ml-risk-agreement.rst #15
Description
I found this doc: ml-risk-agreement.rst easy to read - which I wasn't fully expecting.
I would use this as a reference tool and provide to the researcher just after we've talked through it, and within the TRE as a reminder - but I'd probably add a few things like we expect a separate file of hyperparameters and results of any testing/accuracy results of things they've already done - or don't we need that? Or I would steal the bulk of this doc, attach it to the overall ML risk assessment doc we're using in the TRE and add to it with why we're asking the questions in that document. I don't know if that helps with wording around when to use it.
I'd not mention NDAs specifically, but say if any specific arrangements need to take place for TRE staff to look at code, this is outside the scope of the document.
It'd be useful to have links to examples of the recommendations as good practice,