For most people and organizations, work has continued as before, with AI tools becoming a part of everyday routines. So let me take you on a trip through time, back to the ‘20s of the last century, when the first electric washing machine appeared in American homes. It was, of course, a revolutionary device: my grandmother used to tell me about the long hours she spent washing clothes at the communal washhouse or by the river, and she did not miss it after she started using the automatic machine. However, as the researcher and sociologist Diletta Huuskes explains in her book Technologies of the Revolution (Il Saggiatore 2024) , in 1974 Joann Vanek showed that women's position in domestic work had remained virtually unchanged despite the introduction of household technologies.

The historian and technologist Ruth Schwartz Cowan, in her 1983 book More Work for Mother, again well documented this paradox: with the arrival of the washing machine and all the domestic electric helpers, domestic cleanliness standards increased. If shirts were once changed once a week, they now had to be changed every day. If floors were washed on weekends, now they were washed daily. Work did not diminish; it multiplied, adapting itself to the available tool. As Cowan observed, women worked just as much as before, but now they did it alone, because the appliances had also replaced the help of neighbors, babysitters and domestic workers.

This paradox we observed applies, in a disquieting way, to artificial intelligence too. While AI is meant to speed up and take the load off of employees, just like washing machines, it has instead added new layers of work and expectation. This paradox applies, in a disquieting way, to artificial intelligence, too.

So, the questions we need to ask ourselves and our organizations are: what kind of tool, what kind of “intelligence” are we adding to the team and to our processes?

Not Just Bias: This Is a Question of Power

Talking about artificial intelligence (AI) in the third sector almost always means talking about bias, because AI is made by humans. Humans can be biased creatures with discrimination tendencies, leaving the algorithmic systems to reproduce and amplify the inequalities embedded in the data they are trained on, and even take it further. If no regulations or ethical questions are put in place to govern their functioning, they will simply act as a mirror of these practicalities. In her book Automating inequality (Macmillan 2018), Virginia Eubanks presents the use of algorithmic systems in American public services, and she documents what she calls the "digital poorhouse". It’s a set of data infrastructures that, like the physical poorhouses of the nineteenth century, function to "profile, police, and punish the poor" rather than to serve them. One of the examples quoted in the book is about the Los Angeles's Coordinated Entry System, an AI-assisted tool designed to prioritize housing allocation for homeless people. The system collected deeply sensitive data, discriminatory numbers, mental health histories, legal records, from thousands of people experiencing homelessness and shared it across 168 organizations. It ranked individuals on a vulnerability scale to match them to available housing. The issue, as Eubanks shows, was not the algorithm but the question it was asked to answer: how to manage a shortage of housing more efficiently. The shortage itself was never addressed. The tool optimized the distribution of an inadequate resource while building an enormous surveillance infrastructure around the data of the most vulnerable people in the city.

How much data does the third sector collect about the people it works with, and what is it used for? As we are seeing, addressing bias is not enough. Bias is a technical category, so what we need to talk about is power and inequalities.

A useful compass here is the framework of Data Feminism, developed by researchers Catherine D'Ignazio and Lauren F. Klein in their 2020 book of the same name (MIT Press). Built on intersectional feminist thought, the framework is organized around seven principles: Examine Power, Challenge Power, Make Labor Visible, Consider Context, and Embrace Pluralism. The first principle is foundational: before asking what data can do, ask who holds power over it, who benefits, and who is harmed. As D'Ignazio and Klein write, Data Feminism "is about power, about who has it and who doesn't, and about how those differentials of power can be challenged and changed". This is precisely the lens that the third sector needs when approaching AI, not as passive consumers of a technical tool, but as actors with a responsibility to understand where the source of this power is and to recognize it.

We need to train ourselves, before we train and use the systems, to answer those questions: Who builds these systems? Who owns the data they train on? Who decides which values and priorities are embedded in the design? Who profits from their spread? And who bears the consequences when they go wrong?

When we talk about "recognizing bias" or "improving algorithmic transparency," we are using technical language for what is, at its root, a problem of power and historical injustice.

Rachel Adams, a researcher at the Research ICT Africa institute, in her The New Empire of AI (Polity Press 2024), helps us situate AI not as a neutral tool with some unfortunate side effects, but as an extension of historical structures of power: "Social inequality and economic inequality are structural phenomena” she writes”. They are produced by forms of power that exist at given points in time “whereby decisions are made that favor one group while oppressing or marginalizing another."

For civil society organizations, this is an invitation to act: to ask who benefits from the tools we adopt, to demand transparency from the systems we rely on, and to ensure that the communities we serve are not, once again, the ones left outside.

Civil Society and the Ethics of Refusal

In the same article, Rachel Adams writes that "the new empire of AI, is driven by the logic of expansion and of exponential, perpetual growth. It depends on practices of extraction from the very same places that it seeks to expand into and conquer. It is abstract and invisible: clouds in the ether that are hard to see, and even harder to resist." As the scholar argues, the majority of the world is not simply absent from AI's benefits, but it actively subsidizes them, through labor, land and lives.

For civil society organizations whose mission is precisely to serve and advocate for these communities, this is the core of the matter. An organization using AI to profile the beneficiaries of its services, to select volunteers, or to optimize communications could, without knowing it, systematically exclude the most vulnerable people: precisely the ones it exists to help and work with.

So, a truthful, critical AI literacy is not a technical skill, but a civic and ethical one. Training programs for staff and volunteers should cover not just "how to launch a successful prompt" but "how to evaluate an algorithmic output", "how to recognize a bias", and even "when it is right not to use AI".

Your Feedback Matters

What did you think of this text? Take 30 seconds to share your feedback and help us create meaningful content for civil society!

About The Author

Donata Columbro is a journalist, writer , and science communicator, known as a “data humanizer” for her ability to make complex issues related to data, algorithms, and society accessible. She collaborates with various national publications and teaches data visualization and data storytelling at IULM Milan and Università della Svizzera Italiana.

Sources

Diletta Huyskes, Tecnologia della rivoluzione (Il Saggiatore 2024)

Catherine D'Ignazio, Lauren F. Klein, Data Feminism (MIT Press 2020)

Ruth Schwartz Cowan, More work for mother (Plunkett Lake Press, 1983)

Rachel Adams, The New Empire of AI: The Future of Global Inequality (PolityBooks, 2024)

Disclaimers

This piece of resource has been created as part of the AI for Social Change project within TechSoup's Digital Activism Program, with support from Google.org.

AI tools are evolving rapidly, and while we do our best to ensure the validity of the content we provide, sometimes some elements may no longer be up to date. If you notice that a piece of information is outdated, please let us know at content@techsoup.org.