This article considers some of the risks and challenges raised by the use of algorithm-assisted decision-making and predictive tools by the public sector. Alongside, it reviews a number of ...long-standing English administrative law rules designed to regulate the discretionary power of the state. The principles of administrative law are concerned with human decisions involved in the exercise of state power and discretion, thus offering a promising avenue for the regulation of the growing number of algorithm-assisted decisions within the public sector. This article attempts to re-frame key rules for the new algorithmic environment and argues that 'old' law-interpreted for a new context-can help guide lawyers, scientists and public sector practitioners alike when considering the development and deployment of new algorithmic tools.
This article is part of a discussion meeting issue 'The growing ubiquity of algorithms in society: implications, impacts and innovations'.
Criminal law’s efficient and accurate administration depends to a considerable extent on the ability of decision-makers to identify unique individuals, circumstances and events as instances of ...abstract terms (such as events raising ‘reasonable suspicion’) laid out in the legal framework. Automated Facial Recognition has the potential to revolutionise the identification process, facilitate crime detection, and eliminate misidentification of suspects. This paper commences from the recent decision regarding the deployment of AFR by South Wales Police in order to discuss the lack of underpinning conceptual framework pertinent to a broader consideration of AFR in other contexts. We conclude that the judgment does not give the green light to other fact sensitive deployments of AFR. We consider two of these: a) use of AFR as a trigger for intervention short of arrest; b) use of AFR in an evidential context in criminal proceedings. AFR may on the face of it appear objective and sufficient, but this is belied by the probabilistic nature of the output, and the building of certain values into the tool, raising questions as to the justifiability of regarding the tool’s output as an ‘objective’ ground for reasonable suspicion. The means by which the identification took place must be disclosed to the defence, if Article 6 right to a fair trial is to be upheld, together with information regarding disregarded ‘matches’ and error rates and uncertainties of the system itself. Furthermore, AFR raises the risk that scientific or algorithmic findings could usurp the role of the legitimate decision-maker, necessitating the development of a framework to protect the position of the human with decision-making prerogative.
Algorithm-assisted decision-making in the public sector Oswald, Marion
Philosophical transactions of the Royal Society of London. Series A: Mathematical, physical, and engineering sciences,
09/2018, Volume:
376, Issue:
2128
Journal Article
Peer reviewed
This article considers some of the risks and challenges raised by the use of algorithm-assisted decision-making and predictive tools by the public sector. Alongside, it reviews a number of ...long-standing English administrative law rules designed to regulate the discretionary power of the state. The principles of administrative law are concerned with human decisions involved in the exercise of state power and discretion, thus offering a promising avenue for the regulation of the growing number of algorithm-assisted decisions within the public sector. This article attempts to re-frame key rules for the new algorithmic environment and argues that ‘old’ law—interpreted for a new context—can help guide lawyers, scientists and public sector practitioners alike when considering the development and deployment of new algorithmic tools.
This article is part of a discussion meeting issue ‘The growing ubiquity of algorithms in society: implications, impacts and innovations’.
As the first of its kind in UK policing, the West Midlands Police and Crime Commissioner and West Midlands Police data ethics committee is an ongoing experiment in scrutinising and advising upon AI ...policing projects proposed for real operational environments, with the aim of putting people's rights at the heart of technological development. Using a qualitative action research approach akin to an 'observing participant', this paper suggests that lessons can be learned from the committee's activities in three main areas: i) the contribution to effective accountability in respect of ongoing data analytics projects; ii) the importance of the legal and scientific aspects of the interdisciplinary analysis; and iii) the role of necessity and the human rights framework in guiding the committee's ethical discussion. The big themes underpinning the committee proceedings demonstrate the operationalisation of many of the key factors that must be considered in the human rights necessity test. The technical and statistical aspects of policing AI cannot, and should not, be isolated from the legal, contextual, operational and ethical considerations, as each will influence the other, and thus how technology is evaluated. It is important however that laws applicable to specific policing activities are not overlooked, such as those relating to stop-and-search and its application to algorithmic tools. A three-pillar approach could contribute to achieving trustworthy and accountable use of emerging technologies in UK policing: first, governing law plus guidance and policy interpreted for the relevant context; secondly, standards, both ethical standards attached to personal responsibility and scientific standards; and thirdly, people at every level within policing who are committed to accountability; all of which should be subject to rolling independent oversight. In order to ensure a stable three-pillar approach however, there needs to be fewer generalisations and more specifics as regards the application of relevant law to the deployment of emerging technologies, recognising the importance of a human rights-based approach, combined with agreed scientific standards to help us decide whether things 'work' in the policing context. The West Midlands committee has provided additional and transparent positive pressure on the force to improve – productive challenge - as well as a certain level of validation for the projects that have progressed. A national model based on the West Midlands prototype could contribute to necessary assurance and monitoring and the development of specific policy, provided that the status, resourcing and operational challenges identified in this paper are addressed, and subject to linkage with appropriate regulation and enforcement.
Contemporary discussions and disagreements about the deployment of machine learning, especially in criminal justice contexts, have no foreseeable end. Developers, practitioners and regulators could ...however usefully look back one hundred years to the similar arguments made when polygraph machines were first introduced in the United States. While polygraph devices and machine learning operate in distinctly different ways, at their heart, they both attempt to predict something about a person based on how others have behaved. This paper, through an historical perspective, examines the development of the polygraph within the justice system - both in courts and during criminal investigations - and draws parallels to today's discussion. It can be argued that the promotion of lie detectors supported a reforming legal realist approach, something that continues today in the debates over the deployment of machine learning where 'public good' aims are in play, and raises questions around how key principles of the rule of law can best be upheld. Finally, this paper will propose a number of regulatory solutions informed by the early lie detector experience.
Catalytic P systems and purely catalytic P systems are among the first variants of membrane systems ever considered in this area. These variants of systems also feature some prominent computational ...complexity questions, and in particular the problem if only one catalyst in catalytic P systems and two catalysts in purely catalytic P systems are enough to allow for generating all recursively enumerable sets of multisets. Several additional ingredients have been shown to be sufficient for obtaining such results.
Previously, we could show that using the derivation mode maxobjects, where we only take those multisets of rules which affect the maximal number of objects in the underlying configuration, one catalyst is sufficient for obtaining computational completeness without any other ingredients in catalytic P systems. In this paper we investigate the question whether we can obtain a similar result for purely catalytic P systems, i.e., we show that two catalysts in purely catalytic P systems are enough to allow for generating all recursively enumerable sets of multisets when using specific variants of the maximally parallel derivation mode: we take only those applicable multisets of rules which (i) generate the maximal number of objects, or (ii) yield the maximal difference in the number of objects between the newly generated configuration and the current configuration.
In addition, we also consider non-extendable multisets of rules which (i) generate the minimal number of objects, or (ii) yield the minimal difference in the number of objects between the newly generated configuration and the current configuration.
In all cases, we have also shown that register machines with n decrementable registers can be simulated by simple purely catalytic P systems working in any of these derivation modes using only n catalysts. Hence, simple purely catalytic P systems working in any of these derivation modes are computationally complete.
In the early twentieth century, it was still possible to be relatively anonymous at a large gathering, to be visible, yet not the subject of detailed scrutiny or surveillance. A century on, the ...impact of digital technology has reduced our expectations of privacy, whether physical or online. This article discusses the interpretation of 'private' and 'public' in today's technologically enabled world by reference in particular to case-law on the reasonable expectation of privacy. The article goes on to discuss the potential of technological methods for controlling, blocking and obfuscating digital information and devices as means for individuals to regain control over their privacy, ultimately concluding that these technologies, themselves alone, do not provide a long term solution to privacy harms. Finally, the article puts forward an alternative model for consideration pursuant to which certain information about individuals available to the 'masses' digitally or on the Internet, or which can be generated from such information, should no longer be regarded as 'public' in the sense of there being no privacy in respect of it. Thus, the term 'private' when applied to the digital world must be redefined.
We extend and refine previous results within the general framework for regulated rewriting based on the applicability of rules in sequential grammars 3. Besides the well-known control mechanisms as ...control graphs, matrices, permitting and forbidden rules, partial order on rules, and priority relations on rules we also consider the new variant of activation and blocking of rules as investigated in 1, 2, 4. Moreover, we exhibit special results for strings and multisets as well as for arrays in the general variant defined on Cayley grids of finitely presented groups. Especially we prove that array grammars defined on Cayley grids of finitely presented groups using #-context-free array productions together with control mechanisms as control graphs, matrices, permitting and forbidden rules, partial order on rules, priority relations on rules, or activation and blocking of rules have the same computational power as such array grammars using arbitrary array productions.
PDF
