Data scientists aspire to solve the world’s toughest problems.
But despite good intentions, many "Data for Social Good" projects are fatally flawed.
Data scientists tend to study and collect data from marginalized groups that bear the brunt of social ills —not the powerful groups who create the conditions for those problems to arise in the first place.
When techno-solutions turn a blind eye to the powerful they are doomed to reinforce the status quo.
Computer scientists have tried to address these challenges by developing complex measuring sticks to evaluate an algorithm’s "fairness." But these equations often miss the point.
Justice requires more than just tweaking the numbers.
It requires us to “study up”—to map, measure, and intervene on the ways that powerful actors maintain social inequity through their decisions and policies.
We must "flip the script" to radically re-imagine the terms of our engagement with data.
But studying up is hard. The owners of data often limit access to their information when we study up, because studying up implicates them in the problems that need to be solved.
We need better strategies for engaging in the political economy of data science.
We need a roadmap for studying up.