Must companies take legal issues into account if an employee uses an exoskeleton? Or if the employee is constantly reachable by e-mail and mobile phone? And how about hiring or firing by algorithms? Many businesses will face these workplace questions in the near future, others face them already today. The current law provides some clues for best practices, as Prof. Isabelle Wildhaber explains.
The Future of Workplace-Related Law
The new workplace is constantly changing, inspired by the arrival of new technologies and increasingly uninhibited by geographical boundaries. As new technologies facilitate remote work, the physical and temporal bounds of the workplace are becoming more flexible. Many enterprises are working towards the goal of optimising work and production processes and creating a so-called smart factory in which humans and machines work together and automation is achieved through the deployment of intelligent machines. These intelligent self-learning systems are driven by cloud computing, breakthroughs in sensor technology and the creation of new algorithms that harness the power of big data.
These developments lead to a working environment and working conditions unthinkable at the time the work-related legislation entered into force. However, manufacturers and users of new technologies must nevertheless comply with the legal framework conditions. The challenge for manufacturers and users of modern technologies alike is anticipating legal issues, while work and the workplace are continually changing. Legal risks and imponderables involved with the use of automation processes, robotics and AI in the workplace may be difficult to identify. What must be taken into consideration from a legal perspective so that employers and employees, but also the social partnerships, can continue to thrive in an automated or “bot-based” economy?
Lots of Legal Issues Surrounding Mobile Work
One of the hottest issues currently being discussed in Switzerland is mobile work. Mobile work, in particular home office, has been on the increase since 2000. In November 2016 the Federal Council of Switzerland has published a report on mobile work called „Telearbeit“ (Telework). The report on Telework analysed if there is need for legislative action for mobile work. It concluded that there is no need for legal action and that the existing law suffices. According to the Federal Council certain issues, however, are worth further investigation. These issues are identified, amongst others, as working times and health & safety. The report of the Federal Council is being criticized by the Swiss Federation of Trade Unions. The latter sees an urgent need for legislative action. It accuses the Federal Council of finding legal gaps, yet not being willing to fill them. There are currently no special norms on mobile work in the existing laws. Without any specific agreement, it can indeed be unclear which rules are applicable. Since a high degree of trust, self-discipline and motivation of the employee is necessary in order to do mobile work successfully, it is therefore recommendable to conclude a written agreement on it.
An important issue in mobile work are the limits of employee protection regarding working time regulations. When doing mobile work, the employee can dispose of working times flexibly which one of its great advantages. Nevertheless, the (public labour protection) regulations regarding maximum working hours per week must still be met. Regulations on breaks and resting times as well as the ban on work on Sundays and during the night must be observed. Furthermore, employers must comply with health protection regulations, i.e. they must ensure a correct work environment (rooms, lighting, ventilation, noise prevention) and an ergonomic workplace. All of these duties are difficult to be implemented when employees perform mobile work. The employer is not in a position of direct control at the home of the employee. If there are references that not all working time or health protection requirements are met, the employer must find a remedy or prevent mobile work. In my opinion, when doing mobile work, employees take on more personal responsibility for keeping up with working time regulations and for the protection of their health. The increased need for control when managing employees from a distance raises different questions, such as: Are controlling systems being used by the employer admissible? Should reporting obligations for employees be installed? Are there access rights of the employer to the home office?
Last but not least, the risk of a violation of the obligation of secrecy and of data protection is higher when doing mobile work. A Guideline on this very central issue is recommended. For the same reasons it is advisable that materials, such as laptop, are put at disposal by the employer.
Liability if Algorithms or Robots Cause Damage?
Apart from these unsolved issues regarding mobile work, one of the most discussed legal issues about new technologies at the workplace is: Who should be liable in tort if algorithms or robots cause damage? In our legal system, we treat machines as an extension of the human individuals who set them in motion. If machines cause damage, we attempt to claim compensation from the manufacturer, producer or user. However, the classical approach of torts is difficult to transfer to robots, particularly when they learn independently and become “intelligent”. It can be difficult to establish with hindsight whether an act by the robot causing damage can be attributed to the original programming or subsequent independent learning (“training” through use). Finding “linear” and foreseeable causality will be complicated.
The question is therefore whether the robot itself should be liable. Liability in tort is an aspect of capacity to act. According to the current legal situation, a robot is incapable of providing a legally binding declaration of will due to a lack of legal capacity and capacity for action. Electronic declarations of will or acts must still be attributed to a human. In the future, a robot could be liable like a legal person (similar to a stock company) if it were registered in a public register and had assets. Additionally, it could have an obligatory third party liability insurance. In my view, this sort of an “e-personhood” is quite conceivable in the future.
Health and Safety Measures are Absolutely Necessary
When it comes to workplace safety, the classical liability of the employer for accidents at work and occupational illnesses is important. The employer has an obligation to take all possible preventative measures to protect the health of employees. Robotics impose new requirements on prevention as regards safety at work. Robots have to comply with all regulations regarding product safety and have to take into account the health burdens of the potential operating personnel. When planning construction of a robot, the manufacturer must have a precise idea of the concrete use of the robot in the workplace. A collaboration with the employer may be necessary.
Advances in robotics in the workplace will lead to a massive increase in the interaction between workers and robots. Think about collaborative robots such as Baxter from Rethink Robotics or Yumi from ABB. They are affordable multi-purpose robots for production purposes, friendly to people and easy to program by colleagues for different tasks. This new generation of industrial robots, which collaborate with human individuals, present a challenge to safety at work, as they operate outside safety barriers and cages. With the development of advanced collaborative robots carrying significant weights at high speeds, the risk of significant injury may increase. As Jim McManus, Safety Specialist at the OSHA (Operational Safety and Health Agency) said at the 2016 National Robot Safety Conference in Cincinnati, “If an OSHA compliance officer walks in and sees a robot operating without a guard, he is going to set off alarm bells. There is a robot, so there must be a cage or someone is going to get hurt, right?” To ensure the safety of employees who work with and around collaborative robots, certain norms and standards are required, such as the ISO technical specification on collaborative robots (ISO/TS 15066:2016) dating from 2016.
Safety at work issues are also raised in the context of personal care robots (PCR). These are service robots which perform tasks that make a direct contribution to the improvement of the quality of life of humans (with the exception of medical treatments). They challenge the concept of safety at work because they (1) are used for numerous requirements in environments which are not precisely defined, (2) come into contact with non-specialised users, and (3) share the workspace with humans. This is why formal and international safety standards on PCR were defined in ISO-Norm 13482:2014.
Another challenge to work safety and occupational health are stress-related illnesses. They are on the increase due to today’s constant employee reachability and due to physical and temporal bounds of the workplace becoming more flexible. Constant availability may pose a health threat. These changes raise the question about which public labour regulations are applicable, and in particular which statutes concerning working time apply to a mobile, decentralised workplace. Is an e-mail written on a Sunday night consistent with public labour protections regulations, even though the employee has no Sunday work permission? Does the employee need to be protected from checking e-mails during the week-end or on holidays in order to be able to “de-stress”? In an automated workplace, it seems important to debate how we want to deal with issues such as protecting off time, defining working time, and making sure stress-related illnesses can be prevented.
Surveillance May Have Effects on Employee Health
Yet another issue of work safety and occupational health are monitoring and control mechanisms in the workplace that may have the effect of employee surveillance adversely affecting employee health. Recorded data may provide information about working speed, the health situation or the precision of the working employee. The fact that monitoring is one of the most frequent uses for robotics clearly indicates the relevance of data protection in an age of automation, robotics and AI. The affected person’s right of personality must be set in relation to the employer’s entrepreneurial freedom to organise his or her business. It will be important to precisely define the purposes of the systems used and to delimit the necessary handling of data on the basis of this. Furthermore, it is important for data processing in employment relationships that employees be informed about the use of monitoring systems in advance for reasons of transparency and for the employer to ensure that the employees’ consent is obtained. However, consent can create legal difficulties, amongst other things in the context of big data analyses. Given the balance of power in the employment relationship, consent will seldom be completely voluntary and it is revocable. Even without the employees’ consent it should be attempted to guarantee the legally compliant implementation of the new technologies through a well-considered choice of the data to be stored and possibly by rendering such data anonymous. Personal monitoring should only ever be considered as a “last resort” even in such cases (in compliance with the principle of proportionality).
The affected person’s right of personality must be set in relation to the employer’s entrepreneurial freedom to organise his or her business.
Monitoring has long been a reality, as shown by the example of GPS-supported mobile devices. Such devices, for example, direct “pickers” in the Amazon warehouses to the right shelf in which they find the merchandise ordered by the customer. There are reports from England that the mobile GPS computers raise the alarm if the employee takes a break outside regular break times. Monitoring problems in relation to the right of personality and employee data protection will become more acute in the future in view of new technologies. In the case of an intelligent assembly line in a smart factory which is linked to an assembly line worker and slows down if, for example, the assembly line worker’s pulse becomes too high, health data about the employee is collected which should only be possible under very restricted preconditions in view of the protection of the employee’s personality.
Another example are industrial exoskeletons. Exoskeletons are generally steered through body movements on the basis of cognitive and physical interaction with the human enabled by the drivetrains, sensors and microcontrollers, as well as the intelligent control system. With the exoskeleton tested in Daewoo’s Korean wharfs, wharf workers can carry very heavy loads. It therefore serves to protect the health of the employee, but on the other hand it can involve risks for the right of personality. Is it permissible for the exoskeleton to analyse the posture of the wharf worker and draw conclusions about his psychic condition? Perhaps certain data should not be used at all for the purposes of the employment relationship.
What About Equal Treatment and Discrimination?
Exoskeletons may also raise questions of equal treatment and discrimination. Should exoskeletons be dealt with like parts of the body if they are connected to the human body, even if this is not a permanent state? There is discussion about whether public buildings or rooms can refuse access to the wearers of exoskeletons or may require such exoskeletons to be taken off or deactivated, including for reasons other than security. And can exoskeletons become legally necessary “reasonable accommodations”? Such accommodations compensate for disadvantages and promote factual equality and non-discrimination. This would mean that an employee, e.g. a disabled person or a person suffering from back injuries, would have a claim for reasonable accommodation from the employer in an individual case. Given the continual technological development of aids which compensate for disadvantages and the fact that their cost will reduce significantly in the future, employers should keep an eye on these developments, in particular in the USA, for example. In other countries, such as Switzerland or Germany, legal measures for the promotion of integration of people with disabilities into the employment market are primarily based on the concept of social insurance. Should invalidity or accident insurances approve applications by a paraplegic for the payment of the costs for specific exoskeletons (for example the ReWalk with costs of around EUR 72,000)? Currently, much wearable and human enhancing technology may not be objectively reasonable or may pose undue hardships because of its novelty or cost. However, as this technology becomes more common and prices reduce, it becomes more likely that social insurances or employers may be required to provide it to aid disabled employees in performing their jobs.
The question of equal treatment and discrimination arises with algorithms generally. Algorithms may not be programmed in such a way that they discriminate, neither directly nor indirectly. In the context of the workplace, algorithms must comply with anti-discrimination regulations. Theoretically, algorithms can make decisions based on facts without prejudice by leaving criteria such as the place of origin, age and gender of an applicant out of consideration when making their decisions. In addition, there are technical methods such as “data repair” to exclude indirect discrimination in algorithms based on this data. There is nevertheless extensive criticism of decision-making by algorithms and the assumptions on which algorithms are based are increasingly disputed. After all, algorithms are programmed by human individuals and can therefore reflect their prejudices. The repetition of human behaviour can deepen discrimination which already exists.
In the context of the workplace, algorithms must comply with anti-discrimination regulations.
An example for algorithmic decision-making in the workplace is “hiring by algorithm”, It has become very widespread in large companies in recent years. Apart from non-discrimination, algorithms must be programmed in such a way that no impermissible questions are asked which encroach on constitutionally protected employees’ interests. Questions about private life and character may not influence a decision in favour of or against a job applicant made by an algorithm. This is precisely the problem with numerous personality tests which form part of e-recruitment.
Employment Law Issues Surrounding the “Robo-Boss”
Algorithms may also be used for the process of firing employees (“firing by algorithm”). Algorithms can collect and evaluate huge amounts of data, and use this as a basis to identify development tendencies, predict results, make recommendations and decisions and perform acts. It is therefore definitely conceivable that the decision to fire an employee could be taken by an algorithm in the future. However, whereas an algorithm or a robot can take the decision to fire and make this proposal, it cannot under any circumstances announce a termination of employment because it does not have the legitimate authority to fire a person. As long as there is no “e-personhood”, algorithms or robots do not have the capacity to act. Therefore, dismissal must be announced by an authorised human superior, in other words the managing director, an executive board member or an authorised person. And the responsibility for the dismissal must currently clearly remain with the employer as the natural or legal person behind the robot. This is reasonable because robots are not (yet) in a position to make value judgements and weigh up interests.
Similar thoughts can be given to algorithms or robots issuing instructions to employees. According to a study by the Massachusetts Institute of Technology (MIT) in 2014, humans were happier taking instructions from robots and were both more productive and more satisfied. The Japanese electronics group Hitachi has started to develop an AI system which allocates work tasks to human employees on the basis of data analysis and past working processes. This is supposed to increase productivity by 8%. Such instructions are legally possible. Obviously, they must remain within the framework of the instructions a human superior would be permitted to issue. The instructions of an algorithm or a robot are intended to optimise working processes through data analysis. Again, this raises data protection issues.
All such decisions taken by algorithms or robots – hiring or firing or issuing instructions – raise the additional question about a prohibition of automated individual decisions. On the basis of new information and communication technology, decisions which have legal consequences for the persons affected or have significant adverse effects on them are increasingly taken in automated procedures, also often on the basis of profiles based on statistical data and calculations (profiling). For some time already now, the EU has had provisions on the permissibility of automated individual decisions. The desire for more «humaneness» emanating from a prohibition of automated decisions assumes that human decisions are better than automated decisions, a belief which is being questioned by some, especially outside of the EU.
Replaced by Robots: Dismissals
Let me finish with one last workplace-related legal issue. One of the most frequent discussions about automation, robotics and AI is whether the industry will create more jobs than it eliminates. In any case, many employees will doubtlessly lose their jobs even if new jobs are created. Employers are entitled to change job profiles in line with requirements, to completely replace jobs by automated processes and to announce dismissals. Job profiles will change. If employees cannot fulfil the changed higher expectations in spite of appropriate training and further training programmes offered, dismissals will be inevitable. For example, work with exoskeletons on building sites will present new challenges for employees. Not every building worker will be capable of operating a robot. A building worker who should learn the necessary qualifications (e.g. the functions and operation of the exoskeleton) may take part in a workshop, but be unable to operate the exoskeleton, even after several attempts, and consequently be dismissed by the employer. Is this a redundancy for operational and business reasons where the grounds for dismissal originate from the employer’s sphere, or a dismissal on grounds of personal capability or conduct?
Employers are entitled to change job profiles in line with requirements, to completely replace jobs by automated processes and to announce dismissals.
If managerial authority includes the right to instruct the learning of new qualifications, the primary reason for the dismissal is not the decision by the employer to automate his business, but the lack of suitability or the conduct of the employee himself or herself. Learning new qualifications made necessary by technical progress must be viewed as part of the performance owed. The decisive factor here is the occupational profile at the present time. In the sample case, learning the functions of an exoskeleton is still within the scope of the reasonable duty to learn. This then constitutes dismissal on grounds of personal capability. By contrast, compulsory redundancy for operational reasons involves such significant changes to occupational profiles that they become very different from the job description in the relevant contract of employment and consequently go beyond the originally agreed programme of duties, e.g. if the job requires programming capabilities after automation.