AI Policy

This course neither bans AI nor waves it through. The policy opens up in three stages across the term, tracking what you are able to judge for yourself. Whenever you are unsure whether you may use it, this page is the answer.

Why it is off limits for the first seven weeks

This is not a discipline rule. It is a question of what has to come first. Weeks 1 to 7 are where you develop a feel for code and the ability to debug — where you can look at a fragment and predict what it will print, or read an error and guess which line caused it. You only get there by writing code yourself, getting it wrong yourself, and tracking down the reason yourself.

Hand that stage to an AI and you end up with a pile of code that runs and no way of telling whether it is right. From week 8 onward, that judgement is the single most valuable thing you bring to working with an AI. Models produce code that looks entirely reasonable and falls apart on an edge case. Without those first seven weeks, you will not spot it.

What the declaration should say

From week 8, attach a short declaration to every submission. Three things:

1. What you used — which tool, which model

2. What it produced — name the functions or the specific blocks

3. How you checked it — the tests you ran, the edge cases you tried

The declaration itself costs you nothing. Leaving it out does — because it means you never stopped to think about the division of labour between you and the tool.

Where the line is

Using AI to generate code during Level 0; submitting code you cannot talk through; being unable to explain your own implementation when a TA asks. All three are handled as academic misconduct.

The test is simple: can you answer for every line you hand in?