How I Made This Book

A Transparent Account of AI-Assisted Research in The Hidden Woman

As a researcher grounded in the Genealogical Proof Standard (GPS) and a mental health professional trained to recognize the limits of pattern recognition in human narratives, I approached the reconstruction of Mary Douglas’s life with both methodological rigor and ethical caution. The fragmented archives of Progressive Era reformers—particularly women who altered their names to escape stigma—demand a research strategy that balances technological efficiency with human discernment. This book was produced using AI tools not as a replacement for archival labor, but as an instrument to extend its reach, while subjecting every output to GPS-compliant verification.

AI in Service of the Genealogical Proof Standard

The Genealogical Proof Standard (GPS)—developed by the Board for Certification of Genealogists—requires five elements: reasonably exhaustive research, source citations, analysis and correlation, resolution of conflicts, and a soundly written conclusion. I implemented these standards in collaboration with AI in the following ways:

  1. Reasonably Exhaustive Research AI models were used to perform high-volume, cross-institutional searches across digitized archives (e.g., New York City Municipal Archives, Lilly Library, Presbyterian Historical Society), identifying potential source clusters such as Jarvie Service case files, Upton Sinclair correspondence, and Gertrude Tubby’s psychical research papers. However, I manually verified that no relevant collections were omitted and that all known repositories were queried.
  2. Source Citations AI assisted in generating standardized citations (Chicago Manual of Style, 17th ed.) for primary sources, including newspaper articles, death certificates, and unpublished correspondence. All citations were cross-checked against authoritative style guides and archival finding aids to prevent hallucination or misattribution.
  3. Analysis and Correlation Natural language processing (NLP) tools were used to detect patterns in handwritten notes, coded references (e.g., "Mary D." in Sacco-Vanzetti trial documents), and thematic recurrences across Sinclair’s and Warbasse’s networks. These tools flagged potential connections—such as shared addresses or overlapping references to "Red Mary"—but all correlations were manually reviewed for historical plausibility.
  4. Resolution of Conflicts AI surfaced inconsistencies in dates, names, and biographical details (e.g., discrepancies between 1916 Book News Monthly listings and 1940 U.S. Census records). I then applied contextual historical knowledge—such as the mobility of subscription sales workers in the early 20th century—to resolve conflicts, erring on the side of documented evidence over algorithmic inference.
  5. Soundly Written Conclusion: AI-generated draft narrative sections based on synthesized archival data. These drafts were treated as working hypotheses, not conclusions. I revised and edited each section to reflect the uncertainty inherent in reconstructing a life from scattered fragments, and all speculative language is clearly marked as such.

Risks and Benefits of AI in Archival Research

Benefits

  • Scale and Speed: AI enabled the rapid scanning of thousands of pages of digitized newspapers, government records, and private correspondence across multiple institutions—work that would have taken years manually.
  • Pattern Detection: It identified subtle connections, such as recurring references to "Gertrude Tubby" across unrelated collections, that might have been overlooked.
  • Accessibility: AI tools helped transcribe difficult handwriting in 19th- and early 20th-century documents, making previously inaccessible materials legible.
  • Cross-Referencing: It cross-referenced names, addresses, and dates across disparate sources (e.g., linking Mary Douglas in Washington, D.C., 1940, with Adelaide Branch in Monticello, 1913).

Risks

  • Fabrication of Sources: AI models are prone to hallucinating quotations, dates, and even archival citations. Every AI-generated citation was manually verified against primary or secondary sources.
  • Loss of Context: Archival silences—such as the absence of personal letters from Mary Douglas—cannot be detected by AI. I addressed this by acknowledging gaps and avoiding speculative reconstruction.
  • Bias Amplification: AI trained on historical corpora may replicate gendered or class-based biases in its outputs. I countered this by prioritizing primary sources and critically interrogating AI-generated suggestions.
  • Privacy Violations: Some archival materials contain sensitive personal data. I ensured that no living individuals were inadvertently exposed and that all published materials comply with privacy laws and ethical standards.
  • Over-Reliance on Automation: I treated AI as a research assistant, not an authority. All claims about individuals’ beliefs, motivations, or internal states are grounded in documented behavior or verifiable statements.

Ethical Framework: From Clinician to Scholar

My background as a mental health professional reinforced the importance of non-maleficence—doing no harm—and respect for persons in archival storytelling. The subjects of this book—women who lived in the shadows of Progressive Era reform, who changed their names, who were institutionalized, or who were erased from public memory—deserve narratives that are accurate, compassionate, and free from sensationalism.

AI cannot assess human dignity or ethical weight. It can, however, help us locate the fragments of a life. The reconstruction of Mary Douglas’s story required not only computational power, but also a commitment to narrative ethics—the principle that historical subjects are not puzzles to be solved, but people to be remembered with integrity.

Therefore, while AI assisted in data retrieval, pattern recognition, and draft composition, the final narrative reflects my judgment as a scholar and my ethical responsibility as a storyteller. All errors of fact, interpretation, or tone remain my own.

Final Note on Transparency

This disclosure is itself a form of scholarly accountability. In an era when AI-generated books are increasingly common, I believe transparency about tool use is essential to maintaining public trust in historical scholarship. I invite readers to evaluate this work not only by its conclusions, but by the rigor of the process that produced it.

Narrowsburg, New York
August 6, 2026

 

Tom Rue is an experienced researcher, genealogist, local historian, and licensed mental health professional with more than 40 years of combined experience in archival research, municipal history, and clinical counseling. His dual background is uniquely suited for reconstructing the biography of Adelaide Branch, combining professional-grade archival tracking with a trauma-informed clinical understanding of forced institutionalization, shame narratives, and psychological reintegration.

Historical and Archival Credentials

Rue's work demonstrates how genealogical discipline can restore agency to historically marginalized subjects without resorting to narrative inflation or fictionalizing—a model for intellectual biography rooted in both empathy and evidentiary rigor.

Historical Publications

 

----> Back to the project main portal page for The Adelaide Branch Biography Project