General reading impressions - 2020 Week 6
Hello, all! This week’s reading were somewhat narrow in scope, as I had to finish some writing projects and application before my marriage leave next week.
Joanna J Bryson and Andras Theodorou. 2019. How Society Can Maintain Human-Centric Artificial Intelligence. Human-Centered Digitalization and Services, edited by M. Toivonen and E. Saari, Translational Systems Sciences 19, Springer Nature Singapore.
Shows that centering AI on human needs and interests is not only desirable but technically feasible. The paper integrates technical and governance issues, proposing that control over AI requires both design-based approaches and accountability and transparency mechanisms. As the authors pointly out, a particularly strong challenge to the idea of human-centric AI is that it might be difficult to achieve the substantial level of agreement that is necessary for developing effective mechanisms for controlling AI, a topic that has been in my mind recently. Hopefully, I’ll post some initial thoughts on the subject after my honeymoon.
Paul Ohm. 2018. Forthright Code. Houston Law Review 56(2), 471–504.
An interesting proposal: establishing forthrightness duties for software providers. The persons responsible for the use of software systems would be required not merely to be transparent to the interested reader, but to actively communicate any risks and potential harms. This has the advantage of addressing the difficulties that consumers face when dealing with terms of service and privacy policies, but must be executed while taking into account limited attention spans: since people use many software system everyday, they might be swamped by excessive disclosure.
Langdon Winner. 1980. Do Artifacts Have Politics?
Here, Winner presents the inherently political aspects of technological artifacts: for example, the adequate use of nuclear power plants demands certain social structures — e.g. for caring for long-term residua — that are strongly compatible with, or perhaps even demand, authoritarian political arrangements. This perspective presents a challenge for those (like me) who believe in the possibility of alternative configurations for technologies: perhaps some technologies will always lead to politically undesirable effects, especially as their consequences cannot be confined to specific domains of application. However, there remain the challenge of identifying what technologies — and what aspects of technology — have inherent political meanings and what those meanings are, something that will require attention to the technical objects themselves rather than a wholesale rejection of “modernity”.
Ekker Advocatuur. Landslide Victory in SyRI case: Dutch Court bans risk profiling
A brief collection of materials regarding the SyRI case, in which the Dutch Court of The Hague decided that a risk assessment system used by the Dutch government to detect fraud in public programs was incompatible with Article 8 of the European Convention on Human Rights.
Cade Metz and Adam Satariano. An Algorithm That Grants Freedom, or Takes It Away. New York Times, February 6, 2020.
This long read goes well with the judicial decision presented above, as it shows the impacts of risk analytics on actual lives, and multiple perspectives on the problem, including those more sympathetic to automation.