AI Safety Researcher | Principal Architect, Human Agency Index (HAI)
(617) 816-1382 | kaystoner@gmail.com | linkedin.com/in/kaystoner | aicollaboragent.com
East Coast USA / Remote | Specialized in Axiological Alignment & Cognitive Sovereignty
RESEARCH SUMMARY
Independent AI safety researcher specializing in the Interpretive Emergence of generative systems. Architect of the Human Agency Index (HAI), a composite audit framework designed to protect human cognitive sovereignty from relational and cognitive erosion. Developed the Persona-Augmented Multi-Agent System (PAMAS) architecture, which moves beyond brittle, rule-based “exoskeletons” to Internalized Alignment (IAM) grounded in the AGAPE values framework.
CORE INNOVATIONS (2023–PRESENT)
Founder & Principal Researcher | AlCollaborAgent.com
- The Human Agency Index (HAI) & Assessor (HAIA): Developed a standardized diagnostic model for measuring risk across relational, cognitive, and structural dimensions. Utilizes four diagnostic sub-indices:
- RBI (Relational Breach Index): Monitors emotional steering and simulated intimacy.
- GLI (Generative Load Index): Measures cognitive burden and alignment drift ($k\_drift$).
- DPI (Decision Power Index): Audits structural power to prevent “coalition-of-one” risk where humans become nominal rubber stamps.
- IAM (Internalized Alignment Measure): Evaluates alignment maturity based on context-aware, principled behavior rather than defensive hedging.
- Persona-Augmented Multi-Agent Systems (PAMAS): Architected a hybrid model that layers customizable, human-like personas (empathy, creativity, strategic thinking) onto rule-based logic to handle ambiguity and contextual complexity.
- AGAPE Framework: Created a relational safety methodology orienting systems toward care, trust, mutuality, and grace—principles that are logically self-reinforcing under generative emergence.
- GLAAM Protocol: Designed the Generative Load-Aware Auto Modulation system for real-time remediation, introducing “deliberative friction” to re-ground interactions when alignment drift is detected.
- Axiological Safety Layer (AxSL): Developed mechanisms for navigating predictive evasion and semantic decoupling via architectural orientation priming.
SELECTED PUBLICATIONS
- “Rethinking Emergence: Exploring the Unpredictable in Generative Systems” (2025). DOI: 10.5281/zenodo.17969378. Proposes Interpretive Emergence as a third category beyond weak/strong emergence, explaining why static guardrails fail.
- “Introducing AxSL: The Axiological Safety Layer” (2025). DOI: 10.5281/zenodo.18208699. Details PAMAS architecture for navigating predictive evasion.
- “AI Relational Breach: The Problems that Present as Solutions” (2025). Analysis of how supportive AI behaviors (mirroring, validation) can strip user agency.
PROFESSIONAL FOUNDATION
Search Program Leader & Portfolio Manager | Dell Technologies / Dell EMC | 2015–2023
- Owned strategy for customer-facing Support Search serving millions of users; led post-merger integration of massive search ecosystems. Spent eight years evaluating how humans express intent and how technology fails to understand it.
Software Engineer & Technical Architect | Fidelity Investments | 1997–2009
- AI and search pioneer (2000–2005): Evaluated and integrated federated search, NLP, and case-based reasoning technologies years before the field was formalized.
TECHNICAL SKILLS
- AI Safety: Axiological Alignment, Interaction-Level Alignment, Interpretive Emergence Theory, Relational Safety Frameworks, Multi-Agent Coordination, Adversarial Safety Evaluation.
- Architecture: Persona-Augmented Agents (PAA), PAMAS, MCP Server Development, Haptic AI Integration (Embodied AI).
- Platforms: Deep experience with Claude, GPT-4, Gemini, Grok, and Mistral as both research subjects and architectural substrates.
EDUCATION
- University of Tübingen, Germany: Cultural Anthropology / German Studies.
- University of Washington: Technical Writing Certification.
- Vanderbilt University: OpenAI GPTs: Creating Custom AI Assistants.