Tony Sale
@asale
Epistemic Engineer and independent AI researcher working at the intersection of epistemology, AI safety and ethics, large language models, research methodology, and human-machine knowledge. I design and test practical frameworks for evaluating the reliability of AI-assisted knowledge production, with a focus on evidence provenance, source fidelity, epistemic robustness, reproducibility, and human oversight. I combine philosophical training with hands-on experience in research, LLM workflows, technical communication, and project development, with a focus on turning abstract epistemic problems into testable methodologies and practical tools.
www.tonysale.com$0total balance
$0charity balance
$0cash balance
$0 in pending offers
About Me
RESEARCH & INDEPENDENT PROJECTS
Epistemic Engineering & AI Research | Independent Research | 2023 - Present
• Researching epistemology, ethical AI, trustworthy AI, human-machine knowledge, and the conditions required for reliable AI-assisted knowledge production.
• Developing the Compounding Knowledge Protocol (CKP), a computational framework for structured and verifiable knowledge synthesis across contested research questions, with explicit claim dependencies, evidence provenance, source fidelity, verification, and reproducibility.
• Conducting empirical research on epistemic robustness in large language models, examining how social pressure, authority cues, contextual framing, and protocol variations can affect AI-generated judgments.
• Developing the Epistemic Calibration Audit, a pilot methodology for testing whether frontier LLMs maintain epistemic robustness when exposed to social, authority, and contextual pressure.
• Developing the Protocol Drift Atlas, an open toolkit for measuring protocol sensitivity and detecting systematic changes in LLM safety judgments across evaluation conditions.
• Developing the Aletheia Plurality Audit, a research framework for examining plurality, agreement, and divergence in synthetic governance discussions generated by AI systems.
• Designing research workflows that combine LLMs with explicit verification, structured evidence, provenance tracking, independent evaluation, and human judgment rather than treating model output as authoritative.
No comments yet. Sign in to create one!