Research Guides
Test the Time Left to Finish a Research Review
A researcher can spend several minutes comparing sources without typing. If a review form treats that reading time as abandonment, a session warning can arrive just as the researcher is ready to explain a finding. Test the time available to finish the task, not merely whether the warning appears.
This guide proposes a timeout review for research workflows. It does not claim that Colquhoun.AI provides a particular session-extension or draft-recovery feature, and it is not a full accessibility or security certification.
1. Establish which time limit is being tested
W3C's Timing Adjustable guidance describes alternatives for content-set time limits: switching them off, adjusting them over a sufficiently wide range, or receiving a warning and a simple extension opportunity. For the extension alternative, the criterion specifies at least 20 seconds to respond and at least ten extensions. Defined exceptions include essential limits, unavoidable real-time events and limits longer than 20 hours.
A security-related label alone does not establish an exception. Identify the particular limit and evaluate the applicable requirements with the people responsible for accessibility and security. Do not disable account protections simply to make a usability test pass.
2. Trace a fictional source-review session
Imagine a fictional editor where a researcher enters a comparison note, opens a source in another tab and returns after reading it. The editor warns that the session will end. The researcher activates “Keep reviewing,” but the original expiry still clears the form moments later.
The button's presence did not establish a working extension. The reviewer records the warning time, the action time and the later expiry. In an isolated test environment, they check whether the server actually accepted an extension rather than only whether the interface dismissed the warning.
A repair should reconcile the visible message with the effective session state. Retesting should also check what happened to the draft. Extending access and preserving an unfinished note are related user needs, but they are not the same technical outcome.
3. Use a small timing test plan
- Start condition: sign in with an authorized test account and begin a harmless draft.
- Task: include realistic reading time away from the input field.
- Warning: record when it becomes available and what it says will expire.
- Response window: measure the time actually available to take the offered action.
- Extension: confirm the resulting effective limit, not just the closed dialog.
- Repeat: test the documented extension behavior across successive cycles.
- No response: observe the separate path where the warning is ignored.
Do not infer a successful full-duration test from a shortened demonstration unless the result clearly says it used accelerated timing. A local fixture can expose a logic error quickly, but production configuration still needs a separate, authorized verification.
4. Record enough evidence without recording secrets
Keep event times, safe test identifiers and the visible states needed to reproduce the failure. Exclude session cookies, access tokens and private customer notes from screenshots and shared logs. Use synthetic draft text so the test remains useful without exposing someone's research.
Distinguish session expiry from a network failure or a background job stopping. If the interface says “Your session ended,” but the evidence only shows an unreachable server, preserve that uncertainty rather than recording a definite authentication outcome.
5. Agree on task-level acceptance
A useful acceptance note identifies the tested browser and timing configuration, confirms that the offered extension had its promised effect, and describes the ignored-warning outcome. Record whether the warning and action were operable through the tested input method.
If restoring a draft requires signing in again, test that as a separate authorized journey. Do not call a draft recovered merely because the editor reopened, or call a session extended merely because an alert disappeared.
Keep unresolved questions visible: other devices, multi-tab behavior and untested expiry policies should not inherit a passing result. This makes a timeout finding repairable while keeping the accessibility evidence and the security decision appropriately distinct.