Tal Z. Zarsky

The Trouble with Algorithmic Decisions An Analytic Road Map to Examine Efficiency and Fairness in Automated and Opaque Decision Making

We are currently witnessing a sharp rise in the use of algorithmic decision-making tools. In these instances, a new wave of policy concerns is set forth. This article strives to map out these issues, separating the wheat from the chaff. It aims to provide policy makers and scholars with a comprehensive framework for approaching these thorny issues in their various capacities. To achieve this objective, this article focuses its attention on a general analytical framework, which will be applied to a specific subset of the overall discussion. The analytical framework will reduce the discussion to two dimensions, every one of which addressing two central elements. These four factors call for a distinct discussion, which is at times absent in the existing literature. The two dimensions are (1) the specific and novel problems the process assumedly generates and (2) the specific attributes which exacerbate them. While the problems are articulated in a variety of ways, they most likely could be reduced to two broad categories: efficiency and fairness-based concerns. In the context of this discussion, such problems are usually linked to two salient attributes the algorithmic processes feature—its opaque and automated nature.

Discrimination and Privacy in the Information Society

Vast amounts of data are nowadays collected, stored and processed, in an effort to assist in making a variety of administrative and governmental decisions. These innovative steps considerably improve the speed, effectiveness and quality of decisions. Analyses are increasingly performed by data mining and profiling technologies that statistically and automatically determine patterns and trends. However, when such practices lead to unwanted or unjustified selections, they may result in unacceptable forms of discrimination. Processing vast amounts of data may lead to situations in which data controllers know many of the characteristics, behaviors and whereabouts of people. In some cases, analysts might know more about individuals than these individuals know about themselves. Judging people by their digital identities sheds a different light on our views of privacy and data protection. This book discusses discrimination and privacy issues related to data mining and profiling practices. It provides technological and regulatory solutions, to problems which arise in these innovative contexts. The book explains that common measures for mitigating privacy and discrimination, such as access controls and anonymity, fail to properly resolve privacy and discrimination concerns. Therefore, new solutions, focusing on technology design, transparency and accountability are called for and set forth.

Automated prediction: perception, law, and policy

A few predictions about predictions.

Online Privacy, Tailoring, and Persuasion

This chapter tackles a somewhat neglected realm of the information privacy discourse, by directly examining the specific detriments arising from the systematic uses of personal information collected online. The chapter begins by drawing out the flow of personal information in today’s digital environment, while emphasizing the collection, storage, analysis and subsequent uses of such data. The chapter then focuses on a specific use stemming from the information flow — the ability of online content providers to tailor advertisements and marketing materials for every user. The chapter argues that these forms of advertising are more effective than those practiced in other media, and at times might prove to be unfair and manipulative. Therefore, the chapter states that at times regulatory steps must be taken to mitigate these concerns. Finally, the chapter mentions a recent incident in which the tailoring of advertisements on the basis of personal information has caused a somewhat surprising public outcry, and compares these events with the dynamics addressed above.

Responding to the Inevitable Outcomes of Profiling: Recent Lessons from Consumer Financial Markets, and Beyond

Data profiling practices in today’s information age take many shapes and have many faces. They generate both public intrigue and concern, which in many cases make their way to the media and from there to relevant regulators. These latter entities, in turn, struggle in search of a proper regulatory response to the complicated issues set before them. Profiling presents a policy challenge which is indeed frequently invoked and discussed. However, both the harms it presents and the way in which they should be resolved are extremely difficult to conceptualize. The challenge of regulators and scholars grappling with these issues is three-fold: they (a) must generate a helpful taxonomy for understanding and addressing the various practices and their potential problems. They must also (b) establish which issues could be resolved by internal and external market pressures, as well as indirect regulatory pressures, and which require direct regulatory scrutiny and intervention. Finally, and most importantly, they must (c) formulate (or recommend) regulatory responses at the distinct junctures they deem necessary. In this short chapter I attempt to draw out a brief strategic response to these questions, while relying on previous work. My analysis will focus on the first two elements, while providing merely initial intuitions towards overall solutions. In doing so, I will strive to account for the technological, market and legal developments of the most recent years.

Transparency in Data Mining: From Theory to Practice

A broad variety of governmental initiatives are striving to use advanced computerized processes to predict human behavior. This is especially true when the behavioral trends sought generate substantial risks or are difficult to enforce. Data mining applications are the technological tools which make governmental prediction possible. The growing use of predictive practices premised upon the analysis of personal information and powered by data mining, has generated a flurry of negative reactions and responses. A central concern often voiced in this context is the lack of transparency these processes entail. Although echoed across the policy, legal and academic debate, the nature of transparency in this context is unclear and calls for a rigorous analysis. Transparency might pertain to different segments of the data mining and prediction process. This chapter makes initial steps in illuminating the true meaning of transparency in this specific context and provides tools for further examining this issue.

Data Mining as Search: Theoretical Insights and Policy Responses

Data mining has captured the imagination as a tool which could potentially close the intelligence gap constantly deepening between governments and their new targets – terrorists and sophisticated criminals. It should therefore come as no surprise that data mining initiatives are popping up throughout the regulatory framework. The visceral feeling of many in response to the growing use of governmental data mining of personal data is that such practices are extremely problematic. Yet, framing the notions behind the visceral response in the form of legal theory is a difficult task.

'May the Odds Be Ever in Your Favor': Lotteries in Law

Throughout history, lotteries have been used in numerous legal contexts. However, legal theorists have rarely discussed the role of randomization in law, and have never done so systematically and comprehensively. Against this backdrop, the Article has three underlying goals. First, it fills the aforementioned gap by providing a theoretical framework for assessing lotteries’ role in legal resource allocation. It innovatively integrates fairness and efficiency concerns, critically evaluating and applying insights from various disciplines, including economics, philosophy, political science, psychology, and theology. This multidisciplinary framework — of unprecedented breadth and complexity — provides lawyers and policymakers with a powerful analytical tool for assessing the possible use of random allocation schemes. Second, the Article recognizes the importance and highlights the pervasiveness of lotteries in law. It does so by analyzing and appraising the historical and present role of lotteries in numerous legal contexts through the theoretical prism. It also advocates a cautious expansion of the use of lotteries in other contexts, a notion that runs counter to the basic intuition that the law must be committed to reason and certainty. Third, the Article substantiates a jurisprudentially provocative thesis: While random-based schemes can be and are employed in many settings, there is no consistent set of justifications for all applications. The rationalization is highly varied and context-specific.To construct and apply the theoretical framework, the Article uses the fundamental distinction between fairness and efficiency as a cornerstone. Part I unveils the fairness of random selection as a matter of common perceptions and normative commitments. It starts by showing that lotteries are often perceived as fair allocation methods, especially compared to the alternatives (“positive fairness”). Part I then examines whether the use of lotteries can be justified on the ground of fairness (“normative fairness”). It discusses the outmoded theological justification which associates random selection with divine intervention, the egalitarian argument and its limits, the fairness-related advantages and disadvantages of processual detachment from human agency, and fairness vis-a-vis people who do not take part in the primary allocation, be they allocation candidates or allocators.Part II addresses the advantages and possible drawbacks of random selection in terms of efficiency, compared to conventional alternatives: auctions, need- and merit-based allocations, and queues. It first examines recipients’ ability, ex post, to maximize the utility of the allocated resource, as well as ex post psychological effects of the allocation method. This Part then analyzes ex ante changes in potential recipients’ behavior created by random allocations, also noting the outcomes of the so called “insulation” from power-structures facilitated by random processes. Next, Part II examines the relative advantages and shortcomings of random selection in terms of administrative costs. Finally, it discusses possible effects of random allocations on society at large, such as political economy dynamics, and the potential impact on information flow, public knowledge, and taxation policy.

Liability for Online Anonymous Speech: Comparative and Economic Analyses

This is a pre-edited draft of of an article presented in the special session of the Annual Conference on European Tort Law. The article examines various models for handling the problem of online anonymous defamation from comparative and economic perspectives. The comparative analysis reveals four main paradigms. The US model bars content providers’ indirect liability, but facilitates identification of the speaker. The Israeli model recognises content providers’ fault-based liability but does not provide procedural tools for identifying the speaker. The EU framework enables the victim to request identification of the speaker, and at the same time bring an action against the content provider. Although there is variance among Member States, this model seems to comply with the relevant Directives and European court decisions. The recently-adopted English model (‘residual indirect liability’) enables the victim to pursue a claim against the speaker and, if the speaker is unavailable, imposes liability on the content provider. From an economic perspective, the main problem with exclusively direct liability is the special effort in identifying and pursuing the anonymous speaker. Additional, yet probably less serious, problems are the high likelihood of judgment-proof defendants and high transaction costs which prevent a contractual transfer of the burden to the content provider when it is the cheapest cost avoider. The drawbacks of exclusively indirect liability are the relatively high cost of precautions, the fact that content providers do not capture the full social benefit of their activity, and the asymmetric legal response to errors with respect to ‘defamatoriness.’ Concurrent liability of the speaker and the content provider overcomes the high cost of identifying and pursuing anonymous speakers, and the problem of judgment-proof defendants. It also induces content providers to facilitate identification of anonymous speakers, increasing the likelihood of internalisation by primary wrongdoers. But concurrent liability has potentially conflicting effects on deterrence, and may result in an aggregation of the implementation costs of both direct and indirect liability. The residual indirect liability regime eliminates (or at least reduces significantly) the need for monitoring, and prevents over-deterrence associated with unaccounted benefits and asymmetric response to errors. It also incentivises content providers to reduce the cost of identifying anonymous wrongdoers, and does not raise the characteristic problems of multiple-defendants. This model may raise some difficulties but they seem either insignificant or solvable, making the English model (with some modifications) the most efficient.

The Data Mining Balancing Act

Governments face new and serious risks when striving to protect their citizens. Among the various information technology tools discussed in the political and legal sphere, data mining applications for the analysis of personal information have probably generated the greatest interest. Data mining has captured the imagination as a tool which can potentially close the intelligence gap constantly deepening between governments and their targets. In the US, data mining initiatives are popping up everywhere. The reaction to the data mining of personal information by governmental entities came to life in a flurry of reports, discussions, and academic papers. The general notion in these sources is that of fear and even awe. Striving to understand what lies behind this strong visceral response is difficult and complex. An important methodological step must be part of every one of these inquires mentioned above—the adequate consideration of alternatives. This chapter is devoted to bringing this step to the attention of academics and policy makers.