It strikes me that the latest @NatureComms snafoo (this one about female mentorship) has something in common with the last few - and it’s not just that the analyses provide racist and sexist conclusions. /1

It is that they 1) use data produced for other things, rather than created for the study; 2) assume data is neutral, rather than being shaped by its production process; 3) assume that scale in some ways compensates for bias. /2

Historians use data made in other contexts all the time. It’s what we do. But key to historical research is that data must take account of how it is produced as part of the lens of analysis. And usually this means asking how data is shaped by gender, race, class etc norms /3

Importantly just because you have a LOT of data doesn’t mean that it is somehow more natural or available for analysis outside the lens of its production. It just means you have a lot of data. /4

The problem with the high profile cases coming out of @NatureComms is that they use data produced in one context for another without attending to production. And then are surprised the conclusions are flawed. Applying lessons from historical analysis might help solve this /end