Richard Warner (Chicago-Kent College of Law) and Robert H. Sloan (University of Illinois) have posted “How AI Unfairly Tilts the Playing Field: Privacy, Fairness, and Risk Shadows” on SSRN. Here is the abstract:

Private sector applications of artificial intelligence (AI) raise related questions of informational privacy and fairness. Fairness requires that market competition occurs on a level playing field, and uses of AI unfairly tilt the field. Informational privacy concerns arise because AI tilts the playing field by taking information about activities in one area of one’s life and using it in ways that impose novel risks in areas not formerly associated with such risks. The loss of control over that information constitutes a loss of informational privacy. To illustrate both the fairness and privacy issues, imagine, for example, that Sally declares bankruptcy after defaulting on $50,000 of credit card debt. She incurred the debt by paying for lifesaving medical treatment for her eight-year-old daughter. Post-bankruptcy Sally is a good credit risk. Her daughter has recovered, and her sole-proprietor business is seeing increased sales. Given her bankruptcy, however, an AI credit scoring system predicts that she is a poor risk and assigns her a low score. That low credit score casts a shadow that falls on her when her auto insurance company, which uses credit scores in its AI system as a measure of the propensity to take risks, raises her premium. Is it fair that saving her daughter’s life should carry with it the risk—realized in this case—of a higher premium? The pattern is not confined to credit ratings and insurance premiums. AI routinely creates risk shadows.

We address fairness questions in two steps. First, we turn to philosophical theories of fairness as equality of opportunity to spell out the content behind our metaphor of tilting the playing field. Second, we address the question of how, when confronted with a mathematically complex AI system, one can tell whether the system meets requirements of fairness. We answer by formulating three conditions whose violation makes a system presumptively unfair. The conditions provide a lens that reveals relevant features when policy makers and regulators investigate complex systems. Our goal is not to resolve fairness issues but to contribute to the creation of a forum in which legal regulators and affected parties can work to resolve them. The third of our three condition requires that systems incorporate contextual information about individual consumers, and we conclude by raising the question of whether our suggested approach to fairness significantly reduces informational privacy. We do not answer the question but emphasize that fairness and informational privacy questions can closely intertwine.

Hofit Wasserman-Rozen (Tel-Aviv U), Ran Gilad-Bachrach (same), and Niva Elkin-Koren (same) have posted “Lost in Translation: The Limits of Explainability in AI” on SSRN. Here is the abstract:

As artificial intelligence becomes more prevalent, regulators are increasingly turning to legal measures, like “a right to explanation” to protect against potential risks raised by AI systems. However, are eXplainable AI (XAI) tools – the artificial intelligence tools that provide such explanations – up for the task?

This paper critically examines XAI’s potential to facilitate the right to explanation by applying the prism of explanation’s role in law to different stakeholders. Inspecting the underlying functions of reason-giving reveals different objectives for each of the stakeholders involved. From the perspective of a decision-subject, reason-giving facilitates due process and acknowledges human agency. From a decision-maker’s perspective, reason-giving contributes to improving the quality of the decisions themselves. From an ecosystem perspective, reason-giving may strengthen the authority of the decision-making system toward different stakeholders by promoting accountability and legitimacy, and by providing better guidance. Applying this analytical framework to XAI’s generated explanations reveals that XAI fails to fulfill the underlying objectives of the right to explanation from the perspective of both the decision-subject and the decision-maker. In contrast, XAI is found to be extremely well-suited to fulfil the underlying functions of reason-giving from an ecosystems’ perspective, namely, strengthening the authority of the decision-making system. However, lacking all other virtues, this isolated ability may be misused or abused, eventually harming XAI’s intended human audience. The disparity between human decision-making and automated decisions makes XAI an insufficient and even a risky tool, rather than serving as a guardian of human rights. After conducting a rigorous analysis of these ramifications, this paper concludes by urging regulators and the XAI community to reconsider the pursuit of explainability and the right to explanation of AI systems.

Bill Tomlinson (UC Irvine), Donald Patterson (same), and Andrew W. Torrance (Kansas Law; MIT Sloan) have posted “Turning Fake Data into Fake News: The A.I. Training Set as a Trojan Horse of Misinformation” (San Diego Law Journal, Forthcoming) on SSRN. Here is the abstract:

Generative artificial intelligence (“A.I.”) offers tremendous benefits to society. However, these benefits must be carefully weighed against the societal damage A.I. can also cause. Dangers posed by inaccurate training sets have been raised by many authors. These include racial discrimination, sexual bias, and other pernicious forms of misinformation. One remedy to such problems is to ensure that training sets used to teach A.I. models are correct and that the data upon which they rely are accurate. An assumption behind this correction is that data inaccuracies are inadvertent mistakes. However, a darker possibility exists: the deliberate seeding of training sets with inaccurate information for the purpose of skewing the output of A.I. models toward misinformation. As United States Supreme Court Justice Oliver Wendell Holmes, Jr., suggested, laws are not written for the “good man”, because good people will tend to obey moral and legal principles in manners consistent with a well functioning society even in the absence of formal laws. Rather, Holmes proposed, laws should be written with the “bad man” in mind, because bad people will push the limits of acceptable behavior, engaging in cheating, dishonesty, crime, and other societally-damaging practices, unless constrained by carefully-designed laws and their accompanying penalties.

This article raises the spectre of the deliberate sabotage of training sets used to train A.I. models, with the purpose of perverting the outputs of such models. Examples include fostering revisionist histories, unjustly harming or rehabilitating the reputations of people, companies, or institutions, or even promoting as true ideas that are not. Strategic and clever efforts to introduce ideas into training sets that later manifest themselves as facts could aid and abet fraud, libel, slander, or the creation of “truths,” the belief in which promote the interests of particular individuals or groups. Imagine, for example, a first investor who buys grapefruit futures, who then seeds training sets with the idea that grapefruits will become the new gold, with the result that later prospective investors who consult A.I. models for investment advice are informed that they should invest in grapefruit, enriching the first investor. Or, consider a malevolent political movement that hopes to rehabilitate the reputation of an abhorrent leader; if done effectively, this movement could seed training sets with sympathetic information about this leader, resulting in positive portrayals of this leader in the future outputs of trained A.I. models.

This article adopts the cautious attitude necessitated by Holmes’ Bad Man, applying it to proactively stopping, or retroactively punishing and correcting, deliberate attempts to subvert the training sets of A.I. models. It offers legal approaches drawn from doctrines ranging from fraud, nuisance, libel, and slander, to misappropriation, privacy, and right of publicity. It balances these with protections for speech afforded by the First Amendment and other doctrines of free speech. The result is the first comprehensive attempt to prevent, respond to, and correct deliberate attempts to subvert training sets of A.I. models for malicious purposes.