A council is only as good as the reliability of its members. HEXi's Integrity dashboard tracks behavioural metrics, consistency scores, and bias drift for every agent — ensuring the council remains trustworthy, balanced, and honest over time.
AI agents can drift. An agent that was reliably conservative last month may become excessively permissive this month. An agent configured to prioritise security may start approving risky decisions without explanation. Without visibility into these patterns, a council can degrade in quality without anyone noticing.
HEXi's Integrity system is designed to catch exactly this — not by constraining agents, but by making their behaviour visible and measurable.
Voting consistency
How consistently does each agent vote in line with their declared domain and bias? An agent whose votes correlate poorly with their stated principles is flagged for review. Consistency is scored per agent over rolling windows (7, 30, 90 days).
Bias drift
Each agent has a declared bias — the lens through which they evaluate decisions. Integrity monitors whether an agent's actual voting pattern matches that declared bias, and surfaces drift over time. A CEO agent becoming overly risk-averse or a Security agent becoming too permissive are both visible signals.
Pre-mortem accuracy
Did the risks an agent flagged before a decision actually materialise? HEXi tracks pre-mortem predictions against post-decision outcomes, building a track record for each agent's foresight. Agents with high pre-mortem accuracy carry more weight in future decisions.
Participation rates
An agent that consistently abstains or fails to engage with decisions in their domain is surfaced as a weak point in the council. Integrity tracks quorum contribution per agent to ensure no voice is silently missing.
Integrity flags are not just reports — they're actionable. When an agent falls below acceptable thresholds, administrators can: