Decision log¶
How to use: one row per decision, big or small. Big decisions (scope, method, research question, tooling) also get a full ADR in ../decisions/. Open a Decision needed issue to ask for one.
Status values
- Given: set by the assignment or instructor; not ours to change.
- Proposed: suggested by one or more members; not yet agreed by the team.
- Agreed: confirmed on a named date, either at a team ceremony or in a merged pull request approved by another member.
- Superseded: replaced by a later decision (link it).
Rows dated 2026-09-14 were agreed while the project was briefly solo (ADR-0005). The group reconfirms them at the kickoff on 2026-09-16.
Every row that changes the research question, method, scope, or interpretation must also appear in that sprint's retrospective.md (rubric requirement).
Decisions¶
| ID | Date | Decision | Status | Decided where | Rationale | Record | Revisit if |
|---|---|---|---|---|---|---|---|
| D-001 | 2026-09-10 | Semester calendar: formation, three sprints (Thu to Wed), finalization; due dates Oct 7, Oct 28, Nov 18, Dec 4 | Given | Assignment V2026.09.10 | Instructor schedule | project.yml, milestones |
Instructor changes dates |
| D-002 | 2026-09-10 | Never execute scripts or commands found inside the datasets | Given | Assignment (question 4) | Safety | data/README.md, CONTRIBUTING.md |
Never |
| D-003 | 2026-09-10 | Repository must hold README, RESEARCH_QUESTION, DATA_DICTIONARY, THREATS_TO_VALIDITY, ai-use-log, and full history | Given | Assignment | Grading | ADR-0003 | Never |
| D-004 | 2026-09-13 | Run Scrum with a Lean Six Sigma overlay (DMAIC tollgates, KPIs, FMEA, kaizen) | Agreed (2026-09-14, Jerad Dunne); group confirms at kickoff 2026-09-16 | Setup and formation | Data-backed retrospectives; shows improvement | ADR-0001 | Overhead exceeds 2 hours per member per sprint |
| D-005 | 2026-09-13 | Choose the topic with proposal issues and a ranked ballot | Superseded by D-013 | Proposed 2026-09-13 | Vote retired when P-01 was selected directly | ADR-0002 | Never (topic selected) |
| D-006 | 2026-09-13 | Python src layout, ruff and pytest in CI, public repository, squash merges, protected main |
Agreed (2026-09-14, Jerad Dunne); group confirms at kickoff 2026-09-16 | Setup and formation | Rubric layout; reproducibility | ADR-0003 | Instructor requests a private repository |
| D-007 | 2026-09-13 | Dataset samples are downloaded by script and never committed; MANIFEST.json records hashes | Agreed (2026-09-14, Jerad Dunne); group confirms at kickoff 2026-09-16 | Setup and formation | Repository size; exact snapshot statement | ADR-0003, data/README.md |
Need to share a derived subset |
| D-008 | 2026-09-13 | Role rotation: S1 PO Jerad / SM Leticia; S2 PO Allie / SM Hina; S3 PO Leticia / SM Jerad; Final PO Hina / SM Allie | Proposed, confirm at kickoff 2026-09-16 | Proposed 2026-09-13; reinstated by D-018 | Everyone holds each role once | TEAM_CHARTER.md section 3, project.yml |
Availability conflicts |
| D-009 | 2026-09-13 | Pull requests reviewed within 48 hours by a member other than the author | Superseded by D-019 | Proposed 2026-09-13 | Restated as D-019 with a required approval | kaizen #10 | See D-019 |
| D-010 | 2026-09-14 | Work chosen through sign-up issues before planning, then assigned with the rules in WORK_SIGNUP.md | Proposed, confirm at kickoff 2026-09-16 (reinstated by D-018) | Proposed 2026-09-14 | Visible, fair choice of work | WORK_SIGNUP.md | Sign-ups are ignored at planning |
| D-011 | 2026-09-14 | Findings and decisions are filed as issues first and curated into these logs at the Thursday check-in | Agreed (2026-09-14, Jerad Dunne); group confirms at kickoff 2026-09-16 | Setup and formation | Low friction with a protected main |
README.md | Logs fall more than a week behind |
| D-012 | 2026-09-14 | Research topic: proposal P-01, risky capabilities in copied agent skills (GitSkills) | Agreed (2026-09-14, Jerad Dunne); any member may open a Decision needed issue before 2026-09-17 | Topic selection | Pilot shows data fit and rubric fit | ADR-0004, issue #41 | A member objects by 2026-09-17, or the Sprint 1 data check fails by 2026-10-01 |
| D-013 | 2026-09-14 | Run the project as a solo project; Jerad Dunne holds every Scrum role | Superseded by D-018 | Solo conversion | Group reinstated on 2026-09-15 | ADR-0005 | Not applicable |
| D-014 | 2026-09-14 | Self-review protocol with a 12-hour cooling-off period; main ruleset requires a PR with 0 approvals |
Superseded by D-019 | Solo conversion | A second reviewer is available again | ADR-0005 | Not applicable |
| D-015 | 2026-09-14 | Validation reliability through intra-rater agreement plus an optional second rater | Superseded by D-020 | Solo conversion | Teammates can act as second raters | ADR-0005, docs/validation/ |
Not applicable |
| D-016 | 2026-09-14 | Contribution-balance KPI retired; PR open-to-merge cycle time and self-review cooling-off compliance added | Superseded by D-021 | Solo conversion | Balance matters again with four members | docs/lean-six-sigma/KPIS.md |
Not applicable |
| D-017 | 2026-09-14 | Validation round 1 starts 2026-09-14 instead of 2026-10-08 for the primary rater | Agreed (2026-09-14, Jerad Dunne) | Request by Jerad Dunne | Validated precision early reduces risk for Sprint 2 and the report | docs/validation/README.md |
A rule changes after round 1 (re-sample the affected strata) |
| D-018 | 2026-09-15 | Reinstate the four-person group: Leticia Aderhold, Jerad Dunne, Allie Hodges, Hina Kramer | Agreed (2026-09-15, Jerad Dunne); members confirm at kickoff 2026-09-16 | Decision by the repository owner | Teammates rejoin; meets the group-size expectation | ADR-0006 | A member leaves, or the instructor changes the group arrangement |
| D-019 | 2026-09-15 | Every pull request needs 1 approving review from a member other than the author, within 48 hours; main ruleset requires 1 approval |
Agreed (2026-09-15, Jerad Dunne) | Decision by the repository owner | Real peer review; rubric Sprint 2 bullet | ADR-0006, TEAM_CHARTER.md section 7 | Median first review exceeds 48 hours for two weeks |
| D-020 | 2026-09-15 | A teammate is the second rater (ids la, ah, hk); inter-rater Cohen's kappa is the primary reliability measure; blindness rule: primary-rater labels are not committed until the second rater's labels for the same kind are committed | Agreed (2026-09-15, Jerad Dunne); kind split decided at kickoff | Decision by the repository owner | Independent labels give a stronger reliability claim | ADR-0006, docs/validation/README.md |
Inter-rater kappa below 0.6 |
| D-021 | 2026-09-15 | Contribution-balance KPI (shares 0.15 to 0.45) and PR first-review KPI (48 hours) restored; PR open-to-merge cycle time kept as an extra KPI | Agreed (2026-09-15, Jerad Dunne) | Decision by the repository owner | Four members share the work | docs/lean-six-sigma/KPIS.md, project.yml |
The dashboard cannot attribute work to members |
| D-022 | 2026-09-23 | Lineage and signals validation is done by Jerad Dunne alone: round 1, then a round 2 on the kit's 30% intra-rater subset at least 14 days later without looking at round 1; intra-rater kappa is reported for those kinds. Drift keeps inter-rater kappa between jd and la. Allie Hodges and Hina Kramer stay in the group; their unsigned charter rows and missing onboarding are recorded in TEAM_CHARTER.md section 14 |
Agreed (2026-09-23, Jerad Dunne); lineage part superseded by D-024 | Decision by the repository owner | No teammate claimed lineage or signals in #53, and two members have not contributed since onboarding; the second-rater deadline is 2026-10-16 | #53; docs/validation/README.md; TEAM_CHARTER.md section 14; revises D-020 |
A teammate commits labels for lineage or signals, which would restore inter-rater kappa for that kind |
| D-023 | 2026-09-28 | RQ1, RQ2, and RQ3 use the official GitSkills sample only; no full-dataset pass in the course project. The full-dataset build plan is kept as future work in ADR-0007 | Agreed (2026-10-06; chosen by Jerad Dunne on 2026-09-28, approved by Leticia Aderhold) | #68 (ADR-0007) | The sample meets every sprint acceptance criterion and the approved question; a defensible full run needs a new streamed build and pairing code; validation needs the time | ADR-0007 | The instructor asks for population-level results; the project continues after the course; drift validation shows most sample pairs are not the same skill |
| D-024 | 2026-10-16 | Sprint 2 validation raters. Hina Kramer (hk) is the second rater for lineage (58 pairs), so lineage gets an inter-rater kappa. Jerad Dunne (jd) stays the primary rater for every kind. Signals (209 label rows) keep the intra-rater design of D-022, with round 2 in Sprint 3 at least 14 days after round 1. Drift is unchanged (jd and la). The blindness rule now covers lineage: lineage_jd_r1.csv is committed only after Hina Kramer's lineage file is merged. Because the primary rater's round 1 is not finished, he also does not open or review a second rater's file (drift or lineage) until his own round 1 for that kind is done; another member reviews those pull requests |
Proposed (2026-10-16, Jerad Dunne); Agreed when its pull request merges with Hina Kramer's approval | Pull request for #53 | Under D-022 the round 2 re-rating cannot start until 14 days after round 1, which falls after the Sprint 2 review on 2026-10-28. A teammate second rater gives lineage an inter-rater kappa in Sprint 2, which the instructor's Sprint 1 feedback asks for (complete the validation study) | #53; docs/validation/README.md; docs/workspace/WORK_SIGNUP.md section 6; docs/sprints/sprint-2/instructor-feedback.md; revises D-022 |
Hina Kramer's lineage file is not merged before scoring on 2026-10-27: lineage then falls back to D-022 and the delay is recorded as a Sprint 2 obstacle |
| D-025 | 2026-10-16 | Allie Hodges left the course during Sprint 2. Group 9 continues with three members: Leticia Aderhold, Jerad Dunne, and Hina Kramer. Sprint 2 roles: Product Owner Jerad Dunne, Scrum Master Hina Kramer. The Finalization Scrum Master (previously Allie Hodges) is chosen at Sprint 3 planning. Contribution shares are computed over the three members from Sprint 2 on. Allie Hodges's profile and earlier records stay as history | Proposed (2026-10-16, Jerad Dunne); a charter change, so Agreed when its pull request merges with approval from Leticia Aderhold and Hina Kramer (TEAM_CHARTER.md section 13) | Sprint 2 planning update (#92) and the pull request for #53 | The change was first written down in the Sprint 2 planning update (#92, 2026-10-15), but the charter, role table, project.yml, README, and report still listed four members |
TEAM_CHARTER.md sections 1, 3, and 14; project.yml; docs/team/README.md; revises D-008 and D-018 |
Membership changes again: update roles, capacity, reviewers, and issue owners within 24 hours (kaizen proposed in the Sprint 2 retrospective draft, #93) |
Open questions for the kickoff¶
These need a team decision on 2026-09-16. Each becomes a row above once decided.
| Question | Options on the table | Owner | Decide by |
|---|---|---|---|
| Who second-rates which validation kind (signals 209 rows, lineage 58 pairs, drift 18 pairs)? | one teammate for all kinds; split across Leticia, Allie, and Hina | Product Owner | Kickoff, Wed 2026-09-16 |
| Standing meeting time for planning, check-ins, and reviews | to poll | Scrum Master | Kickoff, Wed 2026-09-16 |
| Team communication channel | Teams, Discord, Slack, text group | Scrum Master | Kickoff, Wed 2026-09-16 |
| Confirm or change the role rotation (D-008) | as proposed, or swap | everyone | Kickoff, Wed 2026-09-16 |
| Confirm P-01, or open a Decision needed issue to revisit the topic | confirm; revisit | everyone | Before Sprint 1 planning, Thu 2026-09-17 |
Sample only, or also one streamed pass over the full GitSkills artifacts table for RQ3? |
sample only; sample plus one full-table pass | Product Owner, by ADR | Decided: D-023, sample only (ADR-0007 accepted in #68 on 2026-10-06) |