Thread by @rachelcoldicutt

Hey! Guess what, I’m still mulling over the National Data Strategy, and I think - thanks to @altrishaw - I have realised what the fundamental flaw is with the aim to use data to build a fairer society, and I need a hand, pls, to articulate that clearly and accurately. (1/n)

I’ve mentioned that the NDS doesn’t sufficiently differentiate between different kinds of data: data about things is viewed as the same kind of utility as data about people - in the sweep of the strategy, roads and bridges don’t seem to be that diff to eg suicide ideation. (2/n)

But that is only one part of the problem - the other part is that it doesn’t differentiate between different kinds of *decision*. And this is where I need a hand - I assume there are useful statistical and/or legal models here to describe this (?). (3/n)

(Caveat: I might be about to get this really badly wrong, but, um, this is why I’m sharing, because I’m doing this as a sort of terrible hobby, and could do with a hand. So be kind pls.) (4/n)

Collecting anonymous personal data to measure things (eg sex and gender diversity in the work place) is a fundamentally different activity to analysing personal data to determine things based on a person’s characteristics, but they often get lumped together. (5/n)

So, as I understand it, while it is important for a business to monitor and manage e.g. sex and gender diversity, it would be discriminatory for the same business to make a hiring decision or a decision about pay based on an individual's sex or gender.

yes? (6/n)

So I have 3 Qs:

1) What are these different models for data-driven decision making called?

2) Any recommendations for existing research on these different classes of decision?

3) Is anonymisation the important differentiator in the underlying data?

(7/n)

And lastly (sorry, yes, still going) if you like scribbling on other people’s google docs, I have tried to spell it out here: docs.google.com/document/d/123… (8/8))