Summary
The episode works through four questions that students find difficult when they use AI assistants to write and check R code. It explains why AI-written code can run without any message and still give wrong answers, with a denominator error built on the real 2021 Canadian Social Connection Survey counts and a public health reporting case in which records were lost between files and central systems. It then examines what makes a second method independent, how to place known-answer tests at a threshold, and how the garden of forking paths multiplies false positives, using a multi-step calculation. A segment on data governance considers what may be shared with an approved assistant, including data dictionaries, synthetic stand-ins and First Nations data under the OCAP principles. The hosts close by debating whether beginners need to write R themselves and agree on a practical rule of predicting each block's output and explaining each line before saving it.