agg
Data Feminism
A way of thinking about data, both their uses and limits, that is informed by direction, experience, a commitment to action, and intersectional feminist thought. The starting point for data feminism is something that goes mostly unacknowledged in data science: power is not distributed equally in the world (D’Ignazio & Klein, 2020, p. 8).
Data science serves the primary goals of the institutions themselves (corporations, governments, and elite research universities)
Think of these goals as three Ss:
Science (universities)
Surveillance (governments)
Selling (corporations)
Algorithmic Bias
Systematic and repeatable errors in a computer system that create unfair outcomes, such as privileging one arbitrary group of users over others
-Algorithms make mathematical, not ethical, decisions
-What resistance to algorithmic harm looks like:
Remembering the un-inevitability of algorithms and AI
Exposing the black box
Advocacy
Algorithms
Historical information used to make a prediction about the future. What you see in your feeds, what is highlighted, and the ads that are displayed to you, those are powered by AI-enabled algorithms (Kantayya & Hoffman, 2020, 25:18)
Machine Learning Algorithms
A scoring system that can predict the probability of the action you are about to perform, which uses patterns from your data to do so
-“Algorithmic technology is being adopted, and there are no safeguards” (Joy Buolamwini, 2020)
-“We need an FDA for algorithms” (Cathy O’Neil, 2020)
-“People are suffering algorithmic harm, they’re not being told what’s happening to them, there’s no appeal system, and there’s no accountability” (Cathy O’Neil, 2020)
-“Everybody has unconscious biases, and people embed their own biases into technology” (Cathy O’Neil, 2020)
Design After Capitalism
Not a revolution but rather a slow transformation that will require collective effort toward remaking work, culture, and society beyond the competitive, free-market models defined by private ownership of the systems of value production (Wizinsky, 2022, p. 3).
Static Data and Changing Social Views
The paper points out that language models might not capture the dynamic nature of language as used in social movements. For instance, the Black Lives Matter movement's influence on Wikipedia's content reflects evolving social narratives, which static datasets might not capture effectively
Curation, Documentation, and Accountability
The authors propose investing significant resources in curating and documenting training data. They argue for a justice-oriented approach to dataset creation, emphasizing the need to avoid 'documentation debt' where datasets are too large and undocumented, making it difficult to understand and address the biases they contain.
Matrix of Domination in Design
Acknowledges that societal power structures influence design.
- Emphasizes inclusive design considering diverse user needs