Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.18284429 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866901362774638592 |
|---|---|
| author | Grillo, Steven |
| author_facet | Grillo, Steven |
| contents | <p>The Anthropocentric Fallacy in Agentic AI Governance</p> <p>Why Human-in-the-Loop Control Fails at Machine Speed</p> <p>Abstract</p> <p> </p> <p>As artificial intelligence systems transition from assistive generation to autonomous, agentic execution, governance architectures inherited from human-centric workflows increasingly impose structural limitations. This paper examines the limitations of Human-in-the-Loop (HITL) governance in high-velocity agentic systems, arguing that biological decision latency and cognitive variance introduce measurable performance degradation and systemic inconsistency.</p> <p>Drawing on findings from cognitive science, systems theory, and ensemble machine learning, we demonstrate that synchronous human oversight is poorly matched to millisecond-scale execution environments. We introduce a machine-speed governance framework, referred to as the Digital Senate, which replaces transactional human gating with adversarial multi-agent consensus operating under deterministic policy constraints. Human oversight is repositioned to an asynchronous audit role—Human-on-the-Rail (HOTR)—preserving accountability without constraining execution velocity.</p> <p>This architecture reframes AI governance as a systems engineering problem rather than a supervisory one, offering improved scalability, consistency, and auditability in autonomous decision pipelines. </p> <p> </p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18284429 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | The Anthropocentric Fallacy in Agentic AI Governance Grillo, Steven <p>The Anthropocentric Fallacy in Agentic AI Governance</p> <p>Why Human-in-the-Loop Control Fails at Machine Speed</p> <p>Abstract</p> <p> </p> <p>As artificial intelligence systems transition from assistive generation to autonomous, agentic execution, governance architectures inherited from human-centric workflows increasingly impose structural limitations. This paper examines the limitations of Human-in-the-Loop (HITL) governance in high-velocity agentic systems, arguing that biological decision latency and cognitive variance introduce measurable performance degradation and systemic inconsistency.</p> <p>Drawing on findings from cognitive science, systems theory, and ensemble machine learning, we demonstrate that synchronous human oversight is poorly matched to millisecond-scale execution environments. We introduce a machine-speed governance framework, referred to as the Digital Senate, which replaces transactional human gating with adversarial multi-agent consensus operating under deterministic policy constraints. Human oversight is repositioned to an asynchronous audit role—Human-on-the-Rail (HOTR)—preserving accountability without constraining execution velocity.</p> <p>This architecture reframes AI governance as a systems engineering problem rather than a supervisory one, offering improved scalability, consistency, and auditability in autonomous decision pipelines. </p> <p> </p> |
| title | The Anthropocentric Fallacy in Agentic AI Governance |
| url | https://doi.org/10.5281/zenodo.18284429 |