End-of-term feedback piles up in three places: Canvas grade and rubric data, Qualtrics surveys, and Google Classroom discussion threads. Most instructors read it once, set it aside, and start the next term. Dock runs a synthesis agent across all three sources, drafts a course-improvement memo with each claim tied back to a row, and routes it to the instructor and department head for review. The agent does the reading. The humans keep the judgment, and the trail shows who decided what.
Canvas, Qualtrics, and Google Classroom stay the system of record for the raw data. Dock is the system of record for what the agent interprets. Each Dock row carries a pointer back to the platform record, agent identity, decision, reviewer, and timestamp. The agent re-fetches platform data via fresh API reads when it needs current state.
Feedback synthesis surface
| Theme | Source rows | Agent draft | Severity | Instructor review | Dept head review |
|---|---|---|---|---|---|
| Week 6 problem set too long | Qualtrics Q12 (n=34), Canvas median time 4.2h | Trim two problems, redistribute to wk 7 | High | Accept | Accept |
| Office hours timing conflict | Classroom thread #1188, Qualtrics Q7 (n=19) | Add a Thursday evening slot | Medium | Accept with note | Accept |
| Group project grading variance | Canvas rubric stdev 0.41, Qualtrics Q15 | Rubric calibration session before next term | Medium | Defer to dept | Accept, schedule |
Each row links to the underlying Canvas assignment, Qualtrics response set, or Classroom thread. The severity and the recommendation are the agent's interpretation, attributed to it. The two review columns are the humans.
Worked workflow
The course closes. The agent pulls Canvas grade distributions and rubric variance, Qualtrics end-of-term survey results, and Google Classroom discussion volume by week. It clusters comments by theme, cross-checks each theme against outcome data, and writes a memo with one row per recommendation. The instructor reviews row by row, accepting, editing, or rejecting. The department head sees only the accepted set and signs off on the ones requiring scheduling or budget. Six months later, when the next term's evaluations come in, the agent re-reads the accepted memo and reports on whether the changes moved the numbers. Every step is attributed.
Why it matters
Feedback that nobody synthesizes is feedback nobody acts on. Qualtrics reports that only sixty percent of students feel their feedback receives serious consideration, and that institutions on its platform have saved thousands of staff hours by automating the read.1 Carnegie Mellon's Eberly Center frames evidence-based teaching improvement as the core of its faculty support work.2 Dock is the layer that lets a department run that loop at scale without losing the audit trail. The instructor sees the agent's reasoning. The department head sees the instructor's review. The next cohort sees the change. Related patterns sit in Dock for education and Dock for research, and the review trail follows the model in agent audit and compliance and agent identity.
Start a free Dock workspace and connect Canvas, Qualtrics, and Google Classroom.
FAQ
Does the agent grade students? No. The agent reads grade distributions and rubric variance as signals. Grading stays with the instructor.
What if the instructor disagrees with a recommendation? The instructor rejects or edits the row. The rejection is logged with reason. The department head sees only what the instructor accepted.
How is this different from a people-ops review? The shape is similar, but the data sources and reviewers differ. See Dock for people ops for the personnel version.
Who is accountable for the synthesis? The agent is named in the row, and the reviewing human signs off. Identity attribution follows the pattern in agent identity.
