🧵 time! I’d love to talk about the responsibilities we have as data practitioners. In this ~~information age~~ I think it’s critical we use data, ML, stats, and algorithms fairly, and with an eye toward making the world better for people.
I found this piece on data visualization very striking: medium.com/nightingale/it…
“Knowledge is never subjective.” As the creator of a graph, you hold the narrative power.
Visualizations don’t merely reflect unchangeable truths: “the visual is really important to our sense of what’s ‘normal.’…we should be analyzing things from the minority perspective, the unexpected position.”
There’s constant chatter in our community about algorithms reflecting harmful stereotypes and systems of oppression. Someone inevitably says there’s nothing we can do about this because models just spit back what they know about the world, so it must be true! It’s JuSt fAcTs!
I think these kinds of comments are either a failure of imagination or made in active bad faith.
Stats was basically invented as a field to correct for crappy real-world data! Aren’t we tasked with solving super hard problems? If our models are perpetuating injustice…we should either correct it immediately, or use another tool.
It’s a hard problem, but there’s lots of research on simple ways to at least attempt to create fairer algorithms, e.g.: ai.google/responsibiliti…
And when simple corrections don’t work, or when the harm caused is too great, I believe we should rethink the whole strategy. What problem are we solving, and at what cost?
https://twitter.com/polotek/status/1307781484158230528?s=20
Even with “simple” statistics (like crime rates), I think we have a responsibility to dig deeply and grapple with the systems (often racist) and big picture that gave us that data:
https://twitter.com/geedee215/status/1268930514976636930?s=21
The journey to a better world is a long one and I think we all (myself included) have a lot of work to do. Let’s get to it 💪
• • •
Missing some Tweet in this thread? You can try to
force a refresh
