The course-wide rules for late work, extensions, academic integrity and getting help.

These rules exist because of what this course grades: a genuine investigation trail, not a correct-looking final state. Each policy below follows from that, rather than restating generic university rules that already apply regardless of this course.

Investigation integrity#

The graded artefact is the investigation you actually carried out — observations, hypotheses, disconfirming tests, evidence, diagnosis, intervention, and verification, in the order you did them. Locating a fault through undirected trial-and-error and then writing up a hypothesis and evidence sequence that was never actually used to find it is a misrepresentation of the work, not a shortcut to the same mark. Assessments 2–4 use lightweight checkpoints and a running log for exactly this reason: not to audit every keystroke, but to fix an honest intermediate record of the investigation against which the final submission can be read.

Collaboration#

All four assessments are individual. Ordinary discussion of course material — what a review session covers, how the six-stage cycle works, what a term in a brief means — follows your university’s normal rules on collaboration versus misconduct; this page does not restate them. What must be your own work, regardless of how much discussion preceded it, is the observation, hypothesis, evidence interpretation, diagnosis, intervention, and verification you submit. Comparing notes on a system with a classmate and then submitting their diagnosis as your own reasoning is not collaboration — it fails the same test as retrofitting a hypothesis after the fact, for the same reason.

Tool use#

No assessment is graded on proficiency with a particular debugger, logging framework, or language — the course is explicit that tools are incidental to a given exercise, not its subject. What’s graded is why you chose a particular observation or tool, what evidence it produced, and how that evidence changed which hypotheses were still live. A dramatic-looking tool used without a stated reason earns nothing that a plain print statement used with one wouldn’t.

AI and coding agents#

AI tools and coding agents may assist with understanding an unfamiliar system, searching a codebase, explaining a language feature, or performing a mechanical code operation. They do not change what you are responsible for: you must be able to account for every piece of evidence you submit, defend every hypothesis you claim to have held, explain every intervention you made, and answer questions about your own reasoning. A conclusion produced by an AI tool is not itself evidence — if you use one, you must independently validate whatever it tells you before it enters your investigation trail, the same way you would treat an unverified claim from any other source.

Uncertainty#

Recording “unknown,” “unverified,” or “insufficient evidence to conclude” is an accepted, expected outcome of a genuine investigation, not a gap to be smoothed over before submission. A claim stated with confidence that matches the evidence you actually have is worth more than a confident claim the evidence doesn’t support — this applies to every assessment in the course, and is graded accordingly rather than penalised for honesty.