Reasonline AgencyThread
See how the work was made.
Students submit their finished work along with the process behind it: AI conversations, drafts, version history, and sources. AgencyThread reads that evidence against a published method and reports, dimension by dimension, how far the record shows the student framing, directing, evaluating, and taking responsibility for the work. You decide what students see.
A report describes evidence about one piece of work. It makes no finding about honesty or misconduct, and it never changes a grade.
Early Access. Pilots with founding instructors start in January 2027.
How it works
- 1
Set the assignment. Choose the evidence students must include and rate the task's stakes, reasoning demand, how much could reasonably go to AI, and what only a person can do.
- 2
Students assemble their evidence once, inside Reasonline or from your LMS.
- 3
AgencyThread writes a report that cites the evidence behind each dimension and says plainly what is missing.
- 4
You review it and choose whether to share it.
- 5
If you also use Assess, George can ask about specific pieces of that evidence.
What goes in an evidence packet
The student assembles one packet per assignment, and both AgencyThread and Assess draw on it. Each item records what Reasonline actually knows about it.
| Evidence | What Reasonline knows | Status |
|---|---|---|
| Final artifact | The file the student submitted. | EARLY ACCESS |
| Learn records | Captured by Reasonline while the work happened, with the conditions of that work attached. | EARLY ACCESS |
| ChatGPT and Claude conversation exports | Supplied by the student. Reasonline reads their structure but can't confirm they are complete. | EARLY ACCESS |
| Other conversation evidence (screenshots, pasted text) | Supplied by the student. | EARLY ACCESS |
| Drafts, notes, and sources | Supplied by the student. | EARLY ACCESS |
| Revision-history exports or screenshots | Supplied by the student. An uploaded Word file or PDF doesn't carry its original history. | EARLY ACCESS |
| Records from an authenticated integration, such as Google Docs version history | Retrieved by Reasonline directly. | IN DEVELOPMENT |
| Evidence that is unavailable or incomplete | The student states what is missing and why. | EARLY ACCESS |
A transcript that arrives through the LMS establishes how it was submitted. It doesn’t establish that its contents are complete or authentic.
The minimum an assignment needs
A finished artifact on its own can’t support a defensible report about the process, so each assignment sets a minimum: the final artifact plus the process evidence you require. You can make some process evidence optional.
Students get clear guidance on removing unrelated personal content from conversations before they submit, and they can state what evidence is unavailable and why. Missing evidence is reported as missing. It is never read as limited participation.
What you get back
The report leads with six dimensions, the evidence cited for each, and a short narrative. For every dimension it keeps three things apart: what the record directly supports, what the reviewer inferred, and what is missing or ambiguous. It also says what evidence would help. Three independent readings are compared; a rating stands only when two agree exactly, and any disagreement is shown in the report rather than hidden.
The composite Human Agency Index comes after the dimensions. HAI describes the extent of substantive human participation that the submitted evidence supports in one work process. It doesn’t measure artifact quality, correctness, learning, expertise, or effort, and it isn’t the share of thinking the student did.
AgencyThread — process report (illustrative)
D3 · Critical Evaluation & Calibrated Reliance
Supported: Rejected the AI’s market-size figure and replaced it with a cited source (conversation, turn 14; draft 2).
Inferred: The later pricing section reflects that correction.
Missing: No record of how the final recommendation was checked. A revision history would help.
How sharing works
- Reports are private to the instructor by default.
- You choose whether to share a report, one at a time or for a whole assignment.
- When a report is shared, the student sees the reasoning behind it and can correct factual errors. Corrections come to you.
- A report never changes a grade, and it is never sent to your LMS as a grade.
- The composite Human Agency Index and the Collaboration Profile appear in the instructor view only. Students see neither.
AgencyThread and Evidence Context do different jobs
Evidence Context records the conditions of the work: assignment rules, permitted assistance, individual or team setting, and the support Reasonline provided. AgencyThread interprets evidence of human participation: how the student framed, directed, evaluated, developed, and took responsibility for the work.
When the course also uses Reasonline Learn, AgencyThread can read relevant Learn records as well as imported evidence. A student who worked inside Reasonline shouldn’t have to reconstruct that work for a second product.
Explore Evidence ContextWith Reasonline Assess
When you have both products, George can add clearly marked process questions to an Assess dialogue. Each question is tied to a specific piece of the student’s submitted evidence, never to the report’s conclusions or scores. The core dialogue stays grounded in the assignment and the artifact, and the report and the dialogue remain separate records for you to compare.
A bounded AI dialogue supplies evidence about understanding and accountability. It can’t fill in process evidence that was never submitted, and some high-stakes uses may still call for a supervised or live conversation.
Where students submit
| Experience | What the student does | Status |
|---|---|---|
| Reasonline assignment | Submits the artifact and evidence inside Reasonline. | EARLY ACCESS |
| LMS launch into Reasonline | Opens a linked assignment from the LMS, then submits inside Reasonline.We name an LMS only after a live test with it. | EARLY ACCESS |
| Native LMS submission | Submits through the LMS's ordinary assignment page; the report returns inside the LMS.Canvas supports this pattern; we're investigating it in November 2026. | IN DEVELOPMENT |
Capabilities at launch
Each capability shows two things: whether you can use it, and how mature the claims about its output are.
| Capability | Access | Claim maturity |
|---|---|---|
| Assignment setup with instructor settings | EARLY ACCESS | — |
| Process report: dimensions, cited evidence, narrative, and what is supported, inferred, or missing | EARLY ACCESS | Claim maturity: Educator review before student use |
| Composite Human Agency Index (HAI), in the instructor view | EARLY ACCESS | Claim maturity: Validation in progress |
| Collaboration Profile, in the instructor view | EARLY ACCESS | Claim maturity: Experimental |
| Instructor chooses what students see | EARLY ACCESS | — |
| Student sees the reasoning and corrects factual errors | EARLY ACCESS | — |
| Process questions in Assess, when the instructor has both products | EARLY ACCESS | — |
| Course rubric result, kept separate from HAI | IN DEVELOPMENT | — |
| Institutional SSO | IN DEVELOPMENT | — |
Data and processing
Evidence is stored in Reasonline’s private storage and read by a third-party AI model provider’s API under Reasonline’s account, which doesn’t use it for training. PDF files are uploaded to the provider’s file service only for the evaluation and set to expire within an hour. Reports are kept in Reasonline, are never used as research data without a separate consent process, and are covered by student data export and erase requests. Only records Reasonline captured itself count as highest-trust evidence; student uploads, including parsed AI conversation exports, are treated as student-supplied.
See Trust & SecurityRequest Reasonline AgencyThread Early Access
Founding Early Access pricing is offered as a written quote while we measure the cost of a report on realistic student evidence. Department pilots are quoted against participants, assignments, and support. Every pilot needs its institution’s approval for the use of student records before students submit.
Reasonline gives instructors evidence of how the work was made and of what the student understands.
AgencyThread supplies the first. Assess supplies the second.