Provides an introduction to the harms that occur when facial recognition technology proves to be inconsistently accurate across genders and different skin tones.
Credit: Community-Based Research and Data Justice Guide, p 12. UBC ORICE.
"Data justice is an approach that redresses ways of collecting and disseminating data that have invisibilized and harmed historically marginalized communities. For decades, if not centuries, data has been weaponized against BIPOC communities, in particular, to reinforce oppressive systems that result in divestment and often inappropriate and harmful policies" (Coalition of Communities of Color. Research Justice). |
"Data are never objective and free from bias or ideological convictions, just like historical sources. Data convey conflicts, hegemonies, and colonialisms. They reinforce cognitive biases and subaltern positions, and can sometimes intensify effects of silencing, both in regard to victims and processes of opposition or even defiance. Recognizing the epistemic baggage in data and data structures is thus a key part of carrying out digitally informed source criticism" (Haslinger, P. Data are Never Neutral. Copernico.eu). |
They may seem so - but is this really the case? Data do not spring out of the ether - to exist they must be generated - generally by:
In the case of machines / AI tools these are programmed by a human research team and / or scrape up data derived from human activity. Every decision that goes into the research design is a reflection of the time and place in which the researcher is working - as well as the researcher's particular interests, objectives, world-views etc., and cannot therefore be considered neutral.
Consider this example from Canada:
"The 2021 Census of Population marked the first time that data on gender were collected, allowing for cisgender and transgender men (and) women, and non-binary people to be captured.... Previous versions of the sex at birth statistical standards did not include the distinction of "at birth." Until 2018, gender did not have its own statistical standard. Previous versions of the classification of gender did not include the distinction between men, women and non-binary people, or the distinction between cisgender, transgender and non-binary people until 2021" (Statistics Canada. Understanding sex at birth and gender of people in Canada.)
It's interesting to consider who gets to decide which types of demographic information are valuable, who (if anyone) benefits from data gaps like these, and what socio-political processes might lead to change. The bottom line: important historical information about Canadians who are transgender or non-binary does not exist because it simply wasn't collected - and the data that were collected before these amendments paint an imprecise picture of the population.
To learn about about the gender-identity options now available to respondents see Statistics Canada: Age, Sex at Birth and Gender Reference Guide: Census of Population 2021 [PDF]. |