New Public Law in Response to New Public Analytics

1 Introduction

At the end of her rich and thought-provoking Montesquieu lecture, Karen Yeung expresses the hope that the analytical construct of ‘New Public Analytics’ (NPA) that she has introduced, can serve as a helpful vehicle for investigation, reflection and critical comparison across the rich variety of policy domains, sites and locations in which NPA is being taken up.¹ This contribution takes up this invitation to reflect upon the consequences of ‘New Public Analytics’ by considering this new phenomenon through the lens of administrative law. Where administrative law has always been concerned with keeping tabs on the public administration, NPA may pose new challenges for administrative law in realizing its mission and even provoke a paradigm shift in administrative law.

After discussing some highlights of Karen Yeung’s introduction of NPA, I will sketch briefly how administrative law has responded to New Public Management (NPM), the predecessor of NPA. Next, I will consider to what extent the ‘new’ challenges of NPA for administrative law really differ from those posed already by NPM. Finally, I will explore which paradigms of administrative law deserve specific attention in order to establish ‘new administrative law’ that is able to meet the challenges of NPA.

2 Characterizing New Public Analytics

When introducing the concept of New Public Analytics (NPA), Yeung has three different objectives in mind. First, she uses NPA in a descriptive sense as an overarching concept to denote the increasing take-up of digital automation and data-driven approaches to public sector decision-making and service delivery across a number of OECD countries from around 2010 onwards. Secondly, she introduces NPA as an analytical construct, comprised of an amalgam of related ideas associated with specific properties, beliefs and practices associated with networked digital technologies, such as governance through algorithms and reliance on datafication. Thirdly, she presents NPA as an emergent yet paradigmatic shift that may be understood as conceptual and ideological successor to its most well-known predecessor of ‘New Public Management’ (NPM).

In explaining this paradigmatic shift, Yeung stresses that NPA appears to be devoid of grand substantive ambitions. Unlike NPM, which explicitly championed the value of efficiency, NPA focuses exclusively on process improvement while claiming to be politically agnostic. This apparently ‘neutral’ orientation impacts the practice of statecraft. Where NPA focuses on process optimization, most attention of public authorities will invariably be directed towards enhancing the benefits for the deploying public sector organization. When technologies that rely on data science methods are introduced and implemented within the public sector, there may be a tendency to frame the relevant problem narrowly in terms of ‘task performance’ rather than a set of broader substantive policy goals that animate the relevant application domain. In other words, the underlying public interests disappear into the background. Apart from its negative impact on statecraft, Yeung also describes changing citizen-state relations under NPA. One issue here is that NPA does not see the citizen-subject as thinking, feeling, embodied individuals with lives of their own who are entitled to be treated as moral agents with dignity and respect. Instead, the datafied citizen-subject is comprised of a set of datapoints, often ‘interpreted’ through the lens of algorithmically generated predictions created primarily to optimize public sector operations. Another issue here is that NPA seems to assume that individualized provision of public services is necessarily superior to universal provision, which is a contestable assumption in her eyes. Finally, NPA goes hand in hand with outsourcing to the private sector, thereby redistributing public and private power. There is a deep concern that this outsourcing erodes public law protections because private providers can evade many of the basic obligations of due process, transparency, accountability, equality, rationality and the like, that would apply if the service was provided directly by government. Even though these public law norms could be imposed in theory as ‘quality’ considerations on the provider via contract to address the constitutional and democratic accountability deficits that may otherwise arise, it turns out to be rather difficult in practice to translate these norms into enforceable contract terms, thereby enabling private service providers to promote their self-interest at the expense of the public interest. The question, then, is how administrative law should respond to this paradigm shift provoked by NPA.

3 Continuity and discontinuity for administrative law

One decade ago, the well-known legal scholar and judge Sabino Cassese published his manifesto ‘New paths for administrative law’.² In this manifesto, Cassese observes two opposite trends in administrative law literature: while one body of literature is emphasizing “the end of administrative law”, another body depicts the rise of “new administrative law”. This new (or postmodern) administrative law is focused on ‘steering’ rather than on ‘ordering’, which used to be the traditional orientation of administrative law, and takes into account the new role of the state as a promoter, as a facilitator, as a risk regulator, and as the helmsman of economy and society.³

According to Cassese, discontinuity in the realm of administrative institutions requires discontinuity in the approaches adopted for studying the new administrative law.⁴ Undoubtedly, there has been discontinuity in the realm of administrative institutions with the advent of New Public Management. As Cassese puts is:

Cassese is certainly not the first administrative law scholar considering the consequences of NPM for administrative law. In 1997, Carol Harlow discussed already the impact of new public management on English administrative law, claiming that a “New Public Law” had emerged in parallel with the New Public Management.⁶ Interestingly, one fundamental characteristic intrinsically linked to NPM at that time, was the increased reliance on information technology, next to (i) a slowing down or reversal of government growth, (ii) privatisation and quasi-privatisation, and (iii) internationalization.⁷ However, in the manifesto of Cassese one decade ago (in 2012), any reference to the impact of information technology or related developments such as digitalization, datafication or automation is still lacking. Apparently, the relative weight of the use of information technology vis-à-vis the other major characteristics of NPM (government decline, privatization, internationalization) has been limited until that moment. Thus, Cassese discusses trends related to these other characteristics of NPM, such as the divide between public law and private law (in line with privatization), the changing role of the executive branch of government and the dynamic interplay between national legal orders.

So from the perspective of administrative law, what is new about NPA in comparison with NPM? To some extent, the challenges posed by NPA have already been considered in response to NPM. This holds in particular for the third implication of NPA that Yeung is discussing: private sector outsourcing and the distribution of public and private power. Since this process of privatization was already present in NPM (e.g. by outsourcing the prison system), it is certainly not unique for NPA. Thus, from the perspective of administrative law, privatization and outsourcing represent continuity rather than discontinuity or a paradigmatic shift. Therefore, other elements of NPA should account for the discontinuity of NPA in comparison with its predecessor of NPM and therefore give rise to a paradigm shift.

Indeed, the other two implications of NPA that Yeung describes (the practice of statecraft and changing citizen-state relations) seem more distinctive for NPA. Even though both the practice of statecraft and citizen-state relations have also been subject to change under NPM with its focus on efficient public service delivery, Yeung argues that unlike NPM, NPA is largely devoid of substantive content by focusing on ‘process optimization’, thereby losing sight of underlying public interests and the citizens involved. Interestingly, where Cassese described how the Weberian “administrative machine” has been replaced under NPM by a more complex triangular relationship between administration, politics and society, with the administration acting as steering facilitator,⁸ the “machine is back” under NPA, although in a different appearance.

4 A paradigm shift for administrative law?

As said, discontinuity in the realm of administrative institutions requires discontinuity in the approaches adopted for studying administrative law. Where Yeung shows that there is again discontinuity in the realm of administrative institutions, this time provoked by NPA, the question is justified whether administrative law itself should also undergo some form of paradigmatic shift in response to New Public Analytics.

In the recent literature on the interplay between ‘algorithmic governance’ or automated decision-making (ADM) and administrative law, a common belief in the ability of administrative law to deal with the new challenges of ADM is visible. Marion Oswald, for example, argues that the principles of administrative law, which are concerned with human decisions involved in the exercise of state power and discretion, offer a promising avenue for the regulation of the growing number of algorithm-assisted decisions within the public sector. According to her, administrative law is flexible enough to respond to many of the challenges raised by the use of algorithms, in particular predictive machine learning algorithms, and can signpost key principles for the deployment of algorithms within public sector settings. To support this claim, she points out that these ‘tech-agnostic’⁹ principles, although derived from historic case law, have already been applied and refined to modern government, to the development of the welfare state, privatization, the development of executive agencies and so on.¹⁰ Other scholars in the field of administrative law show similar optimism. Marion Oswald argues that the more traditional area of administrative law can provide an adequate basis for exercising control over the use of new technologies. According to her, the current administrative law should be understood as a basis for moving forward, rather than as a comprehensive framework which satisfactorily governs public sector ADM.¹¹ Rebecca Williams also acknowledges the potential of administrative law to provide the necessary guidance to enable optimal use of algorithmic systems, provided that it becomes more nuanced and advanced. If administrative law is able to transform, it has the potential not only to do useful work itself in controlling ADM, but also to provide guidance on the interpretation of concepts such as ‘meaningful information’ and ‘proportionality’ within data protection law.¹² Cary Coglianese and David Lehr, finally, argue that nothing about algorithms makes them uniquely or automatically unsuitable for use by administrative agencies in terms of their ability to comport with core legal principles of administrative law. Instead, they observe that these algorithmic systems (with a specific focus on machine learning algorithms) can fit quite comfortably within existing cornerstones of administrative law, such as the principles of nondelegation, due process, antidiscrimination, and transparency.¹³

All in all, more than proposing a paradigm shift, the prevalent opinion among administrative law scholars seems to be one of contextualizing administrative law to tackle the challenges of ADM (or NPA) adequately. In his recent essay ‘Administrative Law in the Automated State’, Cary Coglianese goes even one step further by arguing that a shift to automated administration could even be said to represent something of an ‘apotheosis’ of the principles behind administrative law: automation can keep the exercise of administrative authority democratically accountable, while ensuring that such authority is based on sound expert judgment.¹⁴ He therefore concludes that it is far from clear that the fundamentals of administrative law will change dramatically in an era of algorithmic governance.¹⁵

This line or argumentation, stressing the continuity of administrative law in tackling the challenges posed by NPA, however, is one side of the medal. Even where all the authors discussed above express their confidence in the ability of the principles of administrative law to regulate NPA adequately, each of them emphasizes the need for contextualizing these principles in an algorithmic context. As Coglianese argues:

Similar observations are made by other scholars in administrative law. However, next to the option of contextualizing these ‘tech-agnostic’ principles of administrative law, which is still an example of continuity, there is also the possibility of acknowledging that existing principles of administrative law, even when contextualized, will fall short in regulating NPA. Cobbe, for example, does not exclude that administrative law may need to develop new principles and standards for ADM in addition to existing legal standards.¹⁷ Williams identifies a leading role for administrative law, provided that it becomes ‘more nuanced and advanced’.¹⁸ What then could be the elements of discontinuity in the paradigms of administrative law?

In the literature, several points of attention have been identified, of which only a few will be mentioned briefly here. One of them is the degree of deference by administrative courts when reviewing decisions of administrative authorities. Williams argues that the use of (certain) algorithmic systems could result in a reduction of deference, as what is currently decided broadly, implicitly and even unconsciously by humans within a general sphere of discretion becomes instead a series of detailed, explicit technical decisions which are, by the very fact of their transparency, more open to judicial review. According to Williams, this is even an inevitable part of the process of automation and a necessary price to pay.¹⁹

Another point of attention is the blurring distinction between the general rule (adopted by the legislature) and the individual decision (taken by the executive). According to Cobbe, the fact that individual automated decisions are heavily influenced by the processes and choices around the algorithmic system means that in order to properly evaluate those individual decisions in a ‘bureaucratic’ review it may be necessary to also evaluate some of those broader processes and choices. Thus, the distinction between review of (general) policy and review of individual decisions which exists for human decision-making may be significantly blurred or eroded for ADM.²⁰

Coglianese, finally, identifies as an important challenge for administrative law to find ways to encourage that an automated state is also an empathic state. According to Coglianese, it is this (lack of) empathy where existing principles of administrative law will fall short in an automated state and where the need for greater vision will be needed, as current due process guarantees are insufficient to realize this empathy in a digitalized context.²¹ This element of empathy brings us back to Karen Yeung’s powerful analysis of New Public Analytics: where NPA reduces statecraft to operational decisions aimed at fostering process optimalization, loosing sight of underlying public interests, and where NPA does not see the citizen-subject as thinking, feeling, embodied individuals with lives of their own who are entitled to be treated as moral agents with dignity and respect,²² administrative law needs to secure human empathy by pushing governments forward towards additional opportunities for listening and compassionate responses.²³

5 Goodbye Montesquieu?

In his essay ‘Good-bye Montesquieu’, Bruce Ackerman claims that the ‘holy trinity’ of Montesquieu with its separation of the legislative, executive and judicial branches of government, should be abandoned or at least rethought. According to Ackerman, Montesquieu’s mantra should take account of an institutional world in which independent institutions play increasingly important functions, even though they cannot be classified as legislative or judicial or executive.²⁴ While he does not discuss a revision of the ‘holy trinity’ in the context of digitalization, the paradigmatic shift of New Public Analytics also gives rise to rethink Montesquieu. For example, the blurring distinction under NPA between general rule-making and individual decision-making, especially where machine learning algorithms create general patterns and decision rules on the basis of individual cases, does also challenge Montesquieu’s separation of powers.

Although the challenges posed to administrative law by NPA are to some extent in line with the challenges already posed by NPM, the characteristics of NPA also pose additional challenges for administrative law that have not been tackled before. These challenges can be addressed partly by relying on the flexibility of general principles of administrative law, but this approach might fall short. In those circumstances, general principles of administrative law should not only be rethought, but also rebalanced, possibly resulting in a new trade-off between established general principles of good administration in the new era of New Public Analytics. For that exercise, the analytical construct of NPA introduced by Karen Yeung is very helpful not only to rethink administrative law after ‘the algorithmic turn’, but also to critically interrogate various dimensions of NPA through the lens of administrative law.

Competing Interests

The author has no competing interests to declare.