“The Fallacy of AI Functionality”
I love this paper! I’ve been saying this for years:
Deployed AI systems often do not work. They can be constructed haphazardly, deployed indiscriminately, and promoted deceptively. However, despite this reality, scholars, the press, and policymakers pay too little attention to functionality. This leads to technical and policy solutions focused on “ethical” or value-aligned deployments, often skipping over the prior question of whether a given system functions, or provides any benefits at all. To describe the harms of various types of functionality failures, we analyze a set of case studies to create a taxonomy of known AI functionality issues. We then point to policy and organizational responses that are often overlooked and become more readily available once functionality is drawn into focus. We argue that functionality is a meaningful AI policy challenge, operating as a necessary first step towards protecting affected communities from algorithmic harm.
One mastodon user notes: “My favorite (sarcasm) example of this was police departments buying ML for identifying gunshots. The models were all trained for earthquakes, and the vendor basically repurposed earthquake detection as gunshot detection, made bank, and left departments with a flood of false positives.”(tags: papers false-positives ai ml fail software reliability enshittification)