark their confidence separately from the strength ("high impact, but I'm only moderately sure"). Estimates are collected anonymously, the spread is fed back, and experts revise once — a Delphi-style iteration[1] that dampens the loudest-voice effect.
The elicitation's most valuable output is not the averaged relation scores; it is the transformative flag one analyst raises and others then corroborate: export controls plus energy autarky do not merely each slow the rival — together they could tip it toward a self-sufficient, sanction-proof industrial bloc, a qualitatively different strategic condition. That category-change would have been averaged away by a scoring routine; the elicitation surfaced it because it explicitly asked.
How it works¶
- Pose the relation, not the answer. Each expert is asked whether and how one driver changes another's likelihood, strength, or meaning — framed neutrally so the question does not telegraph the expected reply.
- Record confidence separately from strength. Every judgment carries two marks: how large the effect and how sure the expert is. A confident "small" and an uncertain "huge" must never collapse into one number.
- Iterate anonymously. Collect independently, show the group the distribution, and allow one round of revision — reducing anchoring and status effects without forcing false consensus.
- Hunt explicitly for transformation. A dedicated prompt asks whether any pair, combined, changes the category of the future rather than the magnitude of a driver — the judgments a mean would erase.
Tuning parameters¶
- Panel breadth — how many disciplines and how adversarial the mix. Broader panels catch cross-domain interactions but slow convergence and widen the spread.
- Anonymity and iteration rounds — how insulated experts are from one another and how many revision passes. More insulation curbs bandwagoning; more rounds risk regression to a bland middle.
- Confidence scale resolution — coarse (low/med/high) versus a finer probabilistic scale. Finer scales invite false precision from people who cannot really tell 60% from 70%.
- Disagreement handling — whether persistent splits are averaged, preserved as rival judgments, or escalated for deeper inquiry. Averaging is tidy; preserving is honest.
When it helps, and when it misleads¶
Its strength is reaching the interactions no dataset covers and doing so with the two safeguards evidence-poor foresight most needs: confidence held apart from magnitude, and a deliberate search for category-changing effects. Where the future is genuinely novel, expert judgment elicited well is often the best instrument available.
Its failure mode is the correlated error of a like-minded panel: fifteen experts who trained in the same tradition can be confidently, uniformly wrong, and the method's tidy aggregation launders that shared blind spot into apparent consensus. The classic misuse is treating an elicited number as a measurement rather than a snapshot of belief — quoting "0.7" as if it were observed. The guarding discipline is to keep confidence visibly attached to every relation, recruit for genuine diversity of view, and preserve stubborn disagreements as open questions rather than averaging them into a false middle.
How it implements the components¶
Cross-Impact Expert Elicitation fills the judgment-sourcing end of the archetype:
interaction_relation_record— each elicited judgment becomes a relation record: which driver affects which, in what direction, by what stated mechanism.impact_strength_and_confidence_rating— experts mark magnitude and certainty separately, so weakly held judgments never masquerade as facts.transformative_interaction_effect— a dedicated prompt surfaces the category-changing combinations a scoring average would smooth away.
It does not impose the comparable numeric rubric — the standardized magnitude, sign, and timing scale — that makes any judgment mechanically sortable (time_lag_and_sequence_marker, reinforcement_effect); that is Pairwise Influence Scoring. Elicitation sources the underlying judgments from people; Pairwise Influence Scoring is the rubric that renders them comparable. It also does not stage a live facilitated debate that captures how different actors interpret the same interaction (stakeholder_interpretation_layer) — that is Impact Interaction Workshop.
Related¶
- Instantiates: Cross-Impact Interaction Mapping — supplies the relation records the rest of the map is built from when evidence is thin.
- Sibling mechanisms: Pairwise Influence Scoring · Impact Interaction Workshop · Compound Risk Map · Driver Cluster Heatmap · Driver Network Graph · Trend Interaction Map · Scenario Dependency Diagram · Trigger Dependency Watchlist
Editorial Notes¶
Form Classification¶
Form family: Communication, Facilitation & Learning
Rationale: Cross-Impact Expert Elicitation operates as a designed message, facilitated interaction, ritual, or learning activity that changes shared understanding because it sources judgments about how drivers influence each other from domain experts — capturing the relation, its confidence, and any category-changing interaction — where evidence is thin but expertise is deep.
Independent corroboration: The frozen evidence defines Cross-Impact Expert Elicitation as 'Sources judgments about how drivers influence each other from domain experts — capturing the relation, its confidence, and any category-changing interaction — where evidence is thin but expertise is deep', so its operative form is Communication, Facilitation & Learning.
Nearest alternative: Analysis, Modeling & Optimization — Structured expert interaction elicits judgments; later modeling or aggregation is not the mechanism's defining operation.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Futurism & Strategic Foresight
Origin pattern: Single lineage
Present-day reach: Specialized
Rationale: Eliciting expert estimates of interacting future impacts is principally a foresight method, disciplined by statistical expert-judgment practice.
Related originating lineages:
- Statistics & Experimental Design — Structured expert-judgment methods supply calibration, aggregation, and uncertainty recording.
Review resolution: Eliciting expert estimates of interacting future impacts is principally a foresight method, disciplined by statistical expert-judgment practice.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Dalkey, N., & Helmer, O. "An Experimental Application of the DELPHI Method to the Use of Experts". Management Science 9(3), 458–467 (1963). Uses independent expert judgments with controlled aggregate feedback and revision to limit dominant-person influence. registry ↩