General reading impressions, weeks 7 and 8

Hello all! As mentioned in my previous post, I got married last week. Between my wedding and the honeymoon, I did not read a lot of new stuff (except for advancing on some of my ongoing long reads). So, this set of recommendations will be even sparser than usual.

Papers

Oliver Haimson et al.. 2020 forthcoming. Designing Trans Technology. Accepted to CHI ‘20, April 25–30, 2020, Honolulu, HI, USA.

A study case on the potential and the limits of participatory design approaches for designing technological solutions that address challenges faced by transgender and non-binary people and communities. The paper also presents a solid review of the existing literature on existing technologies built for or by trans and non-binary people, as well as potentially harmful technologies.

Mireille Hildebrandt. 2020. The Artificial Intelligence of European Union Law. German Law Journal 21(SI 1), pp. 74–79.

I am not persuaded by this paper’s arguments regarding the difference between machine intelligence and human intelligence. Proper development of my objections would require more space than the sort of brief recommendation that I intend to give in this series of posts, but its claims regarding machine intelligence somewhat downplay the fact that such systems, even if not intelligent in any meaningful sense of the word, are used in organisational structures which might be said to perform intellectual tasks. However, the author is careful enough to highlight the distinction between the adopted sense of intelligence, which is based on human agents taking an ex-centric position, and agency in general, which does not require intelligence.

The paper still has some very interesting contributions: highlighting the artificial nature of human reasoning, showing how that connects to law as a social artefact, and the general warnings against over-determination, as algorithmic social practices may constrain human agency by fossilising practices in which humans exercise their choices in life (even though it is important to keep in mind that AI is not the only source of over-determination, and law might contribute to that same phenomenon).

Lucas Miotto. 2019. The Good, The Bad, and The Puzzled: Coercion and Compliance.

Miotto argues — mostly against Schauer’s revival of coercion-centred accounts of law — that coercion is not a central factor in the efficacy of legal systems. In the process, he presents an overview of empirical studies regarding the impact of coercion in human compliance with existing norms, and presents a reading of Hart’s puzzled man not as somebody who just follow norms, but as somebody that, after considering all other relevant factors, can still be sufficiently swayed by law. The paper presents a good challenge for those, like me, who hold that an account of law that does not consider coercion is incomplete. Note to self: perhaps it would be instructive to think about coercion along the line adopted by Bainbridge (1983) on the ironies of automation.

John Zerilli et al.. 2019. Algorithm Decision-Making and the Control Problem. Minds & Machines.

As the title suggests, this paper treats of the challenges involved in keeping humans in control of automated decision-making systems, a challenge that applies both to partially automated processes and to the sort of human intervention that was the subject of a paper I presented in the same year. Zerilli et al.’s discussion has many interesting points, such as discussing under which conditions the use of a less-than-optimally reliable system could be acceptable, or the usefulness of task decomposition in obtaining manageable control.

Online reads

Nathan Heller. 2020. Politics Without Politicians. New Yorker.

A nice introduction to Hélène Landemore’s “open democracy” proposal, which relies on transparent structures and randomly-selected representatives. From this layman introduction, I am still reticent about two points: (i) the feasibility of this model in heterogeneous societies, especially in developing countries; (ii) the reasonability of demands this model places upon the selected representatives. Nevertheless, it is an interesting alternative both to traditional electoral representation models and direct democratic deliberation, and I look forward to reading her work.

Marijke Roosen. 2020. What SyRi can teach us about technical solutions for societal challenges. Global Data Justice.

An overview of the SyRi case, in which a Dutch court ruled that the use of an algorithm for benefits fraud detection was incompatible with Article 8 of the European Convention on Human Rights. The court decision emphasised that the safeguards put in place by the Dutch government (or the lack thereof) failed to strike a proper balance between the social interest in combating fraud and the privacy interests of those evaluated by SyRi. As the essay clearly exposes, those violations are particularly dangerous for those who are already in vulnerable positions in society, which highlights the importance of developing systems that benefit, rather than punish, vulnerable populations.

Researcher, Law and Artificial Intelligence

Currently researching the regulation of artificial intelligence at the European University Institute.