After a successful 2016 first edition, our next summer school cohort on The Regulation of Robotics in Europe: Legal, Ethical and Economic Implications will take place in Pisa at the Scuola Sant’Anna, from 3- 8 July.

When the Robolaw project came to an end – and we presented our results before the European Parliament – we clearly perceived that a leap was needed not only in some approaches to regulation but also in the way social scientists, as well as engineers, are trained.

Indeed, in order to undergo technical analysis in law and robotics, without being lured into science fiction, an adequate understanding of the peculiarities of the systems being studied is required. A bottom-up approach, like the one adopted by Robolaw and its guidelines, is essential.

Social scientists, and lawyers in particular, often lack such knowledge and thus tend to either make unreasonable assumptions – of technological developments that are farfetched or simply unrealistic – or misperceive what the pivoting point in the analysis is going to be. The notion of autonomy is a fitting example. The consequence, however, is not simply bad scientific literature, but potentially inadequate policies being developed, thence wrong decision – even legislative ones – being adopted, impacting research and development of new applications, while overlooking relevant issues and impairments.

Similarly, engineers working in robotics are often confronted with philosophical and legal debates involving the devices they research that they are not always equipped to understand. Those debates are instead precious for they allow to identify societal concerns and expectations that can be used to orient research strategically, and engineers ought also participate and have a say.

Ultimately, it is everybody’s interest to better address existing and emerging needs, fulfilling desires and avoiding eliciting – often ungrounded – fears. This is what the European Union understands as Responsible Research and Innovation, but it also the prerequisite for the diffusion of new applications in society and the emergence of a sound robotic industry. Moreover, the current tendency in EU regulation favouring by design approaches – whereby privacy or other rights need to be enforced through the very functioning of the device – require technicians to consider such concerns early on, during the development phase of their products.

A common language needs thus be created, to avoid a babel-tower effect, that preserves each one’s peculiarities and specificities, yet allowing close cooperation.

A multidisciplinary approach, grounded in philosophy – ethics in particular –, law – and law and economics methodologies – economics and innovation management, and engineering is required.

With that idea in mind, we competed and won a Jean Monnet grant – a prestigious funding action of the EU Commission, mainly directed towards the promotion of education and teaching activities – with a project titled: Europe Regulates Robotics and organized the first edition of the Summer School The Regulation of Robotics in Europe: Legal, Ethical and Economic Implications in 2016.

The opening event of the Summer School saw the participation of MEP (Member of the EU Parliament) Mady Delvaux-Stehres, who presented what was then the draft recommendation – now approved – of the EU Parliament on the Civil Law Rules of Robotics, Mihalis Kritikos – a Policy Analyst of the Parliament –, who personally contributed to the drafting of that document, Maria Chiara Carrozza – former Italian minister of University Education and Research, professor of robotics and member of the Italian Senate – discussing Italian political initiatives.  We also had entrepreneurs, such as Roberto Valleggi, and engineers coming from the industry, such as Arturo Baroncelli – from Comau – and academia, Fabio Bonsignorio, who also taught the course.

Classes dealt with methodologies – the Robolaw approach – notions of autonomy, liability – and different liability models – privacy, benchmarking and robot design, machine ethics and human enhancement through technology, innovation management and technology assessment. Students also had the chance to visit the Biorobotics Institute laboratories in Pontedera (directed by Prof. Paolo Dario) and see many different applications and research being carried out, directly explained by the people who work on them.

The most impressive part was, however, our class. We managed to put together a truly international group of young and bright minds, many of which already enrolled in a PhD program – in law, philosophy, engineering and management – coming from Universities such as Edinburgh, London School of Economics, Sorbonne, Cambridge, Vienna, Bologna, Suor Orsola, Bicocca, Milan, Hannover, Pisa, Pavia and Freiburg. Other came from prominent European robotic industries, were practitioners, entrepreneurs, policy makers from EU institutions.

At the end of the Summer School, some presented their research on a broad set of extremely interesting topics, such as driverless car liability and innovation management, machine ethics and the trolley problem, anthrobotics and decision-making in natural and artificial systems.

We had vivid in class debates. A true community was created that is still in contact today. Five of our students actively participated in the 2017 European Robotics Forum in Edinburgh and more are working and publishing on such matters.

We can say we truly achieved our goal! However, the challenge has just begun. We want to reach out to more people and replicate this successful initiative. A second edition of the Summer School will take place again this year in Pisa at the Scuola Sant’Anna from July 3^rd to 8^th and we are accepting applications until June 7^th.

I am certain we will manage again to put together an incredible group of individuals, and I can’t wait to meet our new class. On our side, we are preparing a lot of interesting surprises for them, including the participation of policy makers involved in the regulation of robotics at EU level to provide a first-hand look at what is happening in the EU.

More information about the summer school can be found on our website here.

Registration to the summer school can be found here.

As human-robot interactions become commonplace, MEPs stress that EU-wide rules are needed to guarantee a standard level of safety and security.© AP Images/European Union – EP

Robots are the technology of the future. But the current legal system is incapable of handling them. This generic statement is often the premise for considerations about the possibility of awarding rights (and liabilities) to these machines at some, less-than clearly identified, point in time. Discussing the adequacy of existing regulation in accommodating new technologies is certainly necessary, but the ontological approach is incorrect. Instead a functional approach needs to be adopted, identifying:

1. what rules can be applied to robots (as is);
2. what incentives do such rules provide;
3. whether those incentives are desirable.

The recent Resolution of the European Parliament (henceforth Resolution) has great political relevance and strategic importance in the development of a European Robotic Industry. Its considerations and conclusions will be taken into account in the current position paper.

Issues

The first issue when discussing regulation is that of definitions, for one cannot regulate something without firstly defining it. However, the term robot is technical and encompasses a wide range of applications that have very little in common. For this very reason, it is impossible to develop a unitary body of rules applicable to all kinds of robotic applications, rather different rules should apply to different classes of devices.

The major issue when discussing civil law rules on robotics is that of liability (for damages). Automation might, to some extent, challenge some of the existing paradigms; and increasing human-machine cooperation might cause different sets of existing rules to overlap, leading to uncertainty, thence increased litigation and difficulties in insuring new products.

Connected to the above is robot testing. A clear legal framework for robot testing outside the restricted environment of the laboratory is needed to assess the kind of dangers that might emerge with the use and their statistical frequency (also for insurance purposes). Similarly, standardization and the development of adequate, narrow-tailored technical standards for different kinds of robots is a major concern, both to ensure product safety and the adoption of possible alternatives to existing liability rules.

A possible non-issue when discussing rules for robotics is that of the attribution of personhood. This, if intended in an ontological way, is deprived of any reasonable grounding in both technical, philosophical and legal considerations. Instead, if understood in a purely functional way the attribution of legal personhood (like in the case for corporations) might be open for discussion (in some cases). Considering some more specific kinds of applications, in particular biorobotic devices and the issue of human enhancement, its regulation and management becomes of the greatest importance and quite likely the single most relevant bioethical issues of the nearby future, requiring ad-hoc regulation to be adopted.

Finally, privacy regulation, access to data and data use is of pivotal importance, not only for the development of a European Robotics industry but more broadly for a digital market. All the mentioned issues might fall under some – direct or indirect – competences of the EU and would certainly benefit from regulations adopted at a supranational (thence European) level.

Responses

The Resolution addresses all the above mentioned issues with consistent considerations, depicting an adequate framework for a technical – legal – debate about what narrow tailored sets of rules should be adopted at the EU level. Overall, it is of the greatest political and strategical importance for defining a modern legal system, favorable to the immersion of new technologies and the proliferation of new businesses.

More specifically:

Definitions: there is need for a definition of “robot” that needs to be inclusive. What needs to be avoided are nominalistic discussions which would inevitably emerge as soon as a regulation was adopted (should the notion of robot be too narrow). Debates about whether a robot requires to be autonomous or not, controlled or not, embodied or not are irrelevant from a legal point of view. Instead, such characteristics should allow to distinguish sub-classes of robots that might be regulated unitarily. Thence, next to a broader and all-encompassing definition of robot (that should include software and non-embodied AI), narrower definitions should be elaborated, pooling together those applications that show some relevant similarities and that can be regulated unitarily.

Liability: Human-machine cooperation will cause different sets of rules to overlap (namely product liability rules and traditional tort law principles). This will cause high levels of uncertainty and litigation, delaying innovation. With respect to compensation, it is, in many cases, sensible to separate the function of ensuring product safety from that of providing the victim with compensation. This might justify different alternative solutions to be adopted: liability exemptions for users and/or manufacturers; creation of automatic compensations funds (privately or publicly funded); compulsory insurance provisions. More broadly, the inadequacies of existing rules (in particular product liability rules) might suggest to radically replace a fault based rule with a risk-management approach (based on absolute liability rules) holding liable the party who is better placed to minimize the cost and acquire insurance (Resolution nn. 53, 55). A one-stop-shop approach might be sensible, preventing complex litigation to apportion liability among different players involved. Which solution is preferable depends on the class of applications considered, the market for such products and the possibility to address those risks through insurance (Resolution nn. 57-59).

Testing: a uniform set of rules allowing testing outside the laboratories and even in human environments should be adopted, defining clear standards (in particular with respect to safety, insurance and management of the experiment) thus reducing discretionary powers of local authorities (Resolution n. 23).

Standardization & European Robotics Agency: standards represent the most effective way to ensure high levels of product safety and provide certainty ex ante to manufacturers who conform to them (Resolution n. 22). However, the time required for the adoption of a new standard and its breadth is incompatible with the current pace of technological innovation. A European Robotic Agency, such as the one suggested by the Resolution (nn. 15-17), could have a strategic importance setting a supranational standard, that could be of use beyond European borders. Otherwise, other leading economies will attempt to do the same.

Electronic Personhood: set forth by the Resolution, this notion is purely functional and intends to facilitate the registration, insurance and management of some devices (in particular nonembodied AI) with a legal tool that is equivalent to that used for corporations (so called legal personhood), see Resolution n. 59, let. E) and F).

Human Enhancement: the use of robotics to overcome human limits might become problematic given the lack of a clear sets or rules and criteria that could help discern what kind of manipulations of the human body should be allowed. The constitutional principles of human dignity, equality, and freedom of self-determination, as understood today in the broader bioethical debate are per se insufficient, and narrower criteria ought to be adopted. The legal grounds to justify an intervention by the EU in this field are less evident than in all other matters mentioned, however they can be found in the freedom of movement of EU citizens, which would suggest, to some extent, a uniform framework. With respect to the content of such principles, human dignity ought to be understood as objective and external, limiting self-determination, and reversibility of the intervention onto the body should also be taken into consideration.

Privacy & Free Flow of Data: privacy cannot be granted simply through informed consent. Consent is hardly ever truly informed, and the very possibility to dissent is limited, should one want to use the service or device requiring the collection of personal data for its operation. On the one hand, the current EU Regulation setting forth the “-by design” principle, should be narrowed down through the adoption of specific standards, specifying what satisfies that criteria in different classes of applications (see Recommendation nn. 17-21). On the other hand, consumers should be compensated for allowing access and use of – private and anonymized – data through post-sale services, enriching after-sale duties imposed on the producer.

Contrary to popular belief, the law is faster than any technological development.

If a time machine was invented tomorrow and time travel became reality, every aspect of the machine would already be regulated before news of the device could be shared with the world. If the first time traveller did not come back from his or her trip to the past, that person’s spouse could, under existing legal frameworks, sue the inventors of the time machine and claim them liable for the accident.

Such a claim would subsequently be filed in a court somewhere in the world, and a judge would have to determine whether a rule of negligence or instead an objective standard of liability applies, if a valid contract was entered into between the traveller and the inventors of the machine, and whether the rules governing travel across space also apply to travel across time. Some people would deem the judge’s conclusions correct, and others would deem them wrong, but in any case the decision would still be based on some law and some form of legal reasoning, since the only unacceptable answer by a judge is non liquet.

Given that robotics are already regulated, and that any judge’s decisions on their use will produce incentives to future parties that find themselves in the same conditions, as lawmakers we must shift our question from whether we should regulate to how regulation can most fruitfully be structured.

Most current robotic applications could be considered as “products”, and pursuant to the European Defective Product directive or the US Restatement Third of Torts, could be straightforwardly addressed as such. While the consequences of applying those rules may in some cases lead to reasonable results and thus appear uncontroversial, in other cases they may be problematic, and indeed produce a technology chilling effect (e.g.: in the cases of driverless vehicles and robotic prostheses).

In the problematic cases, ad-hoc regulation may prove reasonable, necessary, and ultimately essential for the correct development of robotic technologies. Precisely because of that, opponents of robotic development would often be better served by legislative inertia rather than intervention. By avoiding the adoption of new rules, the negative incentives triggered by existing regulation would in fact deter the development of those applications.

While product liability rules are normally intended as a tool to incentivize high standards of safety in the design of products, in some cases no actual positive effect is achieved. For instance, after a liability exemption was introduced in the United States shielding the producers of commercial aircrafts from litigation (GARA), the number of accidents did not increase.

For these very reasons, sceptical technologists, engineers and roboticists may come to believe that a thoughtful legislative environment is essential to their work , their research and their business. When adequately pondered and rationally applied, regulation is therefor not an evil, but a useful tool.

How to regulate: The RoboLaw approach

Financed by the European Commission within FP7, the RoboLaw Project adopted a new approach to the problem of how to regulate robotics that is distinctly different from previous methodologies.

Firstly, the term “robot” is not a technical term from either an engineering or a legal perspective, but is derived from science fiction. As such, the RoboLaw project takes the position that there is no purpose in trying to develop an all-encompassing definition of the term “robot” . Rather than attempting to identify a common trait among such varied applications as robotic prostheses, driverless vehicles, softbots, industrial robots, robot companions, and automated vacuum cleaners, the RoboLaw project attempts to examine the peculiarities of each and point out differences between them.

For the purpose of examining the ethical, legal and social implications of robotics, this implies renouncing the idea of developing a uniform solution, a code or set of rules for robots as a single category. RoboLaw did not attempt to further elaborate, substitute, or overcome the laws of robots as Asimov thought of them.

Instead, the RoboLaw project determined that the best approach was to undertake a case-by-case analysis, addressing single kinds – or classes – of applications, pointing out the technical peculiarities of each, and through that, identifying both the ethical and legal implications that the emergence and diffusion of a similar technology may give raise to. Based on these considerations, the RoboLaw guidelines address four different application areas: driverless vehicles, robotic prostheses (and exoskeletons), surgical robots, and robot companions.

Each chapter, devoted to one single kind of device, starts with a technological analysis (conducted by engineers) that summarises the essential aspects of the technology, clarifies how it functions, and points out the most relevant challenges being faced for its advancement. Then, applying existing ethical approaches and theories, the issues are identified and discussed, and policy considerations provided.  Finally, a legal analysis considers applicable rules, determines if the incentives provided are desirable, and offers alternatives where needed.

The chapters then formulate recommendations that a legislator could use to enact ad-hoc rules for the specific robot addressed.

Which technologies to choose

In choosing which robotic technologies to address, RoboLaw took into account the novelty of the application, its possible societal impact, and its relevance.

Three of the applications considered by RoboLaw – robotic prosthetics/exoskeletons, surgical robots and robot companions – fall under the area of healthcare and, according to a recent study by McKinsey, are among the robotics technologies that show the most promise in terms of their ability to improve the quality of life of the great numbers of people.

The fourth – driverless vehicles – could represent a ground-breaking innovation, reduce the number of fatal road accidents, and eventually reshape the concept of transportation, traffic and the way our cities are designed and function.

Because the analysis required a good understanding of the technologies in question, we also considered the specific competencies of the engineers in our research group. As such, several relevant devices were left out, including drones, which are certain to modify many aspects of our future lives, and military robots with their many complex political implications.

Nonetheless, RoboLaw’s methodology can be applied to any robotic device, and should the project receive further funding, other technologies will most certainly be studied with this approach.

When and how to intervene: A functional approach 

The RoboLaw project adopted a functional perspective for deciding when and how to intervene.

It is not a robot’s intrinsic technical quality or characteristic alone that calls for regulation , and the ability of the robot to operate autonomously, and even its ability to learn and adapt its functioning, do not per se suffice in justifying a change in perspective.

Even a robot that can perform complex tasks without human supervision and take decisions towards that end may still not be deemed an agent in a philosophical sense, let alone a legal one. The robot is still an object, a product, a device, not bearing rights but meant to be used. What would justify a shift on a purely ontological basis (thus forcing us to consider the robot as a being provided with rights and duties) is what Gutman, Rathgeber and Syed call ‘strong autonomy’ – namely the ability to decide for one’s self and set one’s own goals. However, at present this belongs to the realm of science fiction, and it can be argued that this is not the direction we desire to take with robots in any case: we want robots to ease our lives, and therefore to do what we decide they should be doing. Should they be free to decide if, how and when to perform what we ask according to their own taste and preferences, then the purpose of developing robotic technologies would be defeated.

If society is to favour or pose some limits to technological development, then the technical aspects of the individual robotic device need to be taken into account, together with other elements, in order to decide when to regulate and in which direction. The market structure, size and condition (which may lead to market failures on one hand, and constitutional principles and fundamental rights on the other hand), do have a bearing in this respect.

In the case of robotic prostheses, which certainly qualify as products, the application of existing rules would most likely discourage their development. Considering the costs associated with research, and the limited market of potential users (at least in the early stages of product development), a producer may decide to never develop a technology in the first place if he is held strictly liable for all damages a prosthesis might cause.

Nonetheless, given the improvement in quality of life such technology could offer amputees, a policy argument could be made to conceive a better liability system where compensation is granted to the victim and the production of safe devices is ensured without holding the researchers and producer liable under all circumstances. Moreover, some legal grounds could be found to actually justify favouring their development (e.g.: art. 4 of the UN Convention on the Rights of People with Disabilities).

Adopting a functional perspective therefore entails analysing the laws that are currently applicable in light of the technological peculiarities and challenges of the device in order to determine the effect that they would produce, the incentives they trigger, and finally, to elaborate a superior solution to the case. The direction to be favoured is determined through the complex weighing of benefits and risks the device would bring about, pursuant to existing principles and rights, in particular as emerging from constitutions and other fundamental rights declarations.

Some examples

To summarize some of the most relevant considerations that are drawn by the guidelines, we could say that definitions are a very relevant issue. Unless driverless vehicles are defined and such definitions used to modify street codes, such devices won’t be allowed to circulate.

Liability rules in many cases could provide a large and material obstacle to the development of desirable applications. Nonetheless, the preferred solutions are technology-specific, and take into account technical peculiarities as well as market structures. For instance, while a large market of users might suggest that a compulsory insurance system requiring owners of driverless vehicles to insure themselves for damages to third parties is sufficient, it does not follow that the same rule could be successfully used with prosthesis or robot companions. The initially more limited market for prostheses, and the potentially very high and uncertain costs existing liability rules could give rise to, would probably discourage private insurance companies from offering such contracts. In the case of robot companions, some more effective solutions could be thought of, including the attribution of legal personhood, similarly to what is done today with corporations.

Since liability pursues safety together with compensation, in some cases the proved inefficacy of existing product liability rules may call for reform. The two issues may be addressed separately: the pursuit of standardization could prove the best approach for the safety, while the adoption of automatic compensation mechanisms (no-fault plans) could ensure victims appropriate compensation.

There are several standardization bodies both at an international and European level (e.g.: respectively International Standard Organization and the European Standard Organizations). Their role should be implemented, and specific authorities should eventually be established with the purpose of determining high safety standards that are narrow enough to be tailored to specific applications – ones that producers can actually conform to. If one takes the current European directives regulating the EC marking, one easily sees that they have too broad a scope. For example, the medical device directive is applicable to anything ranging from a surgical glove to an exoskeleton.

Standardization may also benefit the field of professions that make use of robotic applications, such as surgeons. Requiring doctors to complete specific training in order to obtain a licence to use a specific kind of surgical robot may at once increase the level of performance, reduce accidents and subsequent lawsuits, and encourage the diffusion of such novel techniques.

Can RoboLaw’s guidelines be used outside Europe?  

Though the guidelines were developed for the European Commission and address the European legal system, they could be easily transposed to other systems. The problems emerging robotic technologies give rise to vary little from country to country, and so the solutions proposed by RoboLaw could be somewhat similar. Considering that the main aim of the guidelines is to provide narrow tailored recommendations suggesting the legislators how they shall intervene, such results could be applied – with some necessary adaptation – far beyond European borders.

Moreover, even if the solutions were criticized – and we expect them to be in some cases – they do provide a starting point for a technical debate in legal and philosophical terms around specific solutions to precisely defined problems, which itself is a novel outcome that we truly hope to encourage.

More info

RoboLaw website
BBC Interview
RoboLaw paper presented to the EU Parliament
Robotics Business Review Podcast with Ryan Calo