Karen Yeung and Martin Lodge (Eds), (2019), Oxford University Press, 271 pages
The book Algorithmic Regulation is a collection of 11 very accessible essays on the role of algorithms and big data on regulatory governance. The book is edited by Professor Karen Yeung (Birmingham Law School), one of the leading scholars in regulation and technology, and Professor Martin Lodge (London School of Economics).
Yeung and Lodge define algorithmic regulation as: “decision-making systems that regulate a domain of activity in order to manage risk or alter behaviour through continual computational generation of knowledge from data emitted and directly collected (in real time on a continuous basis) from numerous dynamic components pertaining to the regulated environment in order to identify and, if necessary, automatically refine (or prompt refinement of) the system’s operations to attain a pre-specified goal” (p.5).
The strength of the book, in my reading, is that it moves well beyond both overly hopeful future visions of the opportunities of algorithmic regulation and extremely negative doom and gloom scenarios. Quite practically, the various authors of the book look at algorithmic regulation as a means for “a better, more efficient and accurate, identification of sites that require regulatory attention” (p. 178). In other words, a way of advancing risk-based regulation or improving the allocation of scarce regulatory resources.
Particularly the chapters by Yeung and Lodge are worth a very close read because they take a step back and try to provide as objective as possible answers to questions such as: What are the ethical and normative challenges of algorithmic regulation? How can we best regulate algorithmic regulation (and regulators using it)?
Yeung foresees three significant ethical and normative challenges of algorithmic regulation that regulators need to tackle. First are concerns associated with using algorithmic regulation for decision-making processes. These “may interfere with the fundamental right of individuals to be treated with dignity and respect, without necessarily generating material harm or damage” (p.43). Second are concerns associated with the outcomes of algorithmic regulation. Such outcomes may be erroneous, inaccurate, and discriminatory (because flawed data is used or a poor algorithm is used). Algorithmic regulation may also result in adverse consequences and reinforce existing patterns of injustice against minorities. Third are concerns about how algorithmic regulation will “predict and personalize services offered to individuals” (p.22), mainly when such predictions and personalization build on data that individuals have not shared for these ends (e.g., through their Facebook or Twitter accounts).
Building on these challenges, Lodge (and his essay co-author Dr Andrea Mennicken) provide a set of recommendations for the use of algorithmic regulation. Most importantly, they argue, “is to establish understandings … of bias, perceived fairness of decision-making, explainability of machine learning algorithms, both in terms of transparency and assumed causality, reliability of the decision-making based on such algorithms, and continued external ‘debugging’ so as to ensure the algorithms remain ‘in control’ and uncorrupted” (p.195). To make sure the algorithmic regulation is used fairly and transparently, they call on regulators to pay renewed attention to the ‘typical’ questions of regulating the process of regulation itself. For New Zealand, that could mean including ‘how to deal with algorithmic regulation’ in our broader regulatory stewardship thinking.
The book is relevant to everyone interested in the use of big data, machine learning, and algorithms in regulatory governance. May it be because they worry about these developments, are highly hopeful about their potential, or just want to be objectively informed about the pros and cons that come with them.
Disclaimer In these brief book reviews, I discuss classic and contemporary books that make up the canon of regulatory scholarship. I focus on their central guiding idea or core notions and aim to keep the reviews to around 500 words. Unfortunately, this implies I must sacrifice a considerable amount of detail from the books reviewed.