Attribution & process

This page credits the original writers, documents human contributions to the explorer, and lists the tasks assigned to AI.

Prepared through 2026-11-03. Project direction: Ian G. Williams, using the public credit name supplied for this page.

164human-name entries across local and source credits

78names credited in published local posts

14 + 8CRediT roles + GAIDeT task categories

Three layers of credit

Source writing: the credited people whose posts and linked works form the archive. Project contributions: the documented human direction and critique that shaped this analysis. AI delegation: tasks carried out with the agent and its tools. Being named in the archive does not make someone a coauthor of this analysis, endorse the findings, or establish responsibility for the code.

The source register includes every human-like credited name in the current 430-record inventory, including local writers, named authors of curated source works. Provisional account attributions remain labeled; name strings are not a verified census of distinct people.

CRediT / human contributions

Who contributed to this project?

CRediT’s 14 roles describe types of contribution; they do not decide who qualifies as an author. The table uses short paraphrases of the role descriptions. It maps the available project record, rather than inventing human work for every role.

Ian G. Williams set the research aims, supplied the export, requested the methods and interfaces, reviewed the evolving work, and identified the missing-author problem. The agent drafted and executed much of the implementation. Human direction is distinguished below from hands-on execution and independent verification.

Project-level CRediT mapping, grounded in this conversation and saved artifacts
CRediT role	In plain language	Documented activity and limits	Evidence status
Conceptualization	Set the study’s central aims.	Ian G. Williams defined the cohort-trend and archive-exploration objectives.	Documented direction
Data curation	Organize and annotate usable data.	Ian supplied the historical People directory and confirmed two identity pairs: Margaret/Maggie Schreiner and Anna Schlenz/Anna Corbett. Codex performed the wider extraction, crosswalk and record joins; a complete human record-by-record audit is not documented.	Documented source provision and two identity confirmations; agent implementation
Formal analysis	Apply quantitative or computational methods.	The agent wrote and ran the analysis recipes using Python, NLTK and MALLET. Human recomputation is not documented.	Human execution not established
Funding acquisition	Secure financial support.	No funding-acquisition contribution is established in the available record.	Not documented
Investigation	Gather evidence for the study.	Ian supplied the XML export and historical GCDI People directory. Codex consulted linked institutional sources and extracted directory evidence; separate human source-by-source investigation is not documented.	Human execution not established
Methodology	Choose and shape the approach.	Ian G. Williams requested academic-year cohort comparisons, NLTK/MALLET analysis, and a body-credit re-audit. Implementation details were agent-developed. Ian requested two simulated reviewer perspectives and incorporation of the additional persona review, then prioritized research/examination use. Ian requested the theme-oriented visualization revision; the evidence layers and counting definitions were operationalized by the agent. Ian confirmed academic-year interpretation for the historical directory; Codex operationalized separate role families and evidence-basis comparisons. Ian clarified that student guest authors, including GCDRI participants, may not have been Fellows at publication; these roles must be kept distinct.	Documented direction
Project administration	Coordinate the work.	Ian G. Williams set priorities through successive requests and revisions. Ian designated a public correction contact and retained licensing decisions. Ian selected the project title and authorized a private GCDigitalFellows repository under the authenticated perlsdiver account. Ian authorized public website publishing while keeping the source repository private. Ian required a README changelog for every update, with separate human and AI attribution. Ian instructed preservation and integration of shared GitHub/remote-terminal contributions, with his approval for significant changes. Ian requested repository guidance for assisting review and revision agents; the instructions were agent-drafted. Ian required explicit GAIDeT and CRediT attribution instructions for assisting agents.	Documented direction
Resources	Supply materials or tools.	Ian G. Williams supplied the WordPress XML export. Source writers’ contributions are acknowledged separately below. Ian supplied TTT-Website-review.md and the public contact details.	Documented provision
Software	Build and test programs.	The agent generated the Python/JavaScript implementation. No human programming contribution is claimed from the available record.	Human execution not established
Supervision	Direct and oversee the project.	Ian G. Williams directed the work and requested corrections; this does not establish line-by-line human review of all outputs.	Documented direction
Validation	Check claims, methods or results.	Ian G. Williams identified missing authors and required re-analysis. Numerical tests and browser checks were agent-run; independent human replication is not documented. Ian identified the incorrect account attribution for Linked Data for the Uninitiated (Part 1), named McMichael, and requested this archive-wide correction. Ian required complete article accounting and authorship vetting across the project; requesting this review is distinct from confirming every author.	Documented issue identification
Visualization	Prepare visual presentations.	Ian G. Williams requested interactive charts, annual design controls and the network. The agent implemented the visuals; human chart production is not claimed. Ian additionally requested incorporating themes into the visualizations. Thematic chart implementation and verification were agent-executed, not claimed as human chart production.	Human execution not established
Writing – original draft	Produce initial prose.	The analysis notes, documentation and this page were agent-drafted. Original post writers are credited for their source texts, not automatically as authors of this analysis.	Project human drafting not established
Writing – review & editing	Critique and revise the work.	Ian G. Williams requested clearer explanations, human-readable annotations, and attribution corrections. Human rewriting of every passage is not claimed. Ian authorized implementation of the review recommendations; this is not confirmation of every resulting claim.	Documented critical feedback

This is a documented-process mapping, not a signed contributor agreement. No lead/equal percentages are assigned. No historical writer was contacted for role confirmation. NISO recommends giving contributors an opportunity to review their assigned roles; these entries do not claim that confirmation has occurred. CRediT implementation guidance ↗

GAIDeT / delegated tasks

What did the AI agent do?

GAIDeT documents tasks delegated to generative AI. Here it complements CRediT: human contributions remain in the preceding table, while this table records agent-assisted tasks. Our crosswalk is an application to this project, not a claim that the taxonomies have identical roles.

Project-specific AI-use disclosure

Ian G. Williams directed an AI-assisted workflow using OpenAI Codex (OpenAI); exact underlying model/build not recorded in project artifacts. The agent helped develop methods, generated and ran code, investigated sources, prepared visualizations, drafted explanations, and performed automated checks. The tables below identify the delegated work and its limits.

Human oversight is evidenced by instructions, feedback, and correction requests. Full human verification of every record, model, code line, and statement is not established. Automated tests and agent inspection are not independent human review. The agent is not credited as a human author, and these taxonomies do not transfer accountability to software.

Earlier ChatGPT-assisted analysis is inherited from the project history. Exact model versions and every earlier session cannot be reconstructed from the saved artifacts. Ian identified the separate external review as Deepseek V4.1 Flash 1017 in the CAIL sandbox in OpenWebUI, using a previous-iteration ZIP. That model label is user-supplied; the exact reviewed archive and external session were not independently inspected. This disclosure covers the recorded workflow, not possible AI use by the original post writers. Their historical tool use was not investigated.

GAIDeT task mapping for the agent-assisted project workflow
GAIDeT category	Selected tasks	What occurred here	Oversight or limitation
Conceptualization	Idea generation; Formulating research questions and hypotheses	Suggested additional patterns and questions for exploration.	Ian set the objectives; generated suggestions are not established findings or tested hypotheses.
Literature Review	Literature search and systematization	Located design-history and taxonomy sources and assembled linked documentation. Checked institutional primary sources for the GCDI/Fellows orientation; distinguished inaccessible source pages from retrieved institutional descriptions.	Targeted source discovery, not an exhaustive systematic review.
Methodology	Research design; Selection of research methods	Developed corpus rules, attribution checks, topic summaries and network definitions.	Ian requested the methods and corrections; the agent proposed operational details.
Software Development and Automation	Code generation; Process automation	Generated and executed scripts, interface code, rebuilds and tests. Implemented research-first entry, author search, stable navigation, support documents and offline inspection. Added the thematic interface, pure counting functions, evidence exports and keyword-theme filtering/grouping for the existing index/network. Prepared Git version control, project identity and private organization-repository setup at Ian’s request. Prepared the separate public static website repository and GitHub Pages configuration, including source-link separation and deployment checks. Added complete-coverage verification and source-supported subset filters for the index/network, themes, topic summaries, word counts and writer sensitivity comparison.	Generated code and automated checks remain inspectable; no human code audit is implied.
Data Management	Data cleaning; Data curation and organization; Data analysis; Visualization; Reproducibility testing; Validation	Parsed the XML, prepared metadata, ran NLTK/MALLET recipes, constructed charts, corrected credits and compared outputs. Reviewed the 98-title missing-period list, checked five additional full posts and literal passages, built roster-coverage and credit-uncertainty displays, and preserved original/model baselines. Joined the saved title/tag and broader text flags, built thematic year and overlap counts and writer summaries, kept the 19-post essay layer separate, and checked thematic aggregates against the frozen audit. Organized source and saved evidence for version control while excluding local environments, caches and downloaded software dependencies. On November 2, screened all 430 records’ author fields/full-body credit signals; checked institutional guide page 23 and source biographies; corrected ten additional name records, grouped McMichael variants, and rebuilt attribution-dependent outputs while retaining corpus/model baselines. Reconciled all 857 export items and 436 ordinary posts, documented all 430 category-record decisions and six ordinary-post exclusions, checked named article references (four additional name corrections and five same-name confirmations), classified source-supported/declared-only/unresolved credits, and checked all browser populations and attribution joins.	NLTK, MALLET and counting code execute the recipes. Their numeric outputs are not LLM-generated observations. The latest attribution correction held the fitted corpus/model fixed.
Writing and Editing	Text generation; Proofreading and editing; Summarizing text; Formulation of conclusions	Drafted analysis prose, code comments, guides and this explanation; proposed the six-theme archive interpretation and its conclusions, with a separate scoped contribution declaration. Expanded the essay from 14 to 19 sources, qualified its conclusions, revised the research journeys, and documented the new processes and open decisions. Documented thematic definitions, denominators, attribution and sample-coding limits for nonadvanced readers. Updated the explorer name, README, citation, profile links and process documentation while preserving the source blog’s original title. Documented the second credit-review scope, unresolved cases and the requested error-correction publication. Drafted and edited explicit agent-guide instructions for separate human CRediT contributions and delegated GAIDeT tasks, provenance, confirmation limits and same-change attribution records.	The original archive posts were not generated by this project. Their historical AI use is not assessed. The external review text is attributed separately to the requester-identified Deepseek model; it is not claimed as Codex-generated prose.
Ethics Review	Ethical risk analysis	Considered attribution risks, unsupported author claims, and the separation of drafts and curators. Kept simulated reviewers distinct from actual humans, avoided unverified institutional endorsement and headcounts, preserved historical credits and left licensing for the requester.	This is a limited project-process mapping, not formal ethics approval or a compliance certification.
Supervision	Quality assessment; Identification of limitations; Recommendations	Performed agent-assisted checks and documented ambiguity, exclusions and limits. Ran two separate role-based AI reviewer sub-agents, assessed the supplied manifest-based five-persona simulation, synthesized recommendations and checked implementation. These activities are not human user testing. Tool separation: the supplied five-persona review was produced by Deepseek V4.1 Flash 1017 via OpenWebUI in the CAIL sandbox (requester-reported); Codex produced the two local sub-agent reviews and performed the synthesis and implementation. Checked theme counts, co-occurrence sets, coauthor counting and index integration; withheld unsupported essay overlap estimates. Verified metadata, source-bio and current-version credit distinctions; kept 258 provisional local credits and 53 older Sam account cases explicitly unresolved. Kept 249 account-only and 120 declared-only published credits distinct from 52 source-supported records; no blanket statement that all historical writers are confirmed.	This GAIDeT category names delegated checking tasks; the AI is not a human supervisor or accountable author.

NLTK supplies linguistic processing; MALLET fits statistical topic models; Python and JavaScript perform the saved calculations and interface behavior. These packages are not generative AI coauthors. The post-analysis button opens saved NLTK/MALLET results and performs literal browser search; it does not run a new topic fit or upload an article.

Selected task labels and identifiers follow the public GAIDeT taxonomy. The paper was published online in 2025 and appears in the 2026 journal issue; PubMed verifies the bibliographic record. This page used the public taxonomy and generator documentation; publisher full text was not accessible in this session.

The source-writing credit

All identified human post-author names

Every one of the 164 human-name entries is included below, without a top-author cutoff. The 78 names in the published-local network are part of this larger register; curated works add named source writers. Credit for an original linked work is separate from authorship of its local summary.

Credit to the source writers: We acknowledge the people listed here for the writing attributed to them in the underlying posts and source metadata. Their texts make this archive and its analysis possible. This acknowledgment does not imply their participation in, approval of, or responsibility for the new analysis.

For the source works, Writing – original draft may be a candidate CRediT description of a writer’s contribution, but a byline alone does not establish who drafted, edited, investigated, funded, or supervised the work. We therefore preserve source-writing credit without presenting unconfirmed individual CRediT assignments as facts.

Find a credited person or postCredit context

164 human-name entries shown.

Download the complete source-credit register (CSV) · Download credits and process disclosure (JSON)

Published source links, academic years, attribution status and literal byline evidence are available within each entry. Drafts are metadata-only, with no public link or article body. These are credits as recorded or reviewed, not verified identities or affiliations.