These six cases are constructed teaching composites. Their people, teams, and events are fictional; they are not historical accounts or evidence about any particular field, institution, or person. They complement the module’s documented cases by giving students short, incomplete situations in which to test a Charter procedure. Each case deliberately leaves relevant facts unknown. The goal is a workable response under uncertainty, not discovery of a predetermined moral answer.
A team is assembling a literature-background section due the next morning. Priya uses a generative-AI tool permitted by the course and adds a note saying, “AI-assisted first draft.” The prose is polished, so Mateo begins integrating it while Noor works on the analysis. Mateo then discovers that two cited articles do not exist and three real articles do not support the claims attached to them. Mateo and Noor spend nearly two hours checking every citation and rebuilding several paragraphs; their planned analysis review is postponed. Priya says she understood “first draft” to mean material the team would verify together and did not intend to present it as submission-ready. The team had agreed that AI was allowed and that use must be disclosed, but it had never defined the checks required before work entered the shared draft or how repair labor would be recorded and reassigned.
Write one SBIF comment and one provisional AI/verification norm for the Lecture 1 Charter-norm exit ticket. The norm must name what “ready to share” means, who checks what, and how repair work triggers a workload reset; carry it into the Lecture 2 core and Lecture 3 Appendix A.
Two minutes to mark observation/inference/missing information; three minutes to draft feedback; three minutes to write the norm; one-minute share-out.
Disclosure is relevant but does not by itself settle responsibility, and a flawed draft does not establish deception or justify a categorical judgment about AI. Keep attention on evidence, shared definitions, downstream labor, and repair. Do not use an AI detector as a substitute for checking claims and sources.
Eight minutes before a project submission closes, Lin notices that a shared calculation treats a measurement recorded in milligrams per liter as though it were grams per liter. The value feeds a figure and may change the team’s central conclusion, but nobody can establish the full effect before the deadline. The submission coordinator proposes a two-to-one vote to upload the current version and correct it later if needed. Another member argues that a vote cannot determine whether a calculation is valid. The instructor has not answered an urgent message, and the team does not know whether a late or explicitly qualified submission would be accepted. Its draft Charter says “major decisions are made by majority vote” and “all work will be quality checked,” but it names no pause trigger, decision authority, documentation step, or escalation path for an unresolved high-consequence error.
Rewrite the Lecture 2 core Charter concern clause. Include the trigger, temporary action, responsible person, evidence record, time limit, decision authority, escalation route, and revision step.
One minute to decide what is known; three minutes to map authority and options; three minutes to rewrite the clause; one-minute test by another pair.
There may be more than one defensible action because course rules and technical consequences are incomplete. A robust clause separates a scientific or safety judgment from preferences that may appropriately be voted on and keeps an unresolved concern visible rather than silently treating deadline pressure as evidence.
A team adopts a paid, visually dense project workspace because two members already know it and can activate a free trial. They begin recording meeting decisions, task assignments, and source notes there. After missing two updates, Ren tells the team, “This interface does not work reliably with my access setup. I need an equivalent way to read and enter information.” Ren does not offer personal or medical details. One teammate proposes sending occasional summaries; another offers to screen-share the workspace during meetings. Neither proposal lets Ren search prior decisions or contribute asynchronously, and the free trial will soon require each member to pay. Migrating the records would take time and might lose links. The team selected the tool informally and has no agreed criteria for accessibility, cost, privacy, onboarding, data export, or a fallback if an essential service becomes unavailable.
Draft an Appendix A access-and-fallback procedure covering equivalent participation, cost, trial testing, decision-record location, export, migration owner and deadline, onboarding, and escalation when no suitable tool is immediately available.
One minute to identify the current harm; two minutes to generate functional criteria; three minutes to draft the procedure; one-minute check for equivalent access.
Focus on access requirements and team infrastructure, not on asking a member to justify them through private disclosure. Familiarity and convenience are relevant criteria but not automatic decision rules. Avoid assigning all accommodation, migration, or training labor to the person who encountered the barrier.
A clinician, an engineer, and an ML researcher are developing a model-assisted sensor for a class project. At a progress meeting, all three report that the prototype is “validated.” In the ensuing discussion, they discover that the engineer means the sensor met specified calibration and repeatability tolerances on the bench; the ML researcher means the model reached a prespecified score on held-out data; and the clinician means evidence supports reliable use for the intended population and decision context. Each has completed the test they believed the project plan required, but the tests answer different questions and rely on different assumptions. A project partner now asks whether the device can be described as “validated” in Friday’s presentation. The Charter promises a “validated prototype” but does not define the claim, its evidence boundary, who integrates results across domains, or what language to use when one layer passes and another remains untested.
Add to Appendix B a shared-glossary entry for “validated,” a two-sentence shared outcome, the evidence required at each layer, an integration owner, and an “ask before assuming” or pause rule.
One minute to distinguish the claims; two minutes to map evidence layers; two minutes to draft the glossary entry and rule; one-minute read-back by another discipline.
Do not treat any one definition as inherently superior or imply that these examples exhaust how a discipline uses the term. The useful move is to attach scope, intended use, evidence, and responsibility to an ambiguous label before it becomes a public claim.
Kai developed the initial research question and wrote the first manuscript draft. Leila led data collection, performed the main analysis, and rewrote the results after a method changed. Both entered the project expecting to be first author, but the team never recorded what that expectation meant or when it would be reviewed. During final checks, Morgan independently reruns the analysis, finds a version mismatch, reconstructs the provenance record, and verifies every corrected table. This work prevents the team from reporting outdated results but is barely visible in the current contribution summary. One member proposes co-first authorship for Kai and Leila; another says the first listed name must be chosen immediately. The team has not checked the target journal’s policies, discussed corresponding-author responsibility, or distinguished author eligibility and order from acknowledgments, patent inventorship, data ownership, and other forms of recognition.
Create or revise the Appendix C provisional contribution matrix. Record current contributions, remaining responsibilities, possible recognition forms, venue-policy checks, a review date and triggers, decision participants, and a neutral consultation/escalation route.
Two minutes to separate the credit categories; three minutes to populate the contribution record; three minutes to specify review triggers and process; one-minute check for invisible labor.
Contribution recording improves the conversation but does not mechanically calculate author order or establish patent inventorship. Do not assume co-first authorship is always available, that verification automatically warrants authorship, or that early expectations cannot change. Preserve both contribution-specific recognition and shared accountability.
A team is preparing its final submission when Dana, who administers the analysis environment, reports that a recalibration check failed on one dataset. The cause and effect are unknown. The most senior student replies that the concern has been heard, the deadline is close, and the team must now submit; no one records who is authorized to close the issue or what evidence closure requires. Dana is also the only person with the service credentials and enough pipeline knowledge to reproduce the check. In a variant introduced halfway through discussion, Dana becomes unexpectedly unavailable after reporting the failure. The other members have exported results but not the environment configuration, decision log, or recovery instructions. Their Charter encourages dissent and says files will be backed up, yet it contains no visible-concern register, temporary pause rule, backup role, credential-recovery path, or procedure for operating when a critical expert is absent.
Revise the final Charter package with two executable clauses: an open-concern procedure specifying trigger, owner, visibility, closure evidence, authority, deadline, and escalation; and a redundancy procedure specifying backup people, access, records, rehearsal, and recovery.
Two minutes on the original case; announce the unavailability variant; three minutes to test the Charter; two minutes to revise both clauses; one-minute all-member read-back.
The scenario does not prove that submission must always stop or that seniority is never relevant. It tests whether authority, evidence, and escalation are explicit under pressure. Redundancy is not merely copying files: it includes usable access, current knowledge, privacy and security safeguards, and a practiced recovery path that does not depend on one person.