Where the Receipts Come From

PERS 1506, From Transcript to Track Record, is a one-credit course at Columbus State University in which about a hundred students learn to name a skill, prove it with a specific moment, and say it out loud to someone who could hire them. It reached its midpoint on October 6. This note records what the first eight weeks show. The figures are counted across the whole class; no student is named, and no student's words are quoted. These are observations, not findings.

The numbers

The Midpoint Portfolio closed on October 6. Of the 93 students on the roster at Week 8, 76 submitted, and 73 of those included all four revised cards. Thirty-one earned full marks; the median score was 13.5 of 15. The syllabus that governs the course is posted as Academic Credit, Professional Credibility.

The receipt

Reading the 76 portfolios together, the part of the work that seems hardest is the part the course was designed around. Most students can name a skill, and most write in a voice that sounds like their own: 63 of the 76 read that way to me. The record thins at the receipt. On the evidence criterion, 39 portfolios were strong, 27 were still developing, and 10 made a claim without a specific moment beneath it. One tentative reading is that naming can be taught in eight weeks and proving takes longer. The second half of the term will test it.

Where the receipts come from

The pattern I did not expect concerns where the receipts come from. A rough keyword count across the 304 midpoint cards puts 175 in work or life alone, 50 in both, 50 unclear, and 29 in coursework alone. About one card in ten comes from a class by itself; about one in four mentions a class at all. Where a class does appear, it is mostly in Module 4, as the setting for a setback, a failed exam or a missed deadline, and almost never as the place a skill was built. A course named for the transcript set out to teach students to translate coursework into employer language, and so far they have translated almost everything except coursework. The few coursework-only cards also read, to my eye, as some of the strongest: a graded presentation, a written assignment scored against a rubric, an independent project, each specific enough to be believed.

I can read the gap two ways, and the course's own design points to the second. Life examples may be more vivid, and the course rewards vividness. But the course also never models the move. Every walkthrough example comes from a job or a hobby, and one module asks outright for a skill from outside the classroom. Students did what the examples showed them. The missing move looks like translation rather than selection: they do not lack class examples, they do not yet believe a hiring manager wants to hear about a lab or a problem set. The change I expect to make next time is a walkthrough that takes one assignment and writes it three ways, as the student would say it, as a transcript line, and as an employer would hear it, and a coursework card in the module that already asks what an employer should know, with the job posting beside it.

The AI gap

The AI numbers are harder to interpret. The syllabus asks students, twice, to have an AI draft a version of their skills and then mark every line that is not specifically true of them. Asked at the midpoint how they had used AI, 44 of 76 reported not using it at all. The 32 who did used it mostly on the first card, on module revisions, or on the midpoint revisions themselves, with a few mentions of coaching comments or a discussion post. I do not yet know what to make of the gap, and I would rather say so than guess. Several explanations are plausible, and students have offered some of them, in this course and outside it. Some describe holding back on purpose, writing plainer sentences or even misspelling a word deliberately, because they worry about being accused of using AI, and because the detectors their instructors rely on do not work the way those instructors seem to think. Others have said the course asks for their own story, so they feel little need to perform. Some say they simply do not trust the tools. If those accounts hold, students who skipped the exercise may not have been skipping the work. Among those who did use AI, the reflections suggest they are beginning to notice where their own judgment comes in, and the comments I have left them ask for the same thing in each case: the moment that makes the card theirs.

What the AI got wrong

What students reported the AI getting wrong may turn out to be the most useful part of the midpoint record, and the accounts are consistent. It tended to make them sound more formal than they speak. It generalized the one specific moment that made a card true. In a few cases it supplied details that had not happened: a role a student had never held, or a problem described as solved when the student's own account said it was still being worked on. What they described doing about it is close to what the syllabus hoped for. They removed the invention, restored the specific moment, and rewrote the line the way they would say it aloud. The distinction between smooth and true appears to be one students can reach on their own, at least once they are asked to look for it.

What they want to learn

Asked what they would like to learn about AI, the largest group described the same thing in different words: how to use it without losing their own voice, and how to tell when it is wrong. A smaller group wanted to get better at asking it questions. A visible minority wants nothing to do with it and said so plainly; the environmental cost and simple distrust both came up. These responses will shape the second half of the term, and the December report will say what happened.