This research was scoped, sequenced, and produced by Luke Olson, Jonny Gose, and Kyle VanEtten, in response to a question from Jason Gaboury. The three of us built the working thesis question and the corpus list jointly, with falsifiability designed in from the start: if the primary thesis question could not be provably answered against the corpus, a parallel process was pre-defined to re-pose the question and run a different inquiry against the same evidence. In this case the primary question yielded a high-confidence positive answer and the fallback was not invoked.
An AI agent assisted with the wide retrieval and structured extraction pass across the corpus areas. The model used was a frontier-class large language model selected for its long-context handling and careful synthesis behavior. Mining outputs were never accepted as final claims: every bibliography entry was human-verified by Luke, Jonny, and Kyle; every quoted citation was checked against the rendered source text; every URL was confirmed live or explicitly marked partial-verification at retrieval. Sources that could not be verified at all were excluded from the synthesis rather than carried forward.
The synthesis was produced via a structured multi-corpus mining pass against the working thesis question — What are the most essential moves alumni relations can do to ensure strong institutional affinity AND robust alumni identity, and where do those moves accelerate giving and volunteering? The corpus areas surveyed were: CASE (Council for Advancement and Support of Education) public-summary content; Hanover Research's open-access alumni-giving and engagement reports; Salt & Wine's published brand-thinking; higher-education alumni programs (Notre Dame, Stanford, Wake Forest, with sharpened sub-questions on biggest gifts, donor loyalty, and event attendance); sorority/fraternity and private K-12 alumni research; Social Identity Theory's primary canon (Tajfel, Turner, Hogg, with Choi & Hogg's 2020 meta-analysis); brand community literature (Muñiz & O'Guinn 2001 foundational; Schau 2009 practice-theory); military lifelong-identity research; long-term religious-institution membership; and an alternative-drivers corpus (reciprocity, gratitude, repeated-positive-experience) used as falsification work to stress-test the thesis.
Each source was put through a verification protocol that required a working URL or DOI at retrieval, page-content over URL-slug for any extracted claim, page-attributed quotation for any direct citation, and explicit flagging of partial-verification (paywalled abstract only) versus full-verification (full text retrieved and read). Sources that could not be verified were excluded rather than synthesized.
One note on the bibliography. The original mining pass surveyed roughly thirty sources across the ten corpus areas above. The public bibliography is shorter: we looked at all of these, but not every source turned out to be instructional to the findings, and the bibliography lists only the sources that proved load-bearing. The fuller mining catalog (with verification flags, full-text vs. partial-verification status, and excluded-source notes) is available on request for any reader who wants to retrace the corpus work.
Confidence on the directional findings is calibrated at approximately 92%; the residual 8% lives in the "does this hold for Christian-mission-driven college-ministry alumni specifically" question, which would benefit from direct InterVarsity-data validation.