Roger Clarke's Web-Site
© Xamax Consultancy Pty Ltd, 1995-2021
|Identity Matters||Other Topics||Waltzing Matilda||What's New|
Revised Version of 20 February 2019
(Added ss 5.1 and 5.9, additional citations, editorials, re-formatting)
(Substantial further development of s.7, addition of Table 4 and Appendices)
(Refinements to the 10 and 50 Principles, addition of 25th source-document)
(Addition of 26th source-document)
(Revisions to the 10 Principles, and to the 50 Principles and cross-references)
Roger Clarke **
© Xamax Consultancy Pty Ltd, 2018-19
Available under an AEShareNet licence or a Creative Commons licence.
This document is at http://www.rogerclarke.com/EC/GAIF.html
Organisations across the private and public sectors are looking to use artificial intelligence (AI) techniques not only to draw inferences, but also to make decisions and take action, and even to do so autonomously. This is despite the absence of any means of programming values into technologies and artefacts, and the obscurity of the rationale underlying inferencing using contemporary forms of AI.
To what extent is AI really suitable for real-world applications? Can corporate executives satisfy their board-members that the business is being managed appropriately if AI is inscrutable? Beyond operational management, there are compliance risks to manage, and threats to important relationships with customers, staff, suppliers and the public. Ill-advised uses of AI need to be identified in advance and nipped in the bud, to avoid harm to important values, both corporate and social. Organisations need to extract the achievable benefits from advanced technologies rather than dreaming dangerous dreams.
This working paper first considers several approaches to addressing the gap between the current round of AI marketing hype and the hard-headed worlds of business and government. It is first proposed that AI needs to be re-conceived as 'complementary intelligence', and that the robotics notion of 'machines that think' needs to give way to the idea of 'intellectics', with the focus on 'computers that do'.
A review of 'ethical analysis' of IT's impacts extracts little of value. A consideration of regulatory processes proves to be of more use, but to still deliver remarkably little concrete guidance. It is concluded that the most effective approach for organisations to take is to apply adapted forms of the established techniques of risk assessment and risk management. Critically, stakeholder analysis needs to be performed, and risk assessment undertaken, from those perspectives as well as from that of the organisation itself. This Working Paper's final contribution is to complement that customised form of established approaches to risk by the presentation of a derivative set of Principles for Responsible AI, with indications provided of how those Principles can be operationalised for particular forms of complementary intelligence and intellectics.
The term Artificial Intelligence (AI) was coined in 1955 in a proposal for the 1956 Dartmouth Summer Research Project in Automata (McCarthy et al. 1955). The proposal was based on "the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it". Histories of AI (e.g. Russell & Norvig 2009, pp. 16-28) identify multiple strands, but also multiple re-visits to much the same territory, and a considerable degree of creative chaos.
The over-enthusiasm that characterises the promotion of AI has deep roots. Simon (1960) averred that "Within the very near future - much less than twenty-five years - we shall have the technical capability of substituting machines for any and all human functions in organisations. ... Duplicating the problem-solving and information-handling capabilities of the brain is not far off; it would be surprising if it were not accomplished within the next decade". Over 35 years later, with his predictions abundantly demonstrated as being fanciful, Simon nonetheless maintained his position, e.g. "the hypothesis is that a physical symbol system [of a particular kind] has the necessary and sufficient means for general intelligent action" (Simon 1996, p. 23 - but expressed in similar terms from the late 1950s, in 1969, and through the 1970s), and "Human beings, viewed as behaving systems, are quite simple" (p. 53). Simon acknowledged "the ambiguity and conflict of goals in societal planning" (p. 140), but his subsequent analysis of complexity (pp. 169-216) considered only a very limited sub-set of the relevant dimensions. Much the same dubious assertions can be found in, for example, Kurzweil (2005): "by the end of the 2020s" computers will have "intelligence indistinguishable to biological humans" (p.25), and in self-promotional documents of the current decade.
AI has offered a long litany of promises, many of which have been repeated multiple times, on a cyclical basis. Each time, proponents have spoken and written excitedly about prospective technologies, using descriptions that not merely verged into the mystical, but often crossed the border into the realms of magic and alchemy. Given the habituated exaggerations that proponents indulge in, it is unsurprising that the field has exhibited cyclical 'boom and bust' patterns, with research funding being sometimes very easy to obtain, and sometimes very difficult, depending on whether the focus at the time is on the hyperbole or on the very low delivery-rate against promises.
Part of AI's image-problem is that most of the successes deriving from what began as AI research have shed the name, and become associated with other terms. For example, pattern recognition, variously within text, speech and two-dimensional imagery, has made a great deal of progress, and achieved application in multiple fields, as diverse as dictation, vehicle number-plate recognition and object and facial recognition. Expert systems approaches, particularly based on rule-sets, have also achieved a degree of success. Game-playing, particularly of chess and go, have surpassed human-expert levels, and provided entertainment value and spin-offs, but seem not to have provided the breakthroughs towards posthumanism that their proponents appeared to be claiming for them.
This Working Paper concerns itself with the question of how organisations can identify AI technologies that have practical value, and apply them in ways that achieve benefits, without incurring disproportionate disbenefits or giving rise to unjustified risks. A key feature of AI successes to date appears to be that, even where the technology or its application is complex, it is understandable by people with appropriate technical background, i.e. it is not magic and is not presented as magic, and its applications are auditable. AI technologies that have been effective have been able to be empirically tested in real-world contexts, but under sufficiently controlled conditions that the risks have been able to be managed.
The scope addressed in this Working Paper is very broad, in terms of both technologies and applications, but it does not encompass design and use for warfare or armed conflict. It does, however, include applications to civil law enforcement and domestic national security, i.e. safeguards for the public, for infrastructure, and for public figures.
This working paper commences by considering interpretations of the AI field that may contribute to overcoming its problems and assist in analysing the opportunities and threats that it embodies. Brief scans are undertaken of current technologies that are within the field of view. There are several possible sources of guidance in relation to the responsible use of AI. The paper first considers ethics, and then regulatory regimes. It proposes, however, that the most useful approach is through risk assessment and management processes, but expanding the perspectives from solely that of the organisation itself to also embrace those of stakeholders. The final section draws on the available sources in order to propose a set of principles for the responsible application of AI that are specific enough to guide organisations' business processes.
A major contributor to AI's problems has been the diverse and often conflicting conceptions of what it is, and what it is trying to achieve. The first necessary step is to disentangle the key ideas, and adopt an interpretation that can assist user organisations to appreciate the nature of the technology, and then analyse its potential contributions and downsides.
What does, what could, and what should 'intelligence' mean? What does 'artificial' mean? And are the conventional interpretations of these terms useful to individual organisations, and to the economy and society more generally?
The general sense in which the term 'intelligence' is used by the AI community is that an entity exhibits intelligence if it has perception and cognition of (relevant aspects of) its environment, has goals, and formulates actions towards the achievement of those goals (Albus 1991, Russell & Norvig 2003, McCarthy 2007). Some AI proponents strive to replicate in artefacts the processes whereby human entities exhibit intelligence, whereas others define AI in terms of the artefact's performance rather than the means whereby the performance arises.
The term 'artificial' has always been problematic. The originators of the term used it to mean 'synthetic', in the sense of being human-made but equivalent to human. It is far from clear that there was a need for yet more human intelligence in 1955, when there were 2.8 billion people, let alone now, when there are over 7 billion of us, many under-employed and likely to remain so.
Some proponents have shifted away from human-equivalence, and posited that AI is synthetic, but in some way 'superior-to-human'. This raises the question as to how superiority is to be measured. For example, is playing a game better than human experts necessarily a useful measure? There is also a conundrum embedded in this approach: if human intelligence is inferior, how can it reliably define what 'superior-to-human' means?
An alternative approach may better describe what humankind needs. An idea that is traceable at least to Wyndham (1932) is that " ... man and machine are natural complements: They assist one another". I argued in Clarke (1989) that there was a need to "deflect the focus ... toward the concepts of 'complementary intelligence' and 'silicon workmates' ... to complement human strengths and weaknesses, rather than to compete with them". Again, in Clarke (1993), reprised in Clarke (2014b), I reasoned that: "Because robot and human capabilities differ, for the foreseeable future at least, each will have specific comparative advantages. Information technologists must delineate the relationship between robots and people by applying the concept of decision structuredness to blend computer-based and human elements advantageously".
Adopting this approach, AI needs to be re-conceived such that its purpose is to extend human capabilities. Rather than 'artificial' intelligence, the design objective needs to be 'complementary' intelligence, the essence of which is:
An important category of 'complementary intelligence' is the use of negative-feedback mechanisms to achieve automated equilibration within human-made systems. A longstanding example is the maintenance of ship trim and stability by means of hull shape and careful weight distribution, including ballast. A more commonly celebrated instance is Watts' fly-ball governor for regulating the pressure in a boiler. Of more recent origin are schemes to achieve real-time control over the orientation of craft floating in fluids, and maintenance of their location or path. There are successful applications to deep-water oil-rigs, underwater craft, and aircraft both with and without pilots on board. The notion is also exemplified by the distinction between decision support systems (DSS), which are designed to assist humans make decisions, and decision systems (DS), whose purpose is to make the decisions without human involvement.
Computer-based systems have a clear advantage over humans in contexts in which significant computation is involved, reliability and accuracy are important, and speed of inferencing, decision-making and/or action-taking, are important. This advantage is, however, limited to circumstances in which either a structured process exists or heuristics or purely empirical techniques have been well-demonstrated to be effective.
Further advantages may arise in relation to cost, the delegation to devices of boringly mundane tasks, and the performance by artefacts of tasks that are inherently dangerous, or that need to be performed in environments that are inherently dangerous to humans and/or are beyond their physical capabilities (e.g. environments that feature high pressure such as deep water, low pressure such as space, or high radiation levels both in space and close to nuclear materials). Even where such superiority can be demonstrated, however, the need exists to focus discussion about AI on complementary intelligence, on technologies that augment human capabilities, and on systems that feature collaboration between humans and artefacts.
I contend that the use of the complementary intelligence notion can assist organisations in their efforts to distinguish uses of AI that have prospects for adoption, the generation of net benefits, the management of disbenefits, and the achievement of public acceptability.
The concept of 'automation' is concerned with the performance of a predetermined procedure, or response in predetermined ways to alternative stimuli. It is observable in humans, e.g. under hypnosis, and is designed-into many kinds of artefacts.
The rather different notion of 'autonomy' means, in humans, the capacity for independent decision and action. Further, in some contexts, it also encompasses a claim to the right to exercise that capacity. It is associated with the notions of consciousness, sentience, self-awareness, free will and self-determination. Autonomy in artefacts, on the other hand, lies much closer to the notion of automation. It may merely refer to a substantial repertoire of pre-programmed stimulus-response relationships. Alternatively, it may refer to some degree of adaptability to context, as might arise if some form of machine-learning were included, such that the specification of the stimulus-response relationships change over time depending on the cases handled in the intervening period. Another approach might be to define artefact autonomy in terms of the extent to which a human or some other artefact, does, or even can, intervene in the artefact's behaviour.
In humans, autonomy is best approached as a layered phenomenon. Each of us performs many actions in a subliminal manner. For example, our eye and ear receptors function without us ever being particularly aware of them, and several layers of our neural systems handle the signals in order to offer us cognition, that is to say awareness and understanding, of the world around us.
A layered approach is applicable to artefacts as well. Aircraft generally, including drones, may have layers of behaviour that occur autonomously, without pilot action or even awareness. Maintenance of the aircraft's 'attitude' (orientation to the vertical and horizontal), and angle to the wind-direction, may, from the pilot's viewpoint, simply happen. At a higher level of delegation, the aircraft may adjust the aircraft's flight controls in order to maintain a predetermined flight-path, and in the case of rotorcraft, to maintain the vehicle's location relative to the earth's surface. A higher-order autonomous function is inflight manoeuvring to avoid collisions. At a yet higher level, some aircraft can perform take-off and/or landing autonomously. To date, high-order activities that are seldom if ever autonomous include decisions about when to take off and land, the mission objective, and 'target acquisition' (where to land, where to deliver a payload, which location to direct the payload towards).
At the lower levels, the rapidity with which analysis, decision and action need to be taken may preclude conscious human involvement. At the higher levels, however, a pilot may be able to request advice, to accept or reject advice, to authorise an action recommended by an artefact, to override or countermand a default action, or to resume full control. From the perspective of the drone, its functions may be to perform until its autonomous function is revoked, to perform except where a particular action is over-ridden, to recommend, to advise, or to do nothing.
IEEE, even thought it is one of the most relevant professional associations in the field, made no meaningful attempt to address these issues for decades. It is currently endeavouring to do so. It commenced with a discussion paper (IEEE 2017) which avoids the term AI, and instead uses the term 'Autonomous and Intelligent Systems (A/IS)'. This highlights the need to address both intelligence and autonomy in an integrated manner.
A further factor that has tended to cloud meaningful discussion of responsibility in relation to AI has been inadequate discrimination among the successive phases of the supply-chain from laboratory experiment to deployment in the field, and failure to assign responsibilities to the various categories of entities that are active in each phase.
IEEE's discussion paper (IEEE 2017) recognises that the end-result of successive rounds of R&D is complex systems that are applied in real-world contexts. In order to deliver such systems, however, technology has to be conceived, proven, and embedded in artefacts. It is therefore valuable to distinguish between technology, artefacts that embody the technology, systems that incorporate the artefacts, and applications of those systems. Appropriate responsibilities can then be assigned to researchers, to inventors, to innovators, to purveyors, and to users. Table 1 identifies phases, the output from each phase, and the categories of entity that bear legal and moral responsibility for disbenefits arising from AI.
Installed AI-Based Systems
User Organisations and Individuals
This section has proposed several measures whereby the fog induced by the AI notion can be lifted, and a framework developed for managing AI-based activities. The focus needs to be on complementary intelligence and autonomy, as features of technology, artefacts, systems and applications that support collaboration among all system elements.
AI's scope is broad, and contested. This section identifies areas that have current relevance. Their relevance derives in part from claims of achievement of progress and benefits, and in part from media coverage resulting in awareness among both organisations' staff and the general public. In addition to achieving some level of adoptiong, each faces to at least some degree technical challenges, public scepticism and resistance. Achievement of the benefits that are potentially extractable from these technologies is also threatened by over-claiming, over-reach, and resulting loss of public confidence. This section considers three forms of AI, and then suggests an alternative conceptualisation intended to assist in understanding and addressing the technical, acceptance and adoption challenges.
Robotics originally emerged in the form of machines enhanced with computational capacity. The necessary elements are sensors to acquire data from the robot's environment, computing hardware and software to enable inferences to be drawn and decisions made, and actuators in order to give effect to those decisions by acting on the robot's environment. Robotics has enjoyed its major areas of success in controlled environments such as the factory floor and the warehouse. Less obviously 'robotic' systems include low-level control over the attitude, position and course of craft on or in water and in the air.
The last few years have seen a great deal of coverage of self-driving vehicles, variously on rails and otherwise, in controlled environments such as mines and quarries and dedicated bus routes, and recently in more open environments. In addition, robotics has taken flight, in the form of drones (Clarke 2014a).
Many claims have been made recently about 'the Internet of Things' (IoT) and about systems comprising many small artefacts, such as 'smart houses' and 'smart cities'. For a consolidation and rationalisation of multiple such ideas into the notion of an 'eObject', see Manwaring & Clarke (2015). Many of the initiatives in this area are robotic in nature, in that they encompass all of sensors, computing and actuators.
The term cyborgisation refers to the process of enhancing individual humans by technological means, such that a cyborg is a hybrid of a human and one or more artefacts (Clarke 2005, Warwick 2014). Many forms of cyborg fall outside the field of AI, such as spectacles, implanted lenses, stents, inert hip-replacements and SCUBA gear. However, a proportion of the artefacts that are used to enhance humans include sensors, computational or programmatic 'intelligence', and one or more actuators. Examples include heart pacemakers (since 1958), cochlear implants (since the 1960s, and commercially since 1978), and some replacement legs for above-knee amputees, in that the artificial knee contains software to sustain balance within the joint.
Many such artefacts replace lost functionality, and are referred to as prosthetics. Others, which can be usefully referred to as orthotics, provide augmented or additional functionality (Clarke 2011). An example of an orthotic is augmented reality for firefighters, displaying building plans and providing object-recognition in their visual field. It was argued in Clarke (2014b) that use by drone pilots of instrument-based remote control, and particularly of first-person view (FPV) headsets, represent a form of orthotic cyborgisation.
Artefacts of these kinds are not commonly included in catalogues of AI technology. On the other hand, they have a great deal in common with it, and with the notion of complementary intelligence, and research in the field is emergent (Zhaohui et al. 2016). Cyborgisation has accordingly been defined as being within-scope of the present analysis.
Computing applications for drawing inferences from data began with hard-wired, machine-level and assembler languages (1940-1960), but made great progress with genuinely 'algorithmic programs', in languages such as ForTran (formula translator). That approach involves an implied problem that needs to be solved, and an explicit procedural solution to that problem. During the 1980s, additional means of generating inferences became mainstream, including logic programming and rule-based ('expert') systems. These embody no explicit 'problem' or 'solution'. They instead define a 'problem-domain': some form of modelling of the relevant real world is undertaken, and the model is expressed in a form that enables inferences to be drawn from it.
AI research has delivered a further technique, which accords primacy to the data rather than the model, and has the effect of obscuring the model to such an extent that no humanly-understandable rationale exists for the inferences that are drawn. The relevant branch of AI is 'machine learning' (ML), and the most common technique in use is 'artificial neural networks'. The approach dates to the 1950s, but limited progress was made until sufficiently powerful processors were readily available, from the late 1980s.
Neural nets involve a set of nodes (each of which analogous to the biological concept of a neuron), with connections or arcs among them, referred to as 'edges'. Each connection has a 'weight' associated with it. Each node performs some computation based on incoming data and may as a result adapt its internal state, including the weighting on each connection, and may pass output to one or more other nodes. A neural net has to be 'trained'. This is done by selecting a training method (or 'learning algorithm') and feeding a 'training-set' of data to the network in order to load up a set of weightings on the connections between nodes.
Unlike previous techniques for developing software, neural networking approaches do not begin with active and careful modelling of a real-world problem-solution, problem or even problem-domain. Rather than comprising a set of entities and relationships that mirrors the key elements and processes of a real-world system, a neural network model is simply a list of input variables and a list of output variables (and, in the case of 'deep' networks, intermediary variables). If a model exists, in the sense of a representation of the real world, it is implicit rather than express. The weightings imputed for each connection reflect the characteristics firstly of the training-set that was fed in, and secondly of the particular learning algorithm that was imposed on the training-set.
Although algorithms are used in the imputation of weightings on the connections within a neural net, the resulting software is not algorithmic, but rather empirical. This has led some authors to justify a-theoretical mechanisms in general, and to glorify correlation and deprecate the search for causal relationships and systemic analysis generally (Anderson 2008, Mayer-Schonberger & Cukier 2013).
AI/ML may well have the capacity to discover gems of otherwise-hidden information. However, the inferences drawn inevitably reflect any errors and biasses inherent in the implicit model, in the selection of real-world phenomena for which data was created, in the selection of training-set, and in the learning algorithms used to develop the software that delivers the inferences. Means are necessary to assess the quality of the implicit model, of the data-set, of the data-item values, of the training-set and of the learning algorithm, and the compatibility among them, and to validate the inferences both logically and empirically. Unless and until those means are found, and are routinely applied, AI/ML and neural nets must be regarded as unproven techniques that harbour considerable dangers to the interests of organisations and their stakeholders.
Robotics began with an emphasis on machines being enhanced with computational elements and software. However, the emphasis has been shifting. I contend that the conception now needs to be inverted, and the field regarded as computers enhanced with sensors and actuators, enabling computational processes to sense the world and act directly on it. Rather than 'machines that think', the focus needs to be on 'computers that do'. The term 'intellectics' is a useful means of encapsulating that switch in emphasis.
The term has been previously used in a related manner by Wolfgang Bibel, originally in German (1980, 1989). Bibel was referring to the combination of Artificial Intelligence, Cognitive Science and associated disciplines, using the notion of the human intellect as the integrating element. Bibel's sense of the term has gained limited currency, with only a few mentions in the literature and only a few authors citing the relevant papers. The sense in which I use the term here is rather different:
In the new context of intellectics, artefacts go beyond merely drawing inferences from data, in that they generate a strong impulse for an action to be taken in the real world
I suggest the following criteria for assessing whether an artefact should be classified as falling within the field of intellectics:
As a threshold test, the artefact must at least communicate a recommendation to a human
At a higher level, an artefact makes a decision, which will result in action unless over-ridden or countermanded by a human
At the highest level, an artefact makes a decision, and takes action in the real world to give effect to that decision, without providing an opportunity for a human to prevent the action being taken
The effect of implementing intellectics is to at least reduce the moderating effect of humans in the decision-loop, and even to remove that effect entirely. The emergence of intellectics is accordingly bringing into much stronger focus the legitimacy of the inferencing techniques used, and of the inferences that they are leading to. Among the major challenges involved are the difficulty and expense of establishing reliable software (in particular the size of the training-set required), the low quality of a large proportion of the data on which inferencing depends, the significance of and the approach adopted to empty cells within the data-set, and the applicability of the data-analytic technique to the data to which it is applied (Clarke 2016a, 2016c).
The earlier generations of computer-performed inferencing enabled the expression of humanly-understandable explanations. During the procedural programming era, a set of conditions resulted in an output, and the logic of the solution was express in both the software specification and the source-code. In logic-based programming, 'consequents' could be traced back to 'antecedents', and in rule-based systems, which rules 'fired' in order to deliver the output could be documented (Clarke 1991).
That situation changes substantially with AI/ML and its primary technique, neural nets. The model is at best implicit and may be only very distantly related to the real-world it is assumed to represent, the approach is empirical, it depends on a training-set, and it is not capable of generating a humanly-understandable explanation for an inference that has been drawn. The application of such inferences to decision-making, and to the performance of actions in and on the real world, raises serious questions about transparency (Burrell 2016, Knight 2017). A result of the loss of decision transparency is the undermining of organisations' accountability for their decisions and actions. In the absence of transparency, such principles are under threat as evaluation, fairness, proportionality, evidence-based decision-making, and the capacity to challenge decisions (APF 2013).
Applications of a variety of data analytics techniques are already giving rise to public disquiet, even in the case of techniques that are (at least in principle) capable of generating explanations of decision rationale. The most publicly-visible of these are systems for people-scoring, most prominently in financial credit. There are also applications in 'social credit' - although in this case to date only in the PRC (Chen & Cheung 2017). Similar techniques are also applied in social welfare contexts, sometimes with seriously problematical outcomes (e.g. Clarke 2018a). Concerns are naturally heightened where inferencing technologies are applied to prediction - particularly where the technique's effectiveness is assumed rather than carefully tested, published, and subject to challenge. Such approaches result in something approaching pre-destination, through the allocation of individual people to categories and the attribution of future behaviour, in some circumstances even behaviour of a criminal nature.
There is increasing public pressure for explanations to be provided for decisions that are adverse to the interests of individuals and of small business, especially in the context of inscrutable inferencing techniques such as neural networking. The responsibility of decision-makers to provide explanations is implied by the principles of natural justice and procedural fairness. In the EU, since mid-2018, as a consequence of Articles 13.2(f), 14.2(g) and 15.1(h) of the General Data Protection Regulation (GDPR 2018), access must be provided to "meaningful information about the logic involved", "at least in" the case of automated decisions (Selbst & Powles 2017). On the other hand, "the [European Court of Justice] has ... made clear that data protection law is not intended to ensure the accuracy of decisions and decision-making processes involving personal data, or to make these processes fully transparent [and] a new data protection right, the 'right to reasonable inferences', is needed" (Wachter & Mittelstadt 2019).
Re-conception of the field as Intellectics enables focus to be brought to bear on key issues confronting organisations that apply the outcomes of AI research. Intellectics represents a major power-shift towards large organisations and away from individuals. Substantial pushback from the public needs to be anticipated, and new regulatory obligations may be imposed on organisations. The following sections canvass the scope for these concerns to be addressed firstly by ethics, and secondly through regulatory arrangements.
Both the dated notion of AI and the alternative conceptualisations of complementary intelligence and intellectics harbour potentials for harm. So it is important for organisations to carefully consider what factors constrain their freedom of choice and actions. The following section examines the regulatory landscape. This section first considers the extent to which ethics affects organisational applications of technology.
Ethics is a branch of philosophy concerned with concepts of right and wrong conduct. Fieser (1995) and Pagallo (2016) distinguish 'meta-ethics', which is concerned with the language, origins, justifications and sources of ethics, from 'normative ethics', which formulates generic norms or standards, and 'applied ethics', which endeavours to operationalise norms in particular contexts. In a recent paper, Floridi (2018) has referred to 'hard ethics' - that which "may contribute to making or shaping the law" - and 'soft ethics' - which are discussed after the fact.
From the viewpoint of instrumentalists in business and government, the field of ethics evidences several substantial deficiencies. The first is that there is no authority, or at least no uncontestable authority, for any particular formulation of norms, and hence every proposition is subject to debate. Further, as a form of philosophical endeavour, ethics embodies every complexity and contradiction that smart people can dream up. Moreover, few formulations by philosophers ever reach even close to operational guidance, and hence the sources enable prevarication and provide endless excuses for inaction. The inevitable result is that ethical discussions seldom have much influence on real-world behaviour. Ethics is an intellectually stimulating topic for the dinner-table, and graces ex post facto reviews of disasters. However, the notion of 'ethics by design' is even more empty than the 'privacy by design' meme. To an instrumentalist - who wants to get things done - ethics diversions are worse than a time-waster; they're a barrier to progress.
The occasional fashion of 'business ethics' naturally inherits the vagueness of ethics generally, and provides little or no concrete guidance to organisations in any of the many areas in which ethical issues are thought to arise. Far less does 'business ethics' assist in relation to complex and opaque digital technologies. Clarke (2018b) consolidates a collection of attempts to formulate general ethical principles that may have applicability in technology-rich contexts - including bio-medicine, surveillance and information technology. Remarkably, none of them contain any explicit reference to identifying relevant stakeholders. However, a number of norms are frequently-encountered in these sets. These include demonstrated effectiveness and benefits, justification of disbenefits, mitigation of disbenefits, proportionality of negative impacts, supervision (including safeguards, controls and audit), and recourse (including complaints and appeals channels, redress, sanctions, and enforcement powers and resources).
The related notion of Corporate Social Responsibility (CSR), sometimes extended to include an Environmental aspect, can be argued to have an ethical base. In practice, its primary focus is usually on the extraction of public relations gains from organisations' required investments in regulatory compliance. CSR can, however, extend beyond the direct interests of the organisation to include philanthropic contributions to individuals, community, society or the environment.
When evaluating the potential impact of ethics and CSR, it is important to appreciate the constraints on company directors. They are required by law to act in the best interests of each company of which they are a director. Attention to broad ethical questions is generally extraneous to, and even in conflict with, that requirement, except where a business case indicates sufficient benefits to the organisation from taking a socially or environmentally responsible approach. The primary ways in which benefits can accrue are through compliance with regulatory requirements, and enhanced relationships with important stakeholders. Most commonly, these stakeholders will be customers, suppliers and employees, but the scope might extend to communities and economies on which the company has a degree of dependence.
Given the limited framework provided by ethics, the question arises as to the extent to which organisations are subject to legal and social mechanisms that prevent or constrain their freedom to create technologies, and to embody them in artefacts, systems and applications.
AI seems to have been argued by its proponents to be arriving imminently, on a cyclical basis, roughly every decade since 1956. Despite that, it appears that few regulatory requirements have been designed or modified specifically with AI in mind. One reason for this is that parliaments seldom act in advance of new technologies being deployed.
A 'precautionary principle' has been enunciated, whose strong form exists in some jurisdictions' environmental laws, along the lines of 'When human activities may lead to morally unacceptable harm that is scientifically plausible but uncertain, actions shall be taken to avoid or diminish that potential harm' (TvH 2006). More generally, however, the 'principle' is merely an ethical norm to the effect that 'If an action or policy is suspected of causing harm, and scientific consensus that it is not harmful is lacking, then the burden of proof arguably falls on those taking the action'. Where AI appears likely to be impactful on the scale that its proponents suggest, surely the precautionary principle applies, at the very least in its weak form. On the other hand, the considerable impacts of such AI technologies as automated number-plate recognition (ANPR), 'facial recognition' and drones have not been the subject even of effective after-the-fact regulatory adaptation or innovation, let alone of proactive protective measures.
A large body of theory exists relating to regulatory mechanisms (Braithwaite & Drahos 2000). Regulation takes many forms, including intrinsic and natural controls, self-control, several levels of 'soft' community controls, various kinds of 'formal' or 'hard' regulatory schemes, and regulation by infrastructure or 'code'. An overview of these categories is in Clarke & Bennett Moses (2014), and a relevant analysis is in Clarke (2014c). This section identifies a range of sources that may offer organisations, to some extent guidance, and at least insights into what society expects, and obligations that organisations might be subject to.
AI-based technologies may be subject to intrinsic limitations, or may stimulate natural processes whose effect is to prevent the adoption occurring or continuing, or to curb or mitigate negative impacts. It is appropriate to consider these first. This is because, in the absence of such mechanisms, a case exists for regulatory measures to be devised and imposed; whereas, if adequate intrinsic or natural controls exist, the costs that regulation would impose on all parties are not justifiable. Economists use the term 'market failure' to refer to the absence or ineffectiveness of harm-limitation mechanisms.
A common circumstance is where too much doubt exists about the technology's ability to deliver on its proponents' promises. Another important natural constraint on adoption is economic factors. (The running expenses involved may be too high, or the number of instances it would apply to and/or the benefits to be gained from each intervention may be too small to justify the investment needed to develop the artefact or implement the system). In some circumstances, the realisation of the potential benefits of a technology may be dependent on infrastructure that is unavailable or inadequate. (For example, computing could have exploded in the third quarter of the 19th century, rather than 100 years later, had metallurgy of the day been able to support Babbage's 'difference' and 'analytical' engines). Another form of control is the opposition of players with sufficient market power (including competitors, suppliers, customers and employees) or institutional power (e.g. regulators, financiers, insurers). It has long ben feasible for opponents to stir up public opprobrium through the media, and further opportunities are now provided by social media.
It appears, however, that currently-promoted forms of AI may not be subject to adequate intrinsic and natural controls. The following sub-sections accordingly consider each of the various forms of regulatory intervention, beginning at the apex of the regulatory pyramid with 'hard law'.
In-place industrial robotics, in production-lines and warehouses, is well-established. Various publications have discussed general questions of robot regulation (e.g. Leenes & Lucivero 2014, Scherer 2016, HTR 2018a, 2018b), but fewer identify AI-specific laws. Even such vital aspects as worker safety and employer liability appear to depend not on technology-specific laws, but on generic laws, which may or may not have been adapted to reflect the characteristics of the new technologies.
In HTR (2017), South Korea is identified as having enacted the first national law relating to robotics generally: the Intelligent Robots Development Distribution Promotion Act of 2008. It is almost entirely facilitative and stimulative, and barely even aspirational in relation to regulation of robotics. There is mention of a 'Charter', "including the provisions prescribed by Presidential Decrees, such as ethics by which the developers, manufacturers, and users of intelligent robots shall abide" - but no such Charter appears to exist. A mock-up is at Akiko (2012). HTR (2018c) offers a generic regulatory specification in relation to research and technology generally, including robotics and AI.
In relation to autonomous motor vehicles, a number of jurisdictions have enacted laws. See Palmerini et al. (2014, pp.36-73), Holder et al. (2016), DMV-CA (2018), Vellinga (2017), which reviews laws in the USA at federal level, California, United Kingdom, and the Netherlands, and Maschmedt & Searle (2018), which reviews such laws in three States of Australia. Such initiatives have generally had a strong focus on economic motivations, the stimulation and facilitation of innovation, exemptions from some existing regulation, and limited new regulation or even guidance. One approach to regulation is to leverage off natural processes. For example, Schellekens (2015) argued that a requirement of obligatory insurance was a sufficient means for regulating liability for harm arising from self-driving cars. In the air, legislatures and regulators have moved very slowly in relation to the regulation of drones (Clarke & Bennett Moses 2014, Clarke 2016b).
Automated decision-making about people has been subject to French data protection law for many years. In mid-2018 this became a feature of European law generally, through the General Data Protection Regulation (GDPR) Art. 22, although doubts have been expressed about its effectiveness (Wachter et al. 2017).
On the one hand, it might be that AI-based technologies are less disruptive than they are claimed to be, and that laws need little adjustment. On the other, a mythology of 'technology neutrality' pervades law-making. Desirable as it might be for laws to encompass both existing and future artefacts and processes, genuinely disruptive technologies have features that render existing laws ambiguous and ineffective.
Applications of new technologies are generally subject to existing laws. Particularly with 'breakthrough', revolutionary and disruptive technologies, existing laws are likely to be ill-fitted to the new context, because they were "designed around a socio-technical context of the relatively distant past" (Bennett Moses 2011. p.765), and without knowledge of the new form. In some cases, existing law may hinder new technologies in ways that are unhelpful to both the innovators and those affected by them. In other cases, existing law may have been framed in such a manner that it does not apply to the new form (or judicial calisthenics has to be performed in order to make it appear to apply), even though there would have been benefits if it had done so.
Applications of AI will generally be subject to the various forms of commercial law, particularly contractual obligations including express and implied terms, consumer rights laws, and copyright and patent laws. In some contexts (such as robotics, cyborg artefacts, and AI software embedded in devices), product liability laws may apply. Other laws that assign risk to innovators may also apply, such as the tort of negligence, as may laws of general applicability such as human rights law, anti-discrimination law and data protection law. The obligations that the corporations law assigns to company directors are also relevant. Further sources of regulatory impact are likely to be the laws relating to the various industry sectors within which AI is applied, such as road transport law, workplace and employment law, and health law.
Particularly in common law jurisdictions, there is likely to be a great deal of uncertainty about the way in which laws will be applied by tribunals and courts if any particular dispute reaches them. This acts to some extent as a deterrent against innovation, and can considerably increase the costs incurred by proponents, and delay deployment. From the viewpoint of people who perceive themselves to be negatively affected by the innovation, on the other hand, channels for combatting those threats may be inaccessible, expensive, slow and even entirely ineffectual.
Parliaments struggle to understand and cope with new technologies. An approach to regulation that once appeared to offer promise is co-regulation. Under this arrangement, a parliament establishes a legal framework, including authority, obligations, sanctions and enforcement mechanisms, but without expressing the obligations at a detailed level. This is achieved through consultative processes among advocates for the various stakeholders. The result is an enforceable Code, which articulates general principles expressed in the relevant legislation.
Unfortunately, few instances of effective co-regulation exist, because such processes typically exclude less powerful stakeholders. In any case, there are few signs of parliaments being aware of the opportunity, and of its applicability to Intellectics. In Australia, for example, Enforceable Codes exist that are administered by the Australian Communications and Media Authority (ACMA) in respect of TV and radio broadcasting, and telecommunications, and by the Australian Prudential Regulation Authority (APRA) in respect of banking services. These arrangements succeed both in facilitating business and government activities and in offering a veneer of regulation; but they fail to exercise control over behaviour that the public regards as inappropriate, and hence they have little public credibility.
It is common for parliaments to designate a specialist government agency or parliamentary appointee either to exercise loose oversight over a contested set of activities, or to exercise powers and resources in order to enforce laws or Codes. An important function of either kind of organisation is to provide guidance to both the regulatees and the parties that the scheme is intended to protect. In very few instances, however, does it appear that AI lies within the scope of an existing agency or appointee. Some exceptions may exist, for example in relation to the public safety aspects of drones and self-driving motor vehicles.
As a result, in most jurisdictions, limited guidance appears to exist. For example, six decades after the AI era was launched, the EU has gone no further than a preliminary statement (EC 2018) and a discussion document issued by the Data Protection Supervisor (EDPS 2016). Similarly, the UK Data Protection Commissioner has only reached the stage of issuing a discussion paper (ICO 2017). The current US Administration's policy is entirely stimulative in nature, and mentions regulation solely as a barrier to economic objectives (WH 2018).
Corporations club together for various reasons, some of which can be to the detriment of other parties, such as collusion on bidding and pricing. The activities of industry associations can, however, deliver benefits for others, as well as for their members. In particular, collaborative approaches to infrastructure can improve services and reduce costs for the sector's customers.
It could also be argued that, if norms are promulgated by the more responsible corporations in an industry sector, then misbehaviour by the industry's 'cowboys' would be highlighted. In practice, however, the effect of Industry Codes on corporate behaviour is seldom significant. Few such Codes are sufficiently stringent to protect the interests of other parties, and the absence of enforcement undermines the endeavour. The more marginal kinds of suppliers ignore them, and responsible corporations feel the pinch of competition and reduce their commitment to them. As a result, such Codes act as camouflage, obscuring the absence of safeguards and thereby holding off actual regulatory measures. In the AI field, examples of industry coalitions eagerly pre-countering the threat of regulation include FLI (2017), ITIC (2017), and PoAI (2018).
A more valuable role is played by industry standards. HTR (2017) lists industry standards issued by the International Standards Organisation (ISO) in the AI arena. A considerable proportion of industry standards focus on inter-operability, and on business processes intended to achieve quality assurance. Public safety is also an area of strength, particularly in the field commonly referred to as 'safety-critical systems' (e.g. Martins & Gorschek 2016). Hence some of the physical threats embodied in AI-based systems are able to be avoided, mitigated and managed through the development and application of industry standards; but threats to economic and social interests are seldom addressed.
A role can also be played by professional associations, because these generally balance public needs against self-interest somewhat better than industry associations. Their impact is, however, far less pronounced than that of industry associations. Moreover, the intiatives to date of the two largest bodies are underwhelming, with ACM (2017) using weak forms such as "should" and "are encouraged to", and IEEE (2017) offering lengthy prose but unduly vague and qualified principles. Neither has to date provided the guidance needed by professionals, managers and executives.
It was noted above that Directors of corporations are required by law to pursue the interests of the corporation ahead of all other interests. It is therefore unsurprising, and even to be expected, that organisational self-regulation is almost always ineffectual from the viewpoint of the supposed beneficiaries, and often not even effective at protecting the organisation itself from bad publicity. Recent offerings by major corporations include IBM (Rayome 2017), Google (Pichai 2018) and MS (2018). For an indication of the scepticism with which such documents are met, see Newcomer (2018).
A range of, in most cases fairly vague, principles, have been proposed by a diverse array of organisations. Examples include the European Greens Alliance (GEFA 2016), a British Standard BS 8611 (BS 2016), the UNI Global Union (UGU 2017), the Japanese government (Hirano 2017), a House of Lords Committee (HOL 2018), as interpreted by a World Economic Forum document (Smith 2018), and the French Parliament (Villani 2018).
Although there are commonalities among these formulations, there is also a lot of diversity, and few of them offer usable advice on how to ensure that Intellectics is applied in a responsible manner. The next section draws on the sources identified above, in order to offer practical advice. It places the ideas within a conventional framework, but extends that framework in order to address the needs of all stakeholders rather than just the corporation itself.
One further regulatory element requires consideration. Lessig (1999) popularised the notion of behaviour in socio-technical systems being subject not only to formal law ('East Coast Code'), but also to constraints that exist within computer and network architecture and infrastructure, i.e. standards, protocols, hardware and software ('West Coast Code').
A relevant form that 'West Coast Code' could take is the embedment in robots of something resembling 'laws of robotics'. This notion first appeared in an Asimov short story, 'Runaround', published in 1942; but many commentators on robotics cling to it. For example. Devlin (2016) quotes a professor of robotics as perceiving that the British Standard Institute's guidance on ethical design of robots (BS 2016) represents "the first step towards embedding ethical values into robotics and AI". On the other hand, a study of Asimov's robot fiction showed that he had comprehensively demonstrated the futility of the idea (Clarke 1993). No means exists to encode human values into artefacts, nor to reflect differing values among various stakeholders, nor to mediate conflict among values and objectives.
Ethical analyses offer little assistance, and regulatory frameworks are lacking. It might seem attractive to business enterprises to face few legal obligations and hence to be subject to limited compliance risk exposure. On the other hand, the absence of regulation heightens many other business risks. At least some competitors inevitably exhibit 'cowboy' behaviour, and there are always individuals and groups within each organisation who can be tempted by the promise that AI appears to offer. As a result, there are substantial direct and indirect threats to the organisation's reputation. It is therefore in each organisation's own self-interest for a modicum of regulation to exist, in order to provide a protective shield against media exposés and public backlash.
This section offers guidance to organisations. It assumes that organisations evaluating AI apply conventional environmental scanning and marketing techniques in order to identify opportunities, and a conventional business case approach to estimating the strategic, market-share, revenue, cost and profit benefits that the opportunities appear to offer them. The focus here is on how the downsides can be identified, evaluated and managed.
Familiar, practical approaches to assessing and managing risks are applicable. However, I contend that the conventional framework must be extended to include an important element that is commonly lacking in business approaches to risk. That missing ingredient is stakeholder analysis. Risk assessment and management needs to be performed not only from the business perspective, but also from the perspectives of other stakeholders.
There are many sources of guidance in relation to risk assessment and management. The techniques are well-developed in the context of security of IT assets and digital data, although the language and the approaches vary considerably among the many sources (most usefully: Firesmith 2004, ISO 2005, ISO 2008, NIST 2012, ENISA 2016, ISM 2017). For the present purpose, a model is adopted that is summarised in Appendix 1 of Clarke (2015). See Figure 1.
Existing corporate practice approaches this model from the perspective of the organisation itself. This gives rise to conventional risk assessment and risk management processes outlined in Table 2. Relevant assets are identified, and an analysis undertaken of the various forms of harm that could arise to those assets as a result of threats impinging on, or actively exploiting, vulnerabilities, and giving rise to incidents. Existing safeguards are taken into account, in order to guide the development of a strategy and plan to refine and extend the safeguards and thereby provide a degree of protection that is judged to suitably balance modest actual costs against much higher contingent costs.
Analyse / Perform Risk Assessment
(1) Define the Objectives and Constraints
(2) Identify the relevant Stakeholders, Assets, Values and categories
(3) Analyse Threats and Vulnerabilities
(4) Identify existing Safeguards
(5) Identify and Prioritise the Residual Risks
Design / Initiate Risk Management
(1) Identify alternative Safeguards
(2) Evaluate the alternatives against the Objectives and
(3) Select a Design (or adapt / refine the alternatives to achieve an acceptable Design)
Do / Perform Risk Management
(1) Plan the implementation
(3) Review the implementation
The notion of 'stakeholders' was introduced as a means of juxtaposing the interests of other parties against those of the corporation's shareholders (Freeman & Reed 1983). Many stakeholders are participants in relevant processes, in such roles as employees, customers and suppliers. Where the organisation's computing services extend beyond its boundaries, any and all of those primary categories of stakeholder may be users of the organisation's information systems.
However, the categories of stakeholders are broader than this, comprising not only "participants in the information systems development process" but also "any other individuals, groups or organizations whose actions can influence or be influenced by the development and use of the system whether directly or indirectly" (Pouloudi & Whitley 1997, p.3). The term 'usees' is a usefully descriptive term for these once-removed stakeholders (Clarke 1992, Fischer-Huebner & Lindskog 2001, Baumer 2015).
My first proposition for extension beyond conventional corporate risk assessment is that the responsible application of AI is only possible if stakeholder analysis is undertaken in order to identify the categories of entities that are or may be affected by the particular project (Clarkson 1995). There is a natural tendency to focus on those entities that have sufficient market or institutional power to significantly affect the success of the project. On the other hand, in a world of social media and rapid and deep mood-swings, it is advisable to not overlook the nominally less powerful stakeholders. Where large numbers of individuals are involved (typically, employees, consumers and the general public), it will generally be practical to use representative and advocacy organisations as intermediaries, to speak on behalf of the categories or segments of individuals.
My second proposition is that the responsible application of AI depends on risk assessment processes being conducted from the perspectives of the various stakeholders, to complement that undertaken from the perspective of the corporation. Conceivably, such assessments could be conducted by the stakeholders independently, and fed into the organisation. In practice, the asymmetry of information, resources and power is such that the outputs from independent. and therefore uncoordinated, activities are unlikely to gain acceptance. The responsibility lies with the sponsor of an initiative to drive the studies, engage effectively with the other parties, and reflect their input in the project design criteria and features.
The risk assessment process outlined in Table 2 above is generally applicable. However, my third proposition is that risk assessment processes that reflect the interests of stakeholders needs to be broader than that commonly undertaken within organisations. Relevant techniques include privacy impact assessment (Clarke 2009, Wright & De Hert 2012), social impact assessment (Becker & Vanclay 2003), and technology assessment (OTA 1977). For an example of impact assessment applied to the specific category of person-carrier robots, see Villaronga & Roig (2017). The most practical approach may be, however, to adapt the organisation's existing process in order to encompass whichever aspects of such broader techniques are relevant to the stakeholders whose needs are being addressed.
The results of the two or more risk assessment processes outlined above deliver the information that the organisation needs. They enable the development of a strategy and plan whereby existing safeguards can be adapted or replaced, and new safeguards conceived and implemented. ISO standard 27005 (2008, pp.20-24) discusses four options for what it refers to as 'risk treatment': risk modification, risk retention, risk avoidance and risk sharing. A framework is presented in Table 3 that in my experience is more understandable by practitioners and more readily usable as a basis for identifying possible safeguards.
Existing techniques are strongly oriented towards protection against risks as perceived by the organisation. Risks to other stakeholders are commonly treated as, at best, a second-order consideration, and at worst as if they were out-of-scope. All risk management work involves the exercise of a considerable amount of imagination. That characteristic needs to be underlined even more strongly in the case of the comprehensive, multi-stakeholder approach that I am contending is necessary in the case of AI-based systems.
This section has suggested customisation of existing, generic techniques in order to address the context of AI-based systems. The following section presents more specific proposals.
This section presents a set of Principles for AI. The purpose of doing so is to provide organisations and individuals with guidance as to how they can fulfil their responsibilities in relation to AI and AI-based activities. Because of the broad scope of the AI notion, the considerable diversity among its various forms, and the changes in those forms over time, the Principles proposed below are still somewhat abstract. The intention is to express them in as practically useful a manner as can reasonably be achieved. At the very least, they should provide a firm base for the expression of operational guidance for each specific form of AI.
The Principles in part emerge from the analysis presented in this Working Paper, and in part represent a consolidation of ideas from a suite of previously-published sets of principles. The suite was assembled by surveying academic, professional and policy literatures. Diversity of perspective was actively sought. The sources were governmental organisations (7), non-government organisations (6), corporations and industry associations (5), professional associations (2), joint associations (2), and academics (4). Only sets that were available in the english language were used. This resulted in a strong bias within the suite towards documents that originated in countries whose primary language(s) is or include english. Of the individual documents, 8 are formulations of 'ethical principles and IT'. Extracts and citations are provided at Clarke (2018c). The other 18 claim to provide principles or guidance specifically in relation to AI. Extracts and citations are at Clarke (2018d).
The process of developing the set commenced with themes that derived from the analysis reported on in the earlier sections of this Working Paper. The previously-published sets of principles were then inspected. Detailed propositions within each set were extracted, and allocated to themes, maintaining back-references to the sources. Where items threw doubt on the structure or formulation of the general themes, the schema was adapted in order to sustain coherence and limit the extent to which duplications arise.
The Principles have been expressed in imperative mode, i.e. in the form of instructions, in order to convey that they require action, rather than being merely desirable characteristics, or factors to be considered, or issues to be debated. The full set of Principles, comprising about 50 elements, is in Appendix 1. The 10 over-arching themes are presented in Table 4.
Some of the items that appear in source documents appear incapable of being operationalised. For example, 'human dignity', 'fairness' and 'justice' are vague abstractions that need to be unpacked into more specific concepts. In addition, some items fall outside the scope of the present work. The items that have been excluded from the set in Table 4 are listed in Appendix 2.
In s.2.3 and Table 1 above, distinctions were drawn among the phases of the supply-chain, which in turn produce AI technology, AI-based artefacts, AI-based systems, deployments of them, and applications of them. In each case, the relevant category of entity was identified that bears responsibility for negative impacts arising from AI. In only a few of the 26 documents in the suite were such distinctions evident, however, and in most cases it has to be interpolated which part of the supply-chain the document is intended to address. The European Parliament (CLA-EP 2016) refers to "design, implementation, dissemination and use", IEEE (2017) to "Manufacturers / operators / owners", GEFA (2016) to "manufacturers, programmers or operators", FLI (2017) to researchers, designers, developers and builders, and ACM (2017) to "Owners, designers, builders, users, and other stakeholders". Remarkably, however, in all of these cases the distinctions were only made within a single Principle rather than being applied to the set as a whole.
Some commonalities exist across the source documents. Overall, however, most of the source documents were remarkably sparse, and there was far less consensus that might have been expected more than 60 years after AI was first heralded. For example, only 1 document encompassed cyborgisation (GEFA 2016); and only 2 documents referred to the precautionary principle (CLA-EP 2016, GEFA 2016).
The analysis adopted a conservative approach, whereby a document was scored against a Principle if the idea was in some way evident, even if its coverage of the Principle as a whole was limited. Yet, on average, each Principle was only reflected in 5 of the 26 documents. The most recent document evidenced the largest percentage of the 50 Principles - the European Commission's Draft Ethics Guidelines for Trustworthy AI (EC 2018) - but even that only mustered 28/50 (56%).
It was also striking how few of the 50 Principles were detectable in the majority of the documents. Only 8/26 stipulated 'Conduct impact assessment' (Principle 1.4). Even 'Ensure people's wellbeing ('beneficence')' (4.3) was evident in only 12/26, and only the following four achieved at least half, three of them only just:
Each of the sources naturally reflects the express, implicit and subliminal purposes of the drafters and the organisations on whose behalf they were composed. In some cases, for example, the set primarily addresses just one form of AI, such as robotics or machine-learning. Documents prepared by corporations, industry associations, and even professional associations and joint associations tended to adopt the perspective of producer roles, with the interests of other stakeholders often relegated to a secondary consideration. For example, the joint-association Future Life Institute perceives the need for "constructive and healthy exchange between AI researchers and policy-makers", but not for any participation by stakeholders (FLI 2017 at 3). As a result, transparency is constrained to a small sub-set of circumstances (at 6), 'responsibility' of 'designers and builders' is limited to those roles being mere 'stakeholders in moral implications' (at 9), alignment with human values is seen as being necessary only in respect of "highly autonomous AI systems" (at 10), and "strict safety and control measures" are limited to a small sub-set of AI systems (at 22).
The authors of ITIC (2017) consider that many responsibilities lie elsewhere, and assigns responsibilities to its members only in respect of safety, controllability and data quality. ACM (2017) is expressed in weak language (should be aware of, should encourage, are encouraged) and regards decision opaqueness as being acceptable, while IEEE (2017) suggests a range of important tasks for other parties (standards-setters, regulators, legislatures, courts), and phrases other suggestions in the passive voice, with the result that few obligations are clearly identified as falling on engineering professionals and the organisations that employ them. The House of Lords report might have been expected to adopt a societal or multi-stakeholder approach, yet, as favourably reported in Smith (2018), it appears to have adopted the perspective of the AI industry.
Each of the Principles requires somewhat different application in each phase of the AI supply-chain. An important example of this is the manner in which Principle 7 - Deliver Transparency and Auditability - is intended to be interpreted. In the Research and Invention phases of the technological life-cycle, compliance with Principle 7 requires understanding by inventors and innovators of the AI technology, and explicability to developers and users of AI-based artefacts and systems. During the Innovation and Dissemination phases, the need is for understandability and manageability by developers and users of AI-based systems and applications, and explicability to affected stakeholders. In the Application phase, the emphasis shifts to understandability by affected stakeholders of inferences, decisions and actions arising from at least the AI elements within AI-based systems and applications.
The status of the proposed principles is important to appreciate. They are not expressions of law - although in some jurisdictions, and in some circumstances, some may be legal requirements. They are expressions of moral obligations; but no authority exists that can impose such obligations. In addition, all are contestable, and in different circumstances any of them may be in conflict with other legal or moral obligations, and with various interests of various stakeholders. They represent guidance to organisations involved in AI as to the expectations of courts, regulatory agencies, oversight agencies, competitors and stakeholders. They are intended to be taken into account as organisations undertake risk assessment and risk management, as outlined in s.6 above.
The following Principles apply to each entity responsible for each phase of AI research, invention, innovation, dissemination and application.
AI offers prospects of considerable benefits and disbenefits. All entities involved in creating and applying AI have legal and moral obligations to assess the short-term impacts and longer-term implications, to demonstrate the achievability of the postulated benefits, to be proactive in relation to disbenefits, and to involve stakeholders in the process.
Considerable public disquiet exists in relation to the replacement of human decision-making with inhumane decision-making by AI-based artefacts and systems, and displacement of human workers by AI-based artefacts and systems.
Considerable public disquiet exists in relation to the prospect of humans being subject to obscure AI-based processes, and ceding power to AI-based artefacts and systems.
All entities involved in creating and applying AI have legal and moral obligations to provide safeguards for all human stakeholders, whether as users of AI-based artefacts and systems, or as usees affected by them, and to contribute to human stakeholders' wellbeing.
All entities involved in creating and applying AI have legal and moral obligations to avoid, prevent and mitigate negative impacts on, and to promote the interests of, individuals.
All entities have legal and moral obligations in relation to due process and procedural fairness. These obligations can only be fulfilled if all entities involved in creating and applying AI ensure that humanly-understandable explanations are available to the people affected by AI-based inferences, decisions and actions.
All entities involved in creating and applying AI have legal and moral obligations in relation to the quality of business processes, products and outcomes.
All entities involved in creating and applying AI have legal and moral obligations to ensure resistance to malfunctions (robustness) and recoverability when malfunctions occur (resilience), commensurate with the significance of the benefits, the data's sensitivity, and the potential for harm.
All entities involved in creating and applying AI have legal and moral obligations in relation to due process and procedural fairness. These obligations can only be fulfilled if each entity is discoverable, and each entity addresses problems as they arise.
All entities involved in creating and applying AI have legal and moral obligations in relation to due process and procedural fairness. These obligations can only be fulfilled if the entity implements problem-handling processes, and respects and complies with external problem-handling processes.
The Principles in Table 4 are intentionally framed and phrased in an abstract manner, in an endeavour to achieve applicability to at least the currently mainstream forms of AI discussed earlier - robotics, particularly remote-controlled and self-driving vehicles; cyborgs who incorporate computational capabilities; and AI/ML/neural-networking applications. More broadly, the intention is that they be applicable to what I proposed above as the appropriate conceptualisation of the field - Intellectics.
These Principles are capable of being further articulated into much more specific guidance in respect of each particular category of AI. For example, in a companion project, I have proposed 'Guidelines for Responsible Data Analytics' (Clarke 2018b). These provide more detailed guidance for the conduct of all forms of data analytics projects, including those that apply AI/ML/neural-networking approaches. Areas addressed by the Data Analytics guidelines include governance, expertise and compliance considerations, multiple aspects of data acquisition and data quality, the suitability of both the data and the analytical techniques applied to it, and factors involved in the use of inferences drawn from the analysis.
This paper has proposed that the unserviceable notion of AI should be replaced by the notion of 'complementary intelligence', and that the notion of robotics ('machines that think') is now much less useful than that of 'intellectics' ('computers that do').
The techniques and technologies that emerge from research laboratories offer potential but harbour considerable threats to organisations, and to those organisations' stakeholders. Sources of guidance have been sought, whereby organisations in both the private and public sectors can evaluate the appropriateness of various such technologies to their own operations. Neither ethical analysis nor regulatory schemes deliver what organisations need. The paper concludes that adapted forms of risk assessment and risk management processes can fill the void, and that principles specific to AI can be formulated.
The propositions in this paper need to be workshopped with colleagues in the academic and consultancy worlds. The abstract Principles need to be articulated into more specific expressions that are directly relevant to particular categories of technology, artefacts, systems and applications. The resulting guidance then needs to be exposed to relevant professional executives and managers, reviewed by internal auditors, government relations executives and corporate counsel, and pilot-tested in realistic settings.
Version of 20 February 2019
See here for a PDF version of this Appendix, without cross-references,
The following Principles apply to each entity responsible for each phase of AI research, invention, innovation, dissemination and application. The Principles were derived by consolidating elements from two dozen international sources.
The cross-references following each Principle are to the 'Ethical Principles and IT' sources (Clarke 2018b - E) and 'Principles for AI' sources (Clarke 2018c - P).
1.1 Conceive and design only after ensuring adequate understanding of
purposes and contexts
(E4.3, P5.3, P6.21, P7.1, P15.7, P17.5)
1.2 Justify objectives
1.3 Demonstrate the achievability of postulated benefits
(Not found in any of the documents, but a logical pre-requisite)
1.4 Conduct impact assessment, including risk assessment from all stakeholders'
(E7.1, P3.12, P4.1, P4.2, P6.21, P11.8, P17.5, P18.P0)
1.5 Publish sufficient information to stakeholders to enable them to conduct
(E7.3, P3.7, P4.1, P8.3, P8.4, P8.7)
1.6 Conduct consultation with stakeholders and enable their participation in
(E5.2, E7.2, E8.3, P3.7, P8.6, P8.7, P11.8)
1.7 Reflect stakeholders' justified concerns in the design
(E5.2, E8.3, P3.7, P11.8)
1.8 Justify negative impacts on individuals ('proportionality')
(E3.21, E7.4, E7.5)
1.9 Consider alternative, less harmful ways of achieving the same
2.1 Design as an aid, for augmentation, collaboration and inter-operability
(P4.5, P5.1, P9.1, P9.8 .P14.2, P14.4)
2.2 Avoid design for replacement of people by independent artefacts or systems,
except in circumstances in which those artefacts or systems are demonstrably
more capable than people, and even then ensuring that the result is
complementary to human capabilities
3.1 Ensure human control over AI-based technology, artefacts and systems
(E4.2, E6.1, E6.8, E6.19, P1.4, P2.1, P4.2, P5.2, P6.16, P8.4, P9.3, P12.1, P13.5, P15.4)
3.2 In particular, ensure human control over autonomous behaviour of AI-based
technology, artefacts and systems
(E8.1, P8.4, P10.2, P11.4, P17.12, P18.P4)
3.3 Respect people's expectations in relation to personal data protections
(E5.6, P18.P7), including:
* their awareness of data-usage (E3.6)
* their consent (E3.7, E3.28, E5.3, P3.11, P4.6)
* data minimisation (E3.9)
* public visibility and design consultation and participation (E3.10, E7.2), and
* the relationship between data-usage and the data's original purpose (E3.27)
3.4 Respect each person's autonomy, freedom of choice and right to
(E2.1, E5.3, P3.3, P9.7, P11.3, P18.E3, P18.P6)
3.5 Ensure human review of inferences and decisions prior to action being taken
3.6 Avoid deception of humans
(E4.4, E6.20, P2.5, P18.E4)
3.7 Avoid services being conditional on the acceptance of AI-based artefacts
4.1 Ensure people's physical health and safety ('nonmaleficence')
(E2.2, E3.1, E4.1, E4.3, E5.4, E6.2, E6.9, E6.13, E6.14, E6.18, P1.2, P1.3, P2.1, P3.2, P3.6, P3.9, P3.12, P4.3, P4.9, P6.6, P8.4, P9.4, P10.2, P11.4, P13.5, P14.1, P15.3, P17.8, P18.E2)
4.2 Ensure people's psychological safety (E3.1, E6.9, E6.13, P18.E2), by avoiding negative effects on their mental health, emotional state, inclusion in society, worth, and standing in comparison with other people (E5.4, E6.3)
4.3 Contribute to people's wellbeing ('beneficence')
(E2.3, E3.20, E5.5, P3.1, P3.4, P6.1, P6.14, P6.15, P8.6, P11.6, P12.2, P13.1, P15.1, P16.4, P18.E1)
4.4 Implement safeguards to avoid, prevent and mitigate negative impacts and
(E3.24, E7.6, P6.21, P10.4)
4.5 Avoid violation of trust
4.6 Avoid the manipulation of vulnerable people (E4.4, P4.5, P4.9), e.g. by taking advantage of individuals' tendencies to addictions such as gambling (E6.3, P18.P3), and to letting pleasure overrule rationality
5.1 Be just / fair / impartial, treat individuals equally (E2.4, E2.5, E3.2, E3.16, E3.29, P3.4, P18.E4, P18.P3), and avoid unfair discrimination and bias, not only where they are illegal, but also where they are materially inconsistent with public expectations (ICCPR Arts. 2.1, 3, 26 and 27, E3.16, P3.4, P4.5, P11.5, P15.2, P16.1, P17.4, P18.P5)
5.2 Ensure compliance with human rights laws
(E4.2, P3.5, P3.9, P4.3, P18.E0)
5.3 Avoid restrictions on, and promote, people's freedom of movement
(ICCPR 12, P6.13)
5.4 Avoid interference with, and promote privacy, family, home or reputation
(ICCPR 17, E5.6, P3.11, P6.12, P8.4, P9.6, P13.3, P15.5)
5.5 Avoid interference with, and promote, the rights of freedom of information, opinion and expression (ICCPR 19, P4.6), of freedom of assembly (ICCPR 21, P6.13), of freedom of association (ICCPR 22, P6.13), of freedom to participate in public affairs, and of freedom to access public services (ICCPR 25, P6.13)
5.6 Where interference with human values or human rights is outweighed by other
factors, ensure that the interference is no greater than is justified ('harm
(E3.9, E7.6, P1.3, P3.12)
6.1 Ensure that the fact that a process is AI-based is transparent to all
(E4.4, P4.8, P18.P6)
6.2 Ensure that data provenance, and the means whereby inferences are drawn
from it, decisions are made, and actions are taken, are logged and can be
(E6.6, P2.4, P4.8, P5.2, P6.7, P7.4, P7.6, P8.2, P9.2, P11.1, P11.2, P13.2, P16.2, P16.3, P17.7, P18.P10)
6.3 Ensure that people are aware of inferences, decisions and actions that
affect them, and have access to humanly-understandable explanations of how they
(E3.12, P2.4, P17.1, P18.P6)
7.1 Ensure effective, efficient and adaptive performance of intended
(E6.2, E6.11, P1.6, P4.2, P15.6)
7.2 Ensure data quality and data relevance
(P10.3, P11.2, P17.7, P18.P2)
7.3 Justify the use of data, commensurate with each data-item's sensitivity
(E3.26, E7.4, P18.P7)
7.4 Ensure security safeguards against inappropriate data access, modification
and deletion, commensurate with its sensitivity
(E3.15, P17.9, P18.P8)
7.5 Deal fairly with people (faithfulness, fidelity)
7.6 Ensure that inferences are not drawn from data using invalid or unvalidated
7.7 Test result validity, and address the problems that are detected
(E3.5, P7.7. P9.2, P17.6, P18.P8, P18.P10)
7.8 Impose controls in order to ensure that the safeguards are in place and
(E7.7, P10.2, P18.P8)
7.9 Conduct audits of safeguards and controls
(E7.8, P9.3, P18.P1)
8.1 Deliver and sustain appropriate security safeguards against the risk of
compromise of intended functions arising from both passive threats and active
attacks, commensurate with the significance of the benefits and the potential
to cause harm
(E4.3, E6.11, P1.4, P1.5, P4.9, P6.6, P8.4, P9.5, P18.P8)
8.2 Deliver and sustain appropriate security safeguards against the risk of
inappropriate data access, modification and deletion, arising from both passive
threats and active attacks, commensurate with the data's sensitivity
(E3.15, E6.5, E6.10, P3.11, P9.6, P17.9, P18.P8)
8.3 Conduct audits of the justification, the proportionality, the transparency,
and the harm avoidance, prevention and mitigation measures and controls
8.4 Ensure resilience, in the sense of prompt and effective recovery from
9.1 Ensure that the responsible entity is apparent or can be readily
discovered by any party
(E4.5, E6.4, P2.3, P3.8, P4.7, P8.5, P12.3, P17.3, P18.P1)
9.2 Ensure that effective remedies exist, in the form of complaints processes,
appeals processes, and redress where harmful errors have occurred
(ICCPR 2.3, E3.13, E3.14, E7.7, P3.11, P4.7, P7.2, P8.7, P9.9, P10.5, P11.9, P16.3, P17.5, P18.P1)
10. Enforce, and Accept Enforcement of, Liabilities and Sanctions
10.1 Ensure that complaints, appeals and redress processes operate effectively
(ICCPR 2.3, E7.7)
10.2 Comply with external complaints, appeals and redress processes and outcomes (ICCPR 14), including, in particular, provision of timely, accurate and complete information relevant to cases
The following elements within the sources have not been reflected in Table 4 and Appendix 1.
ACM (2017) 'Statement on Algorithmic Transparency and Accountability' Association for Computing Machinery, January 2017, at https://www.acm.org/binaries/content/assets/public-policy/2017_usacm_statement_algorithms.pdf
Akiko (2012) 'South Korean Robot Ethics Charter 2012' Akiko's Blog, 2012, at https://akikok012um1.wordpress.com/south-korean-robot-ethics-charter-2012/
Albus J. S. (1991) 'Outline for a theory of intelligence' IEEE Trans. Systems, Man and Cybernetics 21, 3 (1991) 473-509, at http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.410.9719&rep=rep1&type=pdf
Anderson C. (2008) 'The End of Theory: The Data Deluge Makes the Scientific Method Obsolete' Wired Magazine 16:07, 23 June 2008, at http://archive.wired.com/science/discoveries/magazine/16-07/pb_theory
APF (2013) 'Meta-Principles for Privacy Protection' Australian Privacy Foundation, March 2013, at https://privacy.org.au/policies/meta-principles/
Baumer E.P.S. (2015) 'Usees' Proc. 33rd Annual ACM Conference on Human Factors in Computing Systems (CHI'15), April 2015
Becker H. & Vanclay F. (2003) 'The International Handbook of Social Impact Assessment' Cheltenham: Edward Elgar, 2003
Bennett Moses L. (2011) 'Agents of Change: How the Law Copes with Technological Change' Griffith Law Review 20, 4 (2011) 764-794, at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2000428
Bibel W. (1980) ''Intellektik' statt 'KI' -- Ein ernstgemeinter Vorschlag' Rundbrief der Fachgruppe Kuenstliche Intelligenz in der Gesellschaft fuer Informatik, 22, 15-16 December 1980
Bibel W. (1989) 'The Technological Change of Reality: Opportunities and Dangers' AI & Society 3, 2 (April 1989) 117-132
Braithwaite B. & Drahos P. (2000) `Global Business Regulation' Cambridge University Press, 2000
BS (2016) 'Robots and robotic devices - Guide to the ethical design and application of robots and robotic systems' BS 8611, British Standards Institute, April 2016
Burrell J. (2016) How the machine 'thinks': Understanding opacity in machine learning algorithms' Big Data & Society 3, 1 (January-June 2016) 1-12
Calo R. (2017) 'Artificial Intelligence Policy: A Primer and Roadmap' UC Davis L. Rev. 51 (2017) 399-404
Chen Y. & Cheung A.S.Y. (2017). 'The Transparent Self Under Big Data Profiling: Privacy and Chinese Legislation on the Social Credit System, The Journal of Comparative Law 12, 2 (June 2017) 356-378, at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2992537
CLA-EP (2016) 'Recommendations on Civil Law Rules on Robotics' Committee on Legal Affairs of the European Parliament, 31 May 2016, at http://www.europarl.europa.eu/sides/getDoc.do?pubRef=-//EP//NONSGML%2BCOMPARL%2BPE-582.443%2B01%2BDOC%2BPDF%2BV0//EN
Clarke R. (1989) 'Knowledge-Based Expert Systems: Risk Factors and Potentially Profitable Application Area', Xamax Consultancy Pty Ltd, January 1989, at http://www.rogerclarke.com/SOS/KBTE.html
Clarke R. (1991) 'A Contingency Approach to the Application Software Generations' Database 22, 3 (Summer 1991) 23-34, PrePrint at http://www.rogerclarke.com/SOS/SwareGenns.html
Clarke R. (1992) 'Extra-Organisational Systems: A Challenge to the Software Engineering Paradigm' Proc. IFIP World Congress, Madrid, September 1992, at http://www.rogerclarke.com/SOS/PaperExtraOrgSys.html
Clarke R. (1993) 'Asimov's Laws of Robotics: Implications for Information Technology' in two parts, in IEEE Computer 26,12 (December 1993) 53-61, and 27,1 (January 1994) 57-66, at http://www.rogerclarke.com/SOS/Asimov.html
Clarke R. (2005) 'Human-Artefact Hybridisation: Forms and Consequences' Proc. Ars Electronica 2005 Symposium on Hybrid - Living in Paradox, Linz, Austria, 2-3 September 2005, PrePrint at http://www.rogerclarke.com/SOS/HAH0505.html
Clarke R. (2009) 'Privacy Impact Assessment: Its Origins and Development' Computer Law & Security Review 25, 2 (April 2009) 123-135, PrePrint at http://www.rogerclarke.com/DV/PIAHist-08.html
Clarke R. (2011) 'Cyborg Rights' IEEE Technology and Society 30, 3 (Fall 2011) 49-57, at http://www.rogerclarke.com/SOS/CyRts-1102.html
Clarke R. (2014a) 'Understanding the Drone Epidemic' Computer Law & Security Review 30, 3 (June 2014) 230-246, PrePrint at http://www.rogerclarke.com/SOS/Drones-E.html
Clarke R. (2014b) 'What Drones Inherit from Their Ancestors' Computer Law & Security Review 30, 3 (June 2014) 247-262, PrePrint at http://www.rogerclarke.com/SOS/Drones-I.html
Clarke R. (2014c) 'The Regulation of of the Impact of Civilian Drones on Behavioural Privacy' Computer Law & Security Review 30, 3 (June 2014) 286-305, PrePrint at http://www.rogerclarke.com/SOS/Drones-BP.html
Clarke R. (2015) 'The Prospects of Easier Security for SMEs and Consumers' Computer Law & Security Review 31, 4 (August 2015) 538-552, PrePrint at http://www.rogerclarke.com/EC/SSACS.html
Clarke R. (2016a) 'Big Data, Big Risks' Information Systems Journal 26, 1 (January 2016) 77-90, PrePrint at http://www.rogerclarke.com/EC/BDBR.html
Clarke R. (2016) 'Appropriate Regulatory Responses to the Drone Epidemic' Computer Law & Security Review 32, 1 (Jan-Feb 2016) 152-155, PrePrint at http://www.rogerclarke.com/SOS/Drones-PAR.html
Clarke R. (2016c) 'Quality Assurance for Security Applications of Big Data' Proc. EISIC'16, Uppsala, 17-19 August 2016, PrePrint at http://www.rogerclarke.com/EC/BDQAS.html
Clarke R. (2018a) 'Centrelink's Big Data 'Robo-Debt' Fiasco of 2016-17' Xamax Consultancy Pty Ltd, January 2018, at http://www.rogerclarke.com/DV/CRD17.html
Clarke R. (2018b) 'Guidelines for the Responsible Application of Data Analytics' Computer Law & Security Review 34, 3 (May-Jun 2018) 467- 476, PrePrint at http://www.rogerclarke.com/EC/GDA.html
Clarke R. (2018c) 'Ethical Principles and Information Technology' Xamax Consultancy Pty Ltd, rev. September 2018, at http://www.rogerclarke.com/EC/GAIE.html
Clarke R. (2018d) 'Principles for AI: A 2017-18 SourceBook' Xamax Consultancy Pty Ltd, rev. September 2018, at http://www.rogerclarke.com/EC/GAI.html
Clarke R. & Bennett Moses L. (2014) 'The Regulation of Civilian Drones' Impacts on Public Safety' Computer Law & Security Review 30, 3 (June 2014) 263-285, PrePrint at http://www.rogerclarke.com/SOS/Drones-PS.html
Clarkson M.B.E. (1995) 'A Stakeholder Framework for Analyzing and Evaluating Corporate Social Performance' The Academy of Management Review 20, 1 (Jan.1995) 92-117 , at https://www.researchgate.net/profile/Mei_Peng_Low/post/Whats_corporate_social_performance_related_to_CSR/attachment/59d6567879197b80779ad3f2/AS%3A530408064417792%401503470545971/download/A_Stakeholder_Framework_for_Analyzing+CSP.pdf
Devlin H. (2016). 'Do no harm, don't discriminate: official guidance issued on robot ethics' The Guardian, 18 Sep 2016, at https://www.theguardian.com/technology/2016/sep/18/official-guidance-robot-ethics-british-standards-institute
DMV-CA (2018) 'Autonomous Vehicles in California' Califiornian Department of Motor Vehicles, February 2018, at https://www.dmv.ca.gov/portal/dmv/detail/vr/autonomous/bkgd
EC (2018) 'Statement on Artificial Intelligence, Robotics and `Autonomous' Systems' European Group on Ethics in Science and New Technologies' European Commission, March 2018, at http://ec.europa.eu/research/ege/pdf/ege_ai_statement_2018.pdf
EDPS (2016) 'Artificial Intelligence, Robotics, Privacy and Data Protection' European Data Protection Supervisor, October 2016, at https://edps.europa.eu/sites/edp/files/publication/16-10-19_marrakesh_ai_paper_en.pdf
ENISA (2016) 'Risk Management:Implementation principles and Inventories for Risk Management/Risk Assessment methods and tools' European Union Agency for Network and Information Security, June 2016, at https://www.enisa.europa.eu/publications/risk-management-principles-and-inventories-for-risk-management-risk-assessment-methods-and-tools
Fieser J. (1995) 'Ethics' Internet Encyclopaedia of Philosophy, 1995, at https://www.iep.utm.edu/ethics/
Firesmith D. (2004) 'Specifying Reusable Security Requirements' Journal of Object Technology 3, 1 (Jan-Feb 2004) 61-75, at http://www.jot.fm/issues/issue_2004_01/column6
Fischer-Huebner S. & Lindskog H. (2001) 'Teaching Privacy-Enhancing Technologies' Proc. IFIP WG 11.8 2nd World Conference on Information Security Education, Perth, 2001, at http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.24.3950&rep=rep1&type=pdf
FLI (2017) 'Asilomar AI Principles' Future of Life Institute, January 2017, at https://futureoflife.org/ai-principles/?cn-reloaded=1
Floridi L. (2018) 'Soft Ethics: Its Application to the General Data Protection Regulation and Its Dual Advantage' Philosophy & Technology 31, 2 (June 2018) 163-167, at https://link.springer.com/article/10.1007/s13347-018-0315-5
Freeman R.E. & Reed D.L. (1983) 'Stockholders and Stakeholders: A New Perspective on Corporate Governance' California Management Review 25:, 3 (1983) 88-106, at https://www.researchgate.net/profile/R_Freeman/publication/238325277_Stockholders_and_Stakeholders_A_New_Perspective_on_Corporate_Governance/links/5893a4b2a6fdcc45530c2ee7/Stockholders-and-Stakeholders-A-New-Perspective-on-Corporate-Governance.pdf
GDPR (2018) 'General Data Protection Regulation' Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the Protection of Natural Persons with Regard to the Processing of Personal Data and on the Free Movement of Such Data, at http://www.privacy-regulation.eu/en/index.htm
GEFA (2016) 'Position on Robotics and AI' The Greens / European Free Alliance Digital Working Group, November 2016, at https://juliareda.eu/wp-content/uploads/2017/02/Green-Digital-Working-Group-Position-on-Robotics-and-Artificial-Intelligence-2016-11-22.pdf
HOL (2018) 'AI in the UK: ready, willing and able?' Select Committee on Artificial Intelligence, House of Lords, April 2018, at https://publications.parliament.uk/pa/ld201719/ldselect/ldai/100/100.pdf
Holder C., Khurana V., Harrison F. & Jacobs L. (2016) 'Robotics and law: Key legal and regulatory implications of the robotics age (Part I of II)' Computer Law & Security Review 32, 3 (May-Jun 2016) 383-402
HTR (2017) 'Robots: no regulatory race against the machine yet' The Regulatory Institute, April 2017, at http://www.howtoregulate.org/robots-regulators-active/#more-230
HTR (2018a) 'Report on Artificial Intelligence: Part I - the existing regulatory landscape' The Regulatory Institute, May 2018, at http://www.howtoregulate.org/artificial_intelligence/
HTR (2018b) 'Report on Artificial Intelligence: Part II - outline of future regulation of AI' The Regulatory Institute, June 2018, at http://www.howtoregulate.org/aipart2/#more-327
HTR (2018c) 'Research and Technology Risks: Part IV - A Prototype Regulation' The Regulatory Institute, March 2018, at http://www.howtoregulate.org/prototype-regulation-research-technology/#more-298
ICO (2017) 'Big data, artificial intelligence, machine learning and data protection' UK Information Commissioner's Office, Discussion Paper v.2.2, September 2017, at https://ico.org.uk/for-organisations/guide-to-data-protection/big-data/
IEEE (2017) 'Ethically Aligned Design: A Vision for Prioritizing Human Well-being with Autonomous and Intelligent Systems (A/IS)' IEEE, Version 2, December 2017, at http://standards.ieee.org/develop/indconn/ec/autonomous_systems.html
ISM (2017) 'Information Security Manual' Australian Signals Directorate, November 2017, at https://acsc.gov.au/infosec/ism/index.htm
ISO (2005) 'Information Technology - Code of practice for information security management', International Standards Organisation, ISO/IEC 27002:2005
ISO (2008) 'Information Technology - Security Techniques - Information Security Risk Management' ISO/IEC 27005:2008
ITIC (2017) 'AI Policy Principles' Information Technology Industry Council, undated but apparently of October 2017, at https://www.itic.org/resources/AI-Policy-Principles-FullReport2.pdf
Knight W. (2017) 'The Dark Secret at the Heart of AI' 11 April 2017, MIT Technology Review https://www.technologyreview.com/s/604087/the-dark-secret-at-the-heart-of-ai/
Leenes R. & Lucivero F. (2014) 'Laws on Robots, Laws by Robots, Laws in Robots: Regulating Robot Behaviour by Design' Law, Innovation and Technology 6, 2 (2014) 193-220
Lessig L. (1999) 'Code and Other Laws of Cyberspace' Basic Books, 1999
McCarthy J. (2007) 'What is artificial intelligence?' Department of Computer Science, Stanford University, November 2007, at http://www-formal.stanford.edu/jmc/whatisai/node1.html
McCarthy J., Minsky M.L., Rochester N. & Shannon C.E. (1955) 'A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence' Reprinted in AI Magazine 27, 4 (2006), at https://www.aaai.org/ojs/index.php/aimagazine/article/viewFile/1904/1802
Manwaring K. & Clarke R. (2015) 'Surfing the third wave of computing: a framework for research into eObjects' Computer Law & Security Review 31,5 (October 2015) 586-603, PrePrint at http://www.rogerclarke.com/II/SSRN-id2613198.pdf
Martins L.E.G. & Gorschek T. (2016) 'Requirements engineering for safety-critical systems: A systematic literature review' Information and Software Technology Journal 75 (2016) 71-89
Maschmedt A. & Searle R. (2018) 'Driverless vehicle trial legislation âÄ" state-by-state' King & Wood Malleson, February 2018, at https://www.kwm.com/en/au/knowledge/insights/driverless-vehicle-trial-legislation-nsw-vic-sa-20180227
Mayer-Schonberger V. & Cukier K. (2013) 'Big Data: A Revolution That Will Transform How We Live, Work and Think' John Murray, 2013
MS (2018) 'Microsoft AI principles' Microsoft, August 2018, at https://www.microsoft.com/en-us/ai/our-approach-to-ai
Newcomer E. (2018). 'What Google's AI Principles Left Out: We're in a golden age for hollow corporate statements sold as high-minded ethical treatises' Bloomberg, 8 June 2018, at https://www.bloomberg.com/news/articles/2018-06-08/what-google-s-ai-principles-left-out
NIST (2012) 'Guide for Conducting Risk Assessments' National Institute of Standards and Technology, Special Publication SP 800-30 Rev. 1, September 2012, at http://csrc.nist.gov/publications/nistpubs/800-30-rev1/sp800_30_r1.pdf
OTA (1977) 'Technology Assessment in Business and Government' Office of Technology Assessment, NTIS order #PB-273164', January 1977, at http://www.princeton.edu/~ota/disk3/1977/7711_n.html
Pagallo U. (2016). 'Even Angels Need the Rules: AI, Roboethics, and the Law' Proc. ECAI 2016
Palmerini E. et al. (2014). 'Guidelines on Regulating Robotics Delivery' EU Robolaw Project, September 2014, at http://www.robolaw.eu/RoboLaw_files/documents/robolaw_d6.2_guidelinesregulatingrobotics_20140922.pdf
Pichai S. (2018) 'AI at Google: our principles' Google Blog, 7 Jun 2018, at https://www.blog.google/technology/ai/ai-principles/
PoAI (2018) 'Our Work (Thematic Pillars)' Partnership on AI, April 2018, at https://www.partnershiponai.org/about/#pillar-1
Pouloudi A. & Whitley E.A. (1997) 'Stakeholder Identification in Inter-Organizational Systems: Gaining Insights for Drug Use Management Systems' European Journal of Information Systems 6, 1 (1997) 1-14, at http://eprints.lse.ac.uk/27187/1/__lse.ac.uk_storage_LIBRARY_Secondary_libfile_shared_repository_Content_Whitley_Stakeholder%20identification_Whitley_Stakeholder%20identification_2015.pdf
Rayome A.D. (2017) 'Guiding principles for ethical AI, from IBM CEO Ginni Rometty', TechRepublic (17 January 2017), at https://www.techrepublic.com/article/3-guiding-principles-for- ethical-ai-from-ibm-ceo-ginni-rometty/
Russell S.J. & Norvig P. (2009) 'Artificial Intelligence: A Modern Approach' Prentice Hall, 3rd edition, 2009
Schellekens M. (2015) 'Self-driving cars and the chilling effect of liability law' Computer Law & Security Review 31, 4 (Jul-Aug 2015) 506-517
Scherer M.U. (2016) 'Regulating Artificial Intelligence Systems: Risks, Challenges, Competencies, and Strategies' Harvard Journal of Law & Technology 29, 2 (Spring 2016) 353-400, at http://euro.ecom.cmu.edu/program/law/08-732/AI/Scherer.pdf
Selbst A.D. & Powles J. (2017) 'Meaningful information and the right to explanation' International Data Privacy Law 7, 4 (November 2017) 233-242, at https://academic.oup.com/idpl/article/7/4/233/4762325
Smith R. (2018). '5 core principles to keep AI ethical'. World Economic Forum, 19 Apr 2018, at https://www.weforum.org/agenda/2018/04/keep-calm-and-make-ai-ethical/
TvH (2006) 'Telstra Corporation Limited v Hornsby Shire Council' NSWLEC 133 (24 March 2006), esp. paras. 113-183, at http://www.austlii.edu.au/au/cases/nsw/NSWLEC/2006/133.htm
UGU (2017) 'Top 10 Principles for Ethical AI' UNI Global Union, December 2017, at http://www.thefutureworldofwork.org/media/35420/uni_ethical_ai.pdf
Vellinga N.E. (2017) 'From the testing to the deployment of self-driving cars: Legal challenges to policymakers on the road ahead' Computer Law & Security Review 33, 6 (Nov-Dec 2017) 847-863
Villani C. (2017) 'For a Meaningful Artificial Intelligence: Towards a French and European Strategy' Part 5 - What are the Ethics of AI?, Mission for the French Prime Minister , March 2018, pp.113-130, at https://www.aiforhumanity.fr/pdfs/MissionVillani_Report_ENG-VF.pdf
Villaronga E.F. & Roig A. (2017) 'European regulatory framework for person carrier robots' Computer Law & Security Review 33, 4 (Jul-Aug 2017) 502-220
Wachter S. & Mittelstadt B. (2019) 'A Right to Reasonable Inferences: Re-Thinking Data Protection Law in the Age of Big Data and AI' Forthcoming, Colum. Bus. L. Rev. (2019), at https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3248829
Wachter S., Mittelstadt B. & Floridi L. (2017) 'Why a Right to Explanation of Automated Decision-Making Does Not Exist in the General Data Protection Regulation' International Data Privacy Law 7, 2 (May 2017) 76-99, at https://academic.oup.com/idpl/article/7/2/76/3860948
Warwick K. (2014) 'The Cyborg Revolution' Nanoethics 8, 3 (Oct 2014) 263-273
WH (2018) 'Summary of the 2018 White House Summit on Artificial Intelligence for American Industry' Office of Science and Technology Policy, White House, May 2018, at https://www.whitehouse.gov/wp-content/uploads/2018/05/Summary-Report-of-White-House-AI-Summit.pdf
Wright D. & De Hert P. (eds) (2012) 'Privacy Impact Assessments' Springer, 2012
Wyndham J. (1932) 'The Lost Machine' (originally published in 1932), reprinted in A. Wells (Ed.) 'The Best of John Wyndham' Sphere Books, London, 1973, pp. 13- 36, and in Asimov I., Warrick P.S. & Greenberg M.H. (Eds.) 'Machines That Think' Holt, Rinehart, and Wilson, 1983, pp. 29-49
Zhaohui W. et al. (2016) 'Cyborg Intelligence: Recent Progress and Future Directions' IEEE Intelligent Systems 31, 6 (Nov-Dec 2016) 44-50
This paper has benefited from feedback from multiple colleagues, and particularly Peter Leonard of Data Synergies and Prof. Graham Greenleaf and Kayleen Manwaring of UNSW. I first applied the term 'intellectics' during a presentation to launch a Special Issue of the UNSW Law Journal in Sydney in November 2017.
Roger Clarke is Principal of Xamax Consultancy Pty Ltd, Canberra. He is also a Visiting Professor in Cyberspace Law & Policy at the University of N.S.W., and a Visiting Professor in the Research School of Computer Science at the Australian National University. He has also spent many years on the Board of the Australian Privacy Foundation, and is Company Secretary of the Internet Society of Australia.
The content and infrastructure for these community service pages are provided by Roger Clarke through his consultancy company, Xamax.
From the site's beginnings in August 1994 until February 2009, the infrastructure was provided by the Australian National University. During that time, the site accumulated close to 30 million hits. It passed 65 million in early 2021.
Sponsored by the Gallery, Bunhybee Grasslands, the extended Clarke Family, Knights of the Spatchcock and their drummer
Xamax Consultancy Pty Ltd
ACN: 002 360 456
78 Sidaway St, Chapman ACT 2611 AUSTRALIA
Tel: +61 2 6288 6916
Created: 11 July 2018 - Last Amended: 20 February 2019 by Roger Clarke - Site Last Verified: 15 February 2009
This document is at www.rogerclarke.com/EC/GAIF.html