You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
$25,000 left to reach the funding goal.
As frontier AI systems transition from passive tools to multi-agent, autonomous workflows, standard governance and legal frameworks face a foundational bottleneck: they rely on outdated, anthropomorphic definitions of agency, intent, and causal responsibility.
This project bridges rigorous philosophical ontology (specifically drawing on Bound Perspective, Presentability, and Architecturalist Freedom) with practical AI governance. The goal is to provide a formal, non-anthropomorphic framework for evaluating model behavior, tracing causal liability in multi-agent networks, and shifting safety design from reactive output-filtering to proactive structural (architectural) constraints.
Over a 3-to-6 month grant period, this project will produce concrete, open-access public deliverables designed for policymakers, safety researchers, and AI labs:
Publication #1 (Month 2): "Attributing Agency to Autonomous Workflows: A Non-Anthropomorphic Liability Model." Establishing how functional bound perspectives map onto multi-agent accountability.
Publication #2 (Month 4): "Presentability and Proof: Epistemological Standards for Latent State Auditing." Proposing structural evaluation criteria for black-box model capabilities.
Policy Brief / EA Forum Series (Month 6): A concise policy brief synthesizing these insights for groups like the Australian AI Safety Institute (AISI) or independent governance researchers, published publicly on the EA Forum / LessWrong.
Research Stipend ($20,000): Dedicated living support allowing 100% focus on writing and publishing these papers over 6 months while preparing for undergraduate/postgraduate pathways.
Operations & Compute ($2,500): Academic literature access, subscriptions, and LLM API testing credits for evaluating model behaviors.
Fiscal/Administrative Buffer ($2,500): Micro-grant processing, open-access publishing fees, and contingencies.
Lead Researcher: Trent Trinder (Sole Independent Researcher).
Track Record & Conceptual Foundation:
I am an independent researcher specializing in ontology, metaphysics, and formal frameworks of agency. Over the past several years, I have authored an extensive original body of philosophical work (Primordial Presence), focusing specifically on inference, structural recognition, bound perspective, and presentability.
My track record demonstrates deep conceptual rigor, self-directed research discipline, and the ability to construct bottom-up theoretical architectures. This project represents the systematic translation of my existing philosophical research into practical, open-access white papers tailored for AI governance, agentic liability, and model evaluation bodies.
Likely Causes of Failure:
Translation / Communication Gap: The core concepts (e.g., Bound Perspective, Presentability) remain overly abstract and fail to resonate with mainstream AI policy analysts or technical safety engineers.
Distribution Bottlenecks: The published white papers do not receive sufficient engagement or traction on public forums (EA Forum / LessWrong) to reach policy decision-makers.
Bandwidth Limitations: Balancing full-time independent research production alongside university preparation delays publication timelines.
Outcomes & Countermeasures if the Project Fails:
Countermeasure: I will seek early feedback from active AI safety researchers and policy reviewers to iterate on language, ensuring jargon is translated into actionable policy frameworks.
Fallback Outcome: Even if high-level policy adoption is limited, the primary deliverables (3 open-access white papers and policy briefs) will be completed and published. These outputs will serve as my core academic portfolio for forthcoming university enrollment (BA in Philosophy) and future institutional fellowship applications (e.g., GovAI).
$0 USD.
This is my first grant application to Manifund or any philanthropic funder in the AI safety/governance ecosystem. All research completed to date has been entirely self-funded and conducted independently.