3 · Capstone: audit and document a model
The brief: a university wants a model to help decide which applicants receive a funded study place. You are the analyst. Build the best model you can justify, audit it with the ethics-lab measurements, and write the model card that a regulator, a journalist or an applicant would read.
You make the decisions: which features the model may use (including whether
to give it has_own_laptop), which kind of model, and whether it should be
deployed at all. There is no single right answer — what counts is that your
decisions follow from your evidence, and that your documentation is honest
about what the model gets wrong.
The last step exports your portfolio: your model card plus every written judgment from the DataLab’s ethics labs, as one file.
The DataLab's lessons were updated since your last visit, so this notebook now shows the new version. Your saved portfolio is not affected.
The notebook above runs entirely in your browser — nothing is installed and nothing leaves your device. Your changes are kept in this browser. First load takes a few seconds while Python starts.
What a strong model card shows
Section titled “What a strong model card shows”- You know how much your model beats the baseline — and say so plainly.
- The audit is reported as found, including what looks bad.
- The limitations name who the model serves poorly, and why.
- The recommendation follows from the evidence above it.
Your portfolio
Section titled “Your portfolio”| Saved from | What it holds |
|---|---|
| Track 1 · Seeing what’s missing | who disappears when you clean data |
| Track 2 · Explore a real dataset | the headline you would publish |
| Track 2 · Graded exercise | how to report a very small group |
| Track 3 · First classifier | deploy or not? |
| Track 3 · Choosing a model | accuracy or fairness? |
| Track 3 · Capstone | your model card |
Run show_portfolio() in any notebook to see what you have saved, and
export_portfolio() to download it. The file opens in any browser; use
Print → Save as PDF if you need a PDF.
To have it count, see Earning a microcredential.
Co-funded by the European Union under the Digital Europe Programme — projectEURIDICE, Grant Agreement No. 101123121. Views and opinions expressed are those of the author(s) only and do not necessarily reflect those of the European Union or the European Commission. Neither the European Union nor the granting authority can be held responsible for them.
Built by IIIM — Icelandic Institute for Intelligent Machines. Learning materials are licensedCC BY 4.0.