Exploring the Puzzles of Decision-Making Algorithms in Public Administration: an Annotated Bibliography
This work is an experimentation around the topic of decision-making algorithms in public administration. I initiated it as I was an intern at Etalab, the French government task force for open data and open government.
My goal is to build a collaborative bibliography gather social science literature around issues of algorithms, artificial intelligence and governance. For a summary of the works I've gathered so far, see the complete bibliography page.
The purpose of this work was threefold:
- Take some critical distance with “State thinking” (sociologist Pierre Bourdieu’s “pensée d’État”) (Bourdieu, 1993). As a student trained in the social sciences, I wanted to take a step back from the beliefs and discourses circulating around algorithms. Perhaps not surprisingly, it was very difficult to achieve while literally working for the State agency in charge of data science in public service, especially as part of my job was to promote the program I was coordinating politically;
- Acquire a broad overview and a better understanding of issues of algorithmic governance;
- Be a useful tool for both my colleagues involved with policy making and my fellow students interested in the digital, politics, society and education. The purpose is to try and understand if and how we can ensure that algorithmic programs are used in a manner that is consistent with what we want from public action.
For now, I organized the work in 5 parts, based around 5 broad questions:
- What do we mean by “algorithms”?
- Why and when are they used in public administration?
- What does the use of decision-making algorithms in public administration reveal about politics? What does “algorithmic governmentality” entail?
- What are the ethical and epistemological concerns associated with algorithmic governmentality? What changes do they bring to public administration?
- How to make public decision-making algorithms more accountable?
I turned each question into a section of this 5-part bibliography. For each section, I gathered literature I found relevant, and selected a few articles I wanted to annotate in-depth. For the non-commented articles, I included a short abstract when available. The annotations are made up of quick summary of the work, evaluation of its overall interest and of its contribution to a work on decision-making algorithms in public administration.
In short, my answers to the aforementioned questions are:
- It is most useful to consider algorithms as sociotechnical assemblages of human and non-human elements;
- We must debunk the myth of neutral, objective and efficient algorithms;
- The social power of algorithms offers an interesting lens through which to observe politics;
- The challenges associated with algorithmic governmentality have to be counterbalanced by a study of algorithms in practice;
- Ensuring the accountability of algorithms entails going beyond transparency.
The bibliography is also available on Zotero.
This is only a V0 of the work. The bibliography is far from complete, the questions it raises are numerous and diverse, the answers I give are only opinions and widely debatable. However, I hope that this work can serve as the basis for discussions and debates. Feel free to contribute by making pull-requests, opening issues, or even forking the project and making it your own! I will also try to create more room for other fields (legal issues and GDPR, anyone?).