My use of AI is human-directed, source-based, and subject to explicit research constraints. I choose the subject, formulate the question or thesis, set the scope and purpose of the work, evaluate the evidence, and decide what conclusions it supports. Depending on the project, AI may help with drafting, organization, formatting, and reference compilation, but it does not replace my scholarly judgment. In the Indian Health Finance series, the writing is my own; AI was used for sources, organization, wording suggestions, and editing.
Intellectual Ownership
I take responsibility for the subjects I choose to study, the questions I ask, the scope and purpose of my research, and the decisions I make about what the evidence supports. I direct the research process, evaluate proposed sources and arguments, and determine the final form and destination of each paper.
The proportion of text written by AI is not, by itself, the measure of authorship. What matters is the intellectual purpose of the work, the research undertaken, my contribution to its analysis and interpretation, and my willingness to stand behind its claims and limitations.
I publish work under My Authorship when it reflects my substantive scholarly contribution and judgment. I use the AI Sandbox for exploratory work, AI-generated drafts, and other resources whose purpose or degree of original contribution makes that the more appropriate designation.
In either case, I remain responsible for representing the work honestly, checking its sources, and making its purpose and limitations clear.
Research Safeguards
I give AI explicit instructions intended to reduce errors and keep the work within the limits of the available evidence. The instructions vary by project and may include the following:
- Source boundaries: Use only material available on edfoxphd.com when that collection is sufficient for the task. When outside sources are permitted, identify them so I can evaluate them.
- No unsupported extrapolation: Do not extend findings, data, or conclusions beyond what the evidence supports.
- No invented information or citations: Do not fabricate facts, references, quotations, or source details. Distinguish verified information from uncertainty and inference.
- No forced symmetry: Do not make an argument artificially balanced, symmetrical, or logically complete merely to produce a smooth narrative. Preserve gaps, contradictions, and unresolved questions when the evidence requires it.
- Preserve uncertainty: Do not fill gaps in the record with plausible-sounding explanations. State when the sources do not support a conclusion.
- Proofreading and consistency: Check spelling, names, quotations, references, and factual consistency.
These safeguards reduce risk; they do not guarantee accuracy. I remain responsible for checking sources, evaluating claims, and deciding what the final paper can legitimately say. I do not assume that an AI-generated citation is valid merely because it looks complete, and I do not treat fluent or logically polished prose as proof that an argument is supported.
Paper-Specific Methods Note
Research constraints differ from one paper to another. Where useful, each paper should state which source limits and verification steps actually applied. The following is a template, not a claim that every safeguard was used in every project.
AI-assisted methods note
Research basis: [site-only / user-supplied sources plus approved outside sources / other defined source set].
AI role: [research assistance / outlining / drafting / editing / formatting / reference compilation].
Constraints applied: [for example, no unsupported extrapolation; preserve uncertainty; do not force symmetry].
Verification: [describe what was checked, including citations, quotations, names, and calculations].
Limitations: [identify unresolved questions, unavailable evidence, or claims that remain unverified].
The methods note should describe what was done for that particular paper, not what I generally ask AI to do. A source boundary or safeguard should be claimed only when it was actually applied.
Publication Paths
My Authorship: for work in which my substantive scholarly contribution, interpretation, synthesis, or argument is central and I am prepared to stand behind the conclusions.
AI Sandbox: for exploratory work, AI-generated drafts, background resources, and analyses whose primary purpose or degree of original contribution makes the Sandbox the more accurate designation. These materials should be labeled clearly and should not be mistaken for my authored scholarship.
Maintaining the Archive: How New Information Is Added and Verified
This website is a curated research archive, not a living news feed. When I consider adding new material—or checking whether existing material still holds—I apply a consistent set of tests.
First, does the new information strengthen the core purpose of the site? The site exists to help Tribal leaders, health directors, program administrators, and specialized researchers understand how Indian health programs are financed, organized, measured, and governed. Material is added only when it advances that understanding in a durable way.
Second, does the new information comport with what is already here? Before anything is added, I ask whether it is consistent with the population definitions, financing architecture, data limitations, and policy pathways already documented. Apparent conflicts are investigated rather than simply overwritten. Gaps, uncertainties, and unresolved questions are preserved when the evidence requires it.
Third, is the source legitimate, traceable, and appropriately weighted? I favor primary administrative data, official tables, Tribal-led analyses, and sources whose limitations can be stated clearly.
Fourth, does the addition respect the experiential knowledge already embedded in the archive? The site carries a point of view shaped by decades of work inside Tribal health programs. That experience includes practical cautions about how information is used with Tribal leaders and a preference for analyses that produce usable results.
Finally, is the addition honest about its own limits? Every significant addition should make clear what it does and does not establish. Where AI tools have been used, their role and the human judgments that governed them are stated. The distinction between work I stand behind as author and exploratory material placed in the AI Sandbox remains sharp.
New information is admitted only when it improves the archive’s reliability as a constrained, experience-informed knowledge base—one that others can use with confidence precisely because its boundaries and standards are explicit.