Joe Tomlinson

The Administrative Justice Trade, with Jed Meers (under contract with BUP)

Are we willing to sacrifice participation for speed of decision-making? Should we regulate AI in government through guidance or legislation? Should complaints bodies spend more time on individual cases or investigating systemic issues? Over the last century, we have come to recognise that the messy realities of administrative justice involve complex trade-offs such as these for public authorities. Yet, administrative justice analysis often stops at observing how intractable this all is, leaving officials to develop strategies to cope with the everyday judgement calls involved in resolving these trade-offs. This book shows how empirical research, particularly experimental survey methods, can help bring public perspectives into how to respond to these challenges in ways that maximise trust and legitimacy, and advance critical insights into how administrative justice works.

WORKING PAPERS

'Waiting for Bureaucratic Justice', with Jed Meers, Michela Tinelli, Sanna-Leea Read, Simon Halliday, and Eppie Leishman 

Accessing public services almost always involves waiting. Across all areas of the administrative state, delays and waiting lists are a common experience for the public accessing systems and a challenge for the officials tasked with improving them. Yet scholarship on bureaucratic justice has largely treated waiting in zero-sum terms: delay is a cost to be minimised in the spirit of ‘justice delayed is justice denied’. Drawing on a discrete choice experiment with a representative sample of the UK public, this article challenges that framing. The results demonstrate that the acceptability of waiting is not fixed but is shaped by the quality of bureaucratic justice in the process as a whole. Where certain process qualities were present (transparency, empathy, voice, personalisation, assistance and responsibility-taking) longer waiting times were more acceptable to participants. On average, participants were willing to wait 2.3 times longer for a process where all of them are present, compared to one where they are absent. These findings suggest that the legitimacy of waiting cannot be understood in isolation from the broader process and that improving bureaucratic justice can pay a double dividend: enhancing overall acceptability while extending the window of tolerable waiting time that makes such improvements possible.

'Hyperefficient Bureaucratic Rationality: Administrative Justice, Fraud and Error, and the Digital Welfare State', with Mark Bennett and Jed Meers

This article analyses administrative justice in large-scale fraud and error detection systems in digitalised social security bureaucracies. Rooted in an original qualitative case study of the UK’s Targeted Case Review programme, we show how new digital infrastructures enable officials to pursue the classic ideal-type of bureaucratic rationality—underpinned by legitimacy claims of enabling accuracy and efficiency—with greater intensity. From the welfare agency’s perspective, these techniques can be rationally seen as successful. However, this hyperefficient form of bureaucratic rationality also intensifies the established risks of injustice inherent in implementing this model, especially for those vulnerable to the learning, compliance, and psychological burdens likely to be generated by exercises of administrative authority. This brings a paradox into focus: implementation experienced as burdensome or unjust risks undermining the claimant trust necessary for the everyday cooperation error prevention systems rely upon. We propose exploring the possibilities of realist reform in this setting: an agenda which, while accepting as credible the need for fraud and error prevention mechanisms in welfare administration, explores whether claimant-centred ideas of justice can be meaningfully accommodated in a process such as the Targeted Case Review. To this end, we present findings from a quantitative survey to explore the possible effects of claimant-centred bureaucratic justice.

'Process Evaluations and System Legitimacy in Reformed Welfare Services,' with Aleksandra Chichoka, Ben Seyd, Jed Meers, Simon Halliday

Within the context of modern, individualised, welfare services, it is important to understand how claimants experience and evaluate the processes through which allocative decisions are made on benefit applications. While an extensive literature has explored the relationship between procedural evaluations and legitimacy assessments across a variety of contexts, this coverage has not extended to service users’ interactions with state welfare services. This study seeks to close this gap, and to further our understanding of how service users evaluate bureaucratic decision processes, through a study of claimants within the UK’s Universal Credit (UC) system. Drawing on qualitative fieldwork, a large-N survey of UC claimants and a conjoint experiment, the study identifies how claimants evaluate different processes within the benefits system, and how these evaluations relate to wider assessments of system legitimacy. The study extends our understanding of procedural justice while also identifying the procedural factors that contribute to the legitimacy of a modern welfare system.

‘Implementing Claimant-Centred Bureaucratic Justice in the Digital Welfare State’, with Jed Meers

In recent years, efforts have been made to articulate a new way of thinking about administrative justice in large-scale public sector systems underpinned by a model often referred to as ‘bureaucratic rationality.’ The aim was to establish whether it is feasible to implement bureaucratic rationality in a way individuals would perceive as more just. In the context of the UK’s social security administration, this line of research has already produced new evidence-based definitions of what claimant-centric implementation means, demonstrated that it appears possible to achieve differential perceptions via service alignment with these definitions, and developed evidence that these perceptions can have associations with substantive beneficial outcomes for authorities through generating pro-system behaviours in line with policy goals. This article advances understanding in this domain by examining, through a series of survey experiments, whether the theory of claimant-centric bureaucratic justice holds up when tested in hypothetical interventions within the Universal Credit digital platform. In doing so, it brings closer the possibility of a social security administration model built on bureaucratic justice.

‘Hard Choices about Soft Law: An Experimental Study of Rule-based Regulation of Artificial Intelligence in Public Administration,’ with Jed Meers, Alexandra Sinclair, and Kimberlee Weatherall

As public administration uptake of artificial intelligence (AI) accelerates, a debate has emerged over whether and how new rules should be developed to govern its use, and in particular what legal form any future regulation should take. This article addresses a central question in this debate: whether any new rules ought to take the form of soft-law guidance or legislation. Using a survey experiment, we investigate whether and how the public responds differently when the same measure—in our experiment, transparency rules for public sector AI use—is described as either (hard) law or (soft) guidelines. We observe that describing proposed rules as laws rather than guidelines causes increased perceptions of administration’s likely effectiveness and ability to promote just treatment of the individual. We further find some evidence that introducing laws or guidelines for AI transparency would not adversely affect perceptions of public-sector AI innovation. This new evidence has implications for the UK’s current regulatory approach—which has thus far produced only soft law under the rubric of promoting a pro-innovation environment—as well as how we generally think about sustaining the legitimacy of administrative governance through this period of technological change.

‘Artificial Intelligence, Administrative Tribunals, and Procedural Justice,’ with Jed Meers

As tribunals handling administrative appeals begin to adopt artificial intelligence (AI), the question of how they can maintain trust and legitimacy comes to the fore. This chapter presents findings from a vignette survey experiment examining the effects of AI adoption on public perceptions of procedural justice in tribunals. We examine differences in procedural justice perceptions related to whether AI is used, specific AI use cases, the tribunal jurisdiction in which AI is used, and between people with and without experience of tribunals. We find that the use of AI significantly reduced scores on every procedural justice measure, but that the strength of this effect depended on context. Building on these empirical findings, we propose four ‘rules of thumb’ that can assist tribunal institutions in navigating dilemmas about whether and how to adopt AI while maintaining trust and legitimacy: (i) assume that AI use in tribunals attracts a penalty in perceptions of procedural justice; (ii) assume that the extent of this penalty varies on contextual features such as use case and jurisdiction; (iii) assume rational predictions, including by lawyers, about how context shapes this variation are not a reliable guide; and (iv) assume that how the use of AI is explained to people affects the size of the AI penalty. 

‘Scans, Streets, and Bicycles: Artificial Intelligence Futures in Administrative Tribunals,’ with Susannah Paul

Artificial intelligence (AI) is increasingly seen as a major force shaping the future development of administrative tribunals. As debate grows over the risks and rewards of this institutional shift, this paper focuses on an under-analysed issue in how adoption is unfolding: the risk that institutional approaches to AI adoption may be constrained by an unduly narrow view of the possible use cases and their value in this setting. Drawing on qualitative socio-legal research with tribunal professionals and users, it develops a bottom-up approach to this aspect of AI adoption, referred to by officials as the “scan” phase of public AI innovation. The paper makes three contributions. First, it outlines and reflects on the value and limits of using socio-legal interview methodologies to inform ‘scan’ activities. Second, drawing on the original qualitative dataset, it presents a novel taxonomy of possible AI use cases identified by participants who engage with tribunals at the everyday, ‘street’ level, moving beyond less systematic speculation on possible uses that has characterised debate thus far. Third, it analyses how participants broadly expect AI-enabled futures in tribunals to unfold, and what this reveals about whether AI is imagined as a driver of automation (replacing functions currently performed by humans) or transformation (functions where AI may be able to perform in ways humans cannot). 

‘The Transformation of the Tribunal Hearing: An Empirical Study of Justice in Online Mental Health Appeals,’ with Magdalena Furgalska and Cassandra Somers-Joce

In recent years, many tribunal hearings have moved online, and many jurisdictions now handle most cases remotely. This transformation of a key institution of justice has received relatively little study so far. Drawing on the largest qualitative study yet undertaken of experiences of remote mental health tribunal hearings, we analyse perspectives on the quality of justice provided by such hearings in respect of three issues: choice, process, and outcome.