Echo chamber dynamics and influence aware evidence aggregation for epistemic integrity in knowledge based systems

Correlated corroboration can make knowledge-based systems overconfident when dependent reports are treated as independent evidence. This paper presents an influence-aware evidence aggregation architecture that makes source dependence explicit and constrains confidence growth under partial dependence knowledge. The architecture combines an influence-indexed evidence graph (EIIG), a generalised estimate of effective independent evidence (EIE) with graded dependence weights, and a bounded confidence-propagation rule (CONFPROP). We evaluate the architecture against three baselines—naive independence aggregation, a Bayesian independence update, and a source-discounted comparator—across four controlled synthetic scenarios and a semi-realistic external case study using 500 claims from the FEVER fact-verification corpus. Under the strongest synthetic dependence regimes, CONFPROP consistently reduces confidence inflation relative to naive aggregation: inflation ratios are 4.00 in a common-parent echo-chamber scenario and 10.00 in a fully duplicated cascade, while no suppression occurs when evidence is genuinely independent. A factorial sensitivity analysis across 648 parameter combinations per stress-test scenario confirms that the discount factor γ decreases with the diversification threshold K and increases with the unknown-status weight α . The architecture provides deterministic, auditable confidence updates with structured monitoring signals, and is designed for modular integration into existing knowledge-based systems operating under uncertain source dependence.