Brief updates (AI, Law, and Otter Things #37)
In this issue, I share some recently published works of mine and a few links to Twitter threads that might interest the reader. And, as usual, dog and otter pictures.
Hi everyone! Once again, I ended up skipping the regular schedule for this newsletter. The blame resides primarily on three factors. The first of them is Guybrush Threepwood, mighty pirate:
It is not that I have been playing Return to Monkey Island non-stop (unlike my previous experiences with Crusader Kings III and my yearly break around the time Football Manager releases a new edition). But the last few weeks have been quite intensive, and I ended up dedicating any spare time to relax and not to this newsletter.
The second factor, as one might expect, is the lovely creature below:
After our summer vacation, she is bolder than she used to be. She now goes outside, plays with toys, and does various things she did not do a month ago. However, this also means I have to get used to the new, daring Winnie and how she is very different from the scared little girl that arrived in our apartment a few months ago.
Finally, the third factor—which, let's be honest, is the relevant one—is overcommitment to projects. After some struggles, documented in previous issues, I am getting much better at saying no to projects, even if they sound interesting. It is not easy, to move from a scarcity of exciting work to a situation where you have more cool things to do than hours in the day. But I am slowly improving my prioritization skills.
The timing for this shift could not be better, of course, as I start thinking about what I might want to do after the PhD. Caring for a dog and some other shifts in my personal life also led to a recalculation of priorities. Nowadays, I substantially reduced the number of projects and activities I am involved in, cutting them to a more manageable level. While the feeling of being swamped has not fully subsided, it makes a world of difference when you have time to pursue other interests. Even if I delay this newsletter a week or two to make room for those things.
Self-plug
Over the last few issues, I have been talking quite a bit about some ongoing projects. Now, a few of them have been published, and so I would like to share them:
Abstract Tax administrations globally increasingly rely on artificial intelligence (AI) systems for automation. However, automation has a huge potential impact on the rights of taxpayers subject to algorithmic assessments, which is compounded by the opacity of complex AI systems. This article argues that adequate protection of taxpayers’ rights demands the use of eXplainable AI (XAI) technologies that can render the functioning and decisions of tax AI systems understandable for taxpayers, administrative appeal bodies and the courts. This demand follows from the constitutional principles that guide taxation. Still, it is insufficiently addressed by soft and hard law instruments on AI, which do not address the particular information needs of the tax domain. To address this gap, the authors conclude the article by mapping technical and legal challenges for the proper application of explanation techniques to tax AI to ensure that automation does not come at the expense of taxpayers’ rights.
Opportunities and limits of European social network regulation
The reuse of metadata from Cadmus, the EUI Research Repository is authorized under the CC0 1.0 Universal (CC0 1.0) Public Domain Dedication
Abstract This chapter argues that social networks are currently undergoing a turn towards adopting procedural safeguards and duties of care regarding the substantive rights of users. The current EU regulatory framework, centred on the eCommerce Directive, was thought for a different online environment. Therefore, it is strained by social networks in ways legislators and courts are currently trying to address. Some of these strains are produced by the institutional design of the regulatory framework, but these institutional factors only become a problem in light of the harms that social networks introduce or amplify. Despite the fact that harmful user behaviour may sometimes be advantageous to social networks (e.g., by attracting certain groups of users), social networks may be induced to adopt content moderation approaches not only in the interest of the users that could be harmed or repelled by such behaviour, but also to avoid losing the liability exemption they enjoy as intermediary carriers of user-generated content. Content moderation may itself introduce risks to users’ rights, and EU courts and legislators have sought to constrain the range of discretion available to moderators. In this context, we argue the regulation of social networks should be perceived as a socio-technical problem, in which neither technical approaches nor general law alone are conducive to socially desirable outcomes. Instead, regulation needs to be aware of the social impacts of platforms, and the role technology can play in amplifying or mitigating them.
In addition, online registration for the REALaw Forum for Young Scholars is now open. Though I am neither a European lawyer nor an administrative lawyer (and, let's face it, not young, either), I will have the opportunity to present some work in progress there. The paper I will present there analyses how the introduction of AI systems into administrative decision-making processes may bring uncertainties for the decision-makers themselves, and not just for the subjects of partially or fully automated decisions.
An Abstract for my paper: Administrative bodies in the European Union and its Member States increasingly rely on artificial intelligence (AI) technologies. These systems are sold, at least partly, as tools to help governments navigate the uncertainties of modern information societies. This paper argues, however, that administrative automation does not necessarily lead to increased certainty in decision-making. Often, the adoption of an AI system is grounded on promises of certainty that go beyond what current technologies can offer. But even when an AI system does not overpromise, its technical complexity can be an obstacle for administrative decision-makers attempting to make sense of the relevant facts and the law. After mapping these two sources of uncertainty, the paper discusses their impact on the legal principles that guide administrative decision-making and proposes how current provisions of EU law may be used to mitigate the uncertainty derived from the use of AI systems in administrative decisions.
Some potentially interesting Twitter threads
One of the best things about Twitter, in my opinion, is that it is an excellent interactive space. Unlike LinkedIn—where we mostly go for performative bragging, which I do not like as much as this newsletter might suggest—Twitter offers something akin to a mensa or coffee machine chat. You can spend some time sharing otter gifs, tweeting about pop culture, and complaining about minor things in life...while interacting with the same interlocutors that can give you deep insights about things you are working on.
(I also love to use Twitter to follow people working in areas in which I have little knowledge, or even interest, but your mileage may vary).
So, from time to time, I come across some interesting reading recommendations or analyses. I try and share them in this newsletter whenever I can, but I haven't done so in a while. So, here are some threads that might be of interest (don't forget to check the replies!)
And now, the otters
See you next time!