feat(iace): risk as confidence range + label in benchmark tab

Report the tool's risk number as a plausible range with a confidence
label instead of a false-precision point value (confidence-aware
tonality — the assessment is confirmed by the DSB / safety expert).

- risk_estimation.go: EstimateConfidence (hoch/mittel/niedrig from how the
  contact mode resolved), EstimateRiskRange (S±1 and aggregate L=F+W+P ±1,
  the empirically validated per-parameter accuracy), RiskLevelRange; share
  the riskBandLabel thresholds with EstimateRiskLevel.
- risk_benchmark.go: RiskComparisonPair gains eng_risk_point/low/high +
  level + level_range + confidence; RiskAgreement gains high_confidence_pct.
- RiskComparison.tsx: per-hazard range "low–high (level range)" + point,
  confidence chip, and an aggregate confidence line; types in useBenchmark.ts.
- Unit tests for the range/confidence helpers.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

ai-compliance-sdk/internal/iace/risk_benchmark.go

@@ -19,6 +19,15 @@ type RiskComparisonPair struct {
 	EngAvoidance   int     `json:"eng_avoidance"`
 	FKScore        float64 `json:"fk_score"`
 	FKBand         string  `json:"fk_band"`
+	// Confidence-aware risk: a point estimate plus a plausible low/high band and
+	// a confidence label, so the tool reports a RANGE (not a false-precision
+	// point) — the assessment is confirmed by the DSB / safety expert.
+	EngRiskPoint      int    `json:"eng_risk_point"`
+	EngRiskLow        int    `json:"eng_risk_low"`
+	EngRiskHigh       int    `json:"eng_risk_high"`
+	EngRiskLevel      string `json:"eng_risk_level"`       // band of the point value
+	EngRiskLevelRange string `json:"eng_risk_level_range"` // e.g. "mittel–hoch"
+	Confidence        string `json:"confidence"`          // hoch / mittel / niedrig
 }
 
 // RiskAgreement aggregates how close the tool's risk numbers are to the GT.
@@ -28,14 +37,15 @@ type RiskAgreement struct {
 	FrequencyWithin1   float64 `json:"frequency_within1"`
 	ProbabilityWithin1 float64 `json:"probability_within1"`
 	AvoidanceWithin1   float64 `json:"avoidance_within1"`
-	RankConcordance    float64 `json:"rank_concordance"` // Fine-Kinney vs GT R
+	RankConcordance    float64 `json:"rank_concordance"`   // Fine-Kinney vs GT R
+	HighConfidencePct  float64 `json:"high_confidence_pct"` // share of matched hazards with "hoch" confidence
 }
 
 // ComputeRiskComparison derives the tool's risk numbers for each matched hazard
 // and compares them to the professional's GT values.
 func ComputeRiskComparison(matched []HazardMatchPair) ([]RiskComparisonPair, RiskAgreement) {
 	pairs := make([]RiskComparisonPair, 0, len(matched))
-	var sevOK, freqOK, probOK, avoidOK, n int
+	var sevOK, freqOK, probOK, avoidOK, n, hiConf int
 	var engFK, gtR []float64
 
 	for _, m := range matched {
@@ -54,11 +64,20 @@ func ComputeRiskComparison(matched []HazardMatchPair) ([]RiskComparisonPair, Ris
 		fk := SuggestFineKinney(cats, scenario, lifecycle, 0)
 		gt := m.GTEntry.RiskIn
 
+		rLow, rPoint, rHigh := EstimateRiskRange(engS, engF, engW, engP)
+		rLevel, rLevelRange := RiskLevelRange(rLow, rPoint, rHigh)
+		conf := EstimateConfidence(cats, scenario)
+		if conf == "hoch" {
+			hiConf++
+		}
+
 		pairs = append(pairs, RiskComparisonPair{
 			HazardName:     m.GTEntry.HazardType,
 			GTSeverity:     gt.S, GTFrequency: gt.F, GTProbability: gt.W, GTAvoidance: gt.P, GTRisk: gt.R,
 			EngSeverity:    engS, EngFrequency: engF, EngProbability: engW, EngAvoidance: engP,
 			FKScore:        fk.Score, FKBand: fk.Band,
+			EngRiskPoint:   rPoint, EngRiskLow: rLow, EngRiskHigh: rHigh,
+			EngRiskLevel:   rLevel, EngRiskLevelRange: rLevelRange, Confidence: conf,
 		})
 
 		if gt.S > 0 {
@@ -88,6 +107,9 @@ func ComputeRiskComparison(matched []HazardMatchPair) ([]RiskComparisonPair, Ris
 		agg.AvoidanceWithin1 = pct(avoidOK, n)
 		agg.RankConcordance = rankConcordance(engFK, gtR)
 	}
+	if len(pairs) > 0 {
+		agg.HighConfidencePct = pct(hiConf, len(pairs))
+	}
 	return pairs, agg
 }