Research library

Papers.

Research papers on expert knowledge, human oversight and how agents learn. Read the arguments, inspect the methods and find the code behind the projects.

Expertise

Paper

Tacit Fragments: Operationalising Tacit Knowledge as a Governed Memory Layer for Agentic AI (opens in a new tab)

Sets out the fragment model, capture methods and review process behind Metis.

Arsalan Shahid, Gordon Suttie, Philip Black and Antonio Garzón-Vico · alphaXiv

Trust

Paper

Collaborative Human-Agent Protocol (CHAP) (opens in a new tab)

A protocol that records approvals, overrides and handoffs between people and AI agents as structured, verifiable events.

Arsalan Shahid, Gordon Suttie and Philip Black · arXiv

Policy paper

Keeping the Human in Command: How Ireland Can Lead on Accountable AI (opens in a new tab)

Makes the case for recording human decisions on AI recommendations, and for Ireland to support open standards that make oversight visible.

Arsalan Shahid · PublicPolicy.ie · UCD Geary Institute

Concept paper

Composable Digital Intelligence (opens in a new tab)

Examines the strengths and limits of different forms of intelligence, and how to combine them with responsibility for the whole system.

Arsalan Shahid, Gordon Suttie and Philip Black · alphaXiv

Agents

Paper

Brevet: Change Control for What Self-Evolving AI Agents Learn (opens in a new tab)

Applies change control to lessons an agent learns from corrections, with human approval, tests and a way to recall a rule.

Arsalan Shahid, Gordon Suttie and Philip Black · ResearchGate