Research on the Optimisation Path of Algorithm Transparency and Board Accountability Mechanisms in an Artificial Intelligence-Assisted Governance Model
Algorithmically-triggered governance remains underdeveloped in AI-integrated institutions, leading to decision delays and unresolved anomalies. This study aimed to evaluate precision, latency, and resolution outcomes of AI-escalation models embedded in institutional board workflows. The study collected 300 AI-generated escalation cases from 30 participants, stratified by legal, administrative, and technical roles. Annotated cases were evaluated using role-calibrated interfaces and tiered trigger thresholds. Models were trained and tested using rule-based routing with bootstrap validation. Comparative analyses (pre- and post-intervention) and subgroup comparisons were performed using ANOVA, the Kruskal–Wallis test, and effect size metrics. All analyses were conducted in Python. Interpretation accuracy ranged from 74.9% to 91.4%, with AI analysts achieving the highest performance. Routing precision peaked at 88.2%, board action success at 82.5%, and tooltip interaction averaged 10.5 clicks. Interface refinement improved task accuracy by 7.0%, decreased latency by 560 ms, and increased routing success by 8.2%. Escalation latency reduced by 630 to 650 ms across domains. Trigger-specific sensitivity was highest for reclassification (92.1%), with final resolution times averaging 4,510 ms. Post-outlier corrections yielded minor gains (<1.2%), and misroutes accounted for the highest unresolved rates (40.0%). Structured escalation protocols notably enhanced AI-governance alignment and reduce critical response delays.
Authors
- Ling Jin (ORCID: https://orcid.org/0009-0003-5387-7594)
Institutions
- Yanbian University (CN)
Publication Details
- Journal
- Administration & Society
- Published
- 2026-11-12
- DOI
- https://doi.org/10.1177/00953997261498380
- Primary Topic
- Ethics and Social Impacts of AI
- Type
- article
- Field-Weighted Citation Impact
- 0.00