Five Things to Consider When You Think About Data Bias (DDN3-V05)

Transcript: Five Things to Consider When You Think About Data Bias

[The title "Five Things to Consider When You Think About Data Bias", with a graphic of three people examining a chart appears]

[The title and graphic slides away to reveal Dr. Rachel Zellars on screen, Assistant Professor, Department of Social Justice and Community Studies, St. Mary's University]

How do we know that public servants are using the right data at their disposal? I want to be as practical as possible here. Two of my teachers, Zoe Marks - let me shout them out -  And Teddy Svoronos are really brilliant researchers at Harvard and they taught me to begin with five organizing principles up front, whenever one is gathering or using data.

So I'll just go through those. First, know your population and specifically how is equity observable, what sensitivities in that group exist, and how are you then studying this group in context?

So that's the first. Second, examine your data distribution. This is something I teach very practically every summer in a group to teenagers in my Freedom School program. What does the mean in your data set conceal? So when I teach income inequality to young people, I don't teach the average or the mean income over a period of time. I usually work around, work with 40-year periods, but rather I teach the quintile because it's only through those quintiles where you can see the radical income inequality when it's graphed in North America. Right?

So in Canada and the U.S. So and it really resonates, of course, with what we intuitively feel. The rich are getting richer, the poor are getting poorer. So really distinguishing between how you chart, or how are you paying attention to the mean? And then what is that concealing? In the case of income inequality a lot.

Thirdly, intersectional approach. Always take an intersectional approach. Inequality interacts with multiple identity categories, and one of our biased tendencies as researchers and people in general is to flatten or simplify marginalized groups of people while affording great detail nuance to whiteness, to the status quo.

Black women, for example, live with a multiplicity of identities. So we need disaggregated data in public service on Black women, Black disabled women, Black French-speaking women, Black queer women and yes, Black disabled French-speaking queer women in public service. We need those data sets. And this is where I'll just make a plug for qualitative data as well. Very useful.

Fourth. What's the data gap? Where is the data missing? This one can make me cry. So let me take a deep breath first. I've spent the last few months scouring through past Royal Commission studies in public service. In the early 1980s, one of the first collections of Black feminist writings in the United States was contained in a book called All the Women Are White and All the Blacks Are Men, But Some of Us Are Brave. Historically, the most marginalized have been studied the least, or simply not studied at all. And so, in public service, we have enormous and greatly disaggregated sets on those valued as universal. Thus, the title all the women are White and all the men are Black. So, when attending to the data gaps, we have to be super mindful of a few things. First, what biases are we bringing to the subject matter at the start? I'll talk about bias in a bit. Secondly, what is the availability and the quality with a capital Q of the data that does currently exist? The quality.

Thirdly, how do we attend to the gaps, the absences, the silences that exist in data sets?
Primary sources and archives. This is such a big question, and I'm just going to shout out, you know, the scholars that have taught me the most about how to read those silences, absences and gaps. Michel-Rolph Trouillot, Saidiya Hartman's work, Thavolia Glymph, and our own brilliant historian here in Canada, Charmaine Nelson. And it's Thavolia Glymph who taught me years ago better than anyone that what we ignore as researchers, what we ignore, reveals much more than what we actually give our attention to.

And then the final thing that I'll say about this. Two more points. Fourthly, in terms of data reliability collection, be accountable to the communities that you're studying.

So important. And then the final thing I'll say is, defining equity and success carefully with with that data.

I love this quotation so much: "Equality in does not mean equity out". That's from Zoe Marks. How are we defining success? Are we using the status quo or whiteness to be blunt as the standard and as researchers gathering and figuring out how to use data, are we paying close attention to both the disproportions and the significance and meaning of those disproportions? So those are five sort of organizing principles that I hope are useful. Thank you.

[The five organizing principles appear at the same time on screen:

1. Know your population.
2. Examine your data distribution.
3. Intersectional approach.
4. What is the data gap?
5. Define equity and success carefully.]