A risk assessment system deploys 5 parallel Claude agents to independently evaluate the same loan application. The architect must aggregate their risk scores (each returns High/Medium/Low with a confidence score) into a single decision. Scores arrive as: 3x High (confidence 0.9, 0.85, 0.8), 1x Medium (confidence 0.7), 1x Low (confidence 0.6). Which aggregation strategy is most appropriate for a financial risk context where false negatives (approving risky loans) are costly?

The conservative consensus rule is correct because in financial risk contexts, the cost of false negatives, such as approving a risky loan, typically exceeds the cost of false positives, such as rejecting a safe loan. A conservative consensus rule that elevates to the highest risk category when any agent exceeds a confidence threshold, such as 0.75, ensures that high-confidence risk signals are not diluted by lower-confidence dissenting agents. This asymmetric rule reflects the asymmetric cost structure. The simple majority-vote approach is wrong because it ignores confidence scores, so a 0.6-confidence Low vote and a 0.9-confidence High vote carry equal weight, which is inappropriate when confidence information is available. The confidence-weighted voting approach is wrong because it still allows low-confidence dissenting votes to reduce the overall score below what the high-confidence majority indicates, and in a risk-averse context, this averaging is dangerous. The median-confidence approach is wrong because it is a heuristic that conflates the confidence of individual assessments with the risk category selection; the correct risk output is High in this case, but the reasoning based on median confidence is incorrect and would fail in other distributions.