Law by Algorithm: An Introduction

This post is part of a special series including contributions to the OBLB Annual Conference 2022 on ‘Personalized Law—Law by Algorithm’, held in Oxford on 16 June 2022. This post comes from Gerhard Wagner, co-author with Horst Eidenmüller of ‘Law by Algorithm’, the book discussed on the second panel of the conference.

I. Law by Algorithm

For quite some time, the disruptive effects of digital technology on the economy and on society were limited to jobs that required much routine and little sophistication. These times are over. The digital transformation has reached the higher professions, such as doctors, accountants, and lawyers. Only priests seem to remain unaffected, at least for the time being. In addition, one standard element of science fiction literature, namely andromorph artefacts that look, act and behave like humans, seems to finally appear on the horizon, together with other computer-driven appliances that are capable of moving around, such as autonomous cars.

How shall the legal system react to the digital transformation in the sense just described? – This question is the thread running through the chapters of ‘Law by Algorithm’. In seeking to find answers, we seek to avoid extreme approaches. In general as well as in legal discourse, problems of the digital transformation are often approached from either an utopian or a dystopian vantage point. Some commentators fear that digital technology will destroy fundamental values and goals such as individual autonomy, protection from wrongful harm, and fair resolution of disputes. They paint the picture of a dark future, where robots haven taken over control and are dominating humans. On this account, the proverbial ‘ePerson’, once created, will become emancipated and force its will on humans. The institutions that we know and the values embodied in them will be transformed into something else. In contrast, the utopian view expects a bright future for humanity, now that it can team up with digital artefacts to further its ends. Online shopping will continue to dramatically increase consumer welfare, autonomous cars will bring down the number and severity of traffic accidents, corporations will operate more efficiently, and disputes will be resolved without delay and at incredibly low cost.

‘Law by Algorithm’ does not pretend to have a looking glass and to know with acceptable certainty what the future will bring. If anything, it seems reasonable to expect that digital technology holds the potential of improving the lot of the human race, but that it also poses a threat. Economic and social transformations are usually a mixed bag, and cannot be judged as either only ‘good’ or ‘bad’. As a consequence, ‘Law by Algorithm’ takes a nuanced, multi-faceted approach. To begin with, we primarily try to take stock, ie to thoroughly investigate what is really ‘going on’. The second step is to identify the problems of digitization in each distinct context. Some of these problems have already materialized, others must be expected to become visible in the near future. Ultimately, the goal is to develop tailor-made solutions to the problems that we identified. In doing so, our focus is on the core areas of private law, ie contract, tort, and corporate law, and not on regulation. While regulation obviously plays an increasing role in the legal response to digitization, as epitomized by the recent enactment of the European Union’s Digital Markets Act, we count on private law tools such as disclosure mandates, rights of withdrawal, contractual choice between different options, tort-based liability, but also on legal guideposts for the various mechanisms of alternative dispute resolution offered by private entities.

In the following sections, I will set out our approach by discussing three developments and issues which have become central to the debate about ‘Law by Algorithm’: the market for dispute resolution, liability for autonomous systems, and consumer contracts / online shopping.

II. The Market for Dispute Resolution

Much of the discussion on dispute resolution is dominated by the vision of a ‘digital judge’ or, somewhat more modest, ‘digital courts’. Depending on the spirit and temper of the respective commentator, the digital version of what used to be judicial dispute resolution is presented as panacea or anathema to the long-standing problems of cost, delay, complexity of proceedings and uncertainty of outcome.

‘Law by Algorithm’ takes another view. We look at dispute resolution not as a monopoly controlled by the state but as a market in which several actors, public and private, compete with one another. And we predict that digitization will lead to a dramatic shift in market shares, away from the public courts, to the various private suppliers of dispute resolution services. In short, digitization will lead to a radical privatization of dispute resolution.

The range of digital tools that help to settle disputes, or to avoid them in the first place, is broad already. Smart contracts promise automated contract performance at low cost and under perfectly defined circumstances. No court needs to become involved, and no judgment needs to be executed, as the contract ‘executes itself’. Large businesses that face consumers have begun to offer an online complaints mechanism that is designed to deal with fairly large numbers of customer complaints in an expedient and cost-efficient manner. As consumers as users are held in high esteem in the online industries, internal complaints management is designed to keep customers happy and around. Again, no court is needed to ensure this kind of consumer satisfaction.

From internet-based complaints management it is only a small step to online dispute resolution in its already established form of the European Regulation (EU) No. 524/2013. Online dispute resolution in the technical sense implies the involvement of a neutral third party that helps the parties to resolve their dispute. The main playground for online dispute resolution is conciliation and mediation. It may be expected that online arbitration will experience much growth, and it is here that an algorithmic decisionmaker might take office very soon. These mechanisms of online dispute resolution promise dramatic reductions in cost and time, but not only this. Since the beginning of time, dispute resolution processes of any sort were cramped between the conflicting goals of accuracy of outcome on the one hand and cost and delay on the other. Today, for the first time in history, the combination of internet-based dispute resolution procedures with an algorithmic arbitrator or conciliator promises to deliver reasonable or even optimal accuracy of outcome at trivial cost.

But still, it would be naïve to believe that the new forms of dispute resolution just described come without any problems of their own. Smart contracts not only offer automated performance, they also force it on whichever party is liable to perform. The potential for abuse is obvious. Customer complaints handling mechanisms make many customers happy, but may be particularly unfair to poor or low-spending customers. And online dispute resolution is clearly in need of a legal framework that ensures procedural fairness, and not only procedural efficiency.

III. Liability for Autonomous Systems

Autonomous digital systems that are able to direct movable hardware have been confined to science fiction for a long time. But now they are literally ‘hitting the road’. Autonomous vehicles promise huge savings in accident costs, as they eliminate the two main causes of traffic accidents: drunk driving and speeding as well as other violations of rules of the road. Similarly, robot surgeons such as the Da Vinci Surgical System, mainly used for prostatectomies, offer significant advantages in precision compared to traditional ‘manual’ surgery.

The anticipated reduction in the number of traffic victims and the severity of injuries is certainly good news. Without a doubt, however, accidents will continue to occur. When an autonomous digital system causes harm, the question arises: who is liable? The European Parliament has floated the bold idea that the digital artefact itself should be held liable, under the guise of an ‘ePerson’. This far-reaching proposal raises the question whether the liability system already in force is fit for the task or in need of reform. In fact, the arrival of autonomous digital systems marks a radical shift in control. Whereas analogue machines and appliances are controlled by their users, or operators for that matter, digital systems are controlled by software that is off limits to the user. The once-so-proud driver of a car is reduced to the role of a ‘mere’ passenger. The real ‘driver’ of an autonomous car is the manufacturer who uses the same algorithm to control a whole fleet of vehicles.

This shift in control caused by the introduction of autonomous cars and other machines into society is a major challenge to existing liability rules. If anything, it suggests to take a closer look at the rules governing the responsibility of manufacturers, ie the law of products liability. The EU already operates a uniform regime of products liability under Directive 85/374/EEC, but it remains doubtful whether it applies to manufacturers of software, particularly where such software has not become embedded into hardware. In addition, one of the main elements of products liability, namely the concept of product defect, is anything but straightforward in its application to algorithms. The EU Commission hesitates to reform the Directive on products liability and believes that new liability rules targeting the operator of so-called AI systems are needed. Given that digital systems are associated with a shift in control, away from users/operators and towards manufacturers, this idea warrants some scepticism.

IV. Consumer Contracts / Online Shopping

Online shopping develops rapidly as it offers huge benefits to consumers. It opens up a new world to many buyers, helps to compare prices and allows to search for the best bargain. However, contracting in the digital world also poses new threats. Much of the internet business is controlled by online sellers that possess large amounts of customer data. The analysis of such data yields information about the customer that sellers never had before. It is no exaggeration to say that, today, (online) sellers know more about their customers than these customers know about themselves. Informational asymmetries of this kind are new. And the law is not prepared to deal with them.

One risk raised by digital transactions is that businesses pocket all of the surplus created by contracts of exchange. In the brick-and-mortar world, the market price protected infra-marginal consumers to benefit from a few well-informed, price-sensitive buyers who pulled the market in the right direction. Nowadays, with all the data that are available, every individual becomes a market in itself. Inasmuch as the seller has the information necessary to estimate the reservation price of an individual customer, first-degree price discrimination is within reach. With first-degree price discrimination one hundred percent of the joint surplus generated by a particular transaction is pocketed by one of the parties.

Big data has paved the way to another business strategy that hurts consumers, namely the deliberate exploitation of cognitive biases. If sellers know the times during which individual customers are in a ‘hot state’, ie when they have difficulties in resisting certain offers, they have the chance to pitch their offers exactly at such ‘hot’ times. In extreme situations, a person who tries to quit smoking may be presented with a special offer of cigarettes right at the time when she is most vulnerable to fall back on her old habit.

Finally, algorithmic shopping may lead consumers into the same echo chambers that are known from the political sphere. The intake of political information that was curated by a digital platform to fit the pre-existing preferences of the addressee has led to a divided public and increasingly radical views held by individuals at both ends of the political spectrum. Transposed to the commercial sphere, algorithmic consumption will only deepen pre-existing preferences with a view to certain consumption choices. Tell your digital butler to order pizza once, and you will get it for the rest of your life.

In thinking about potential remedies, we focus again on common-law mechanisms, such as a right to withdraw from consumer contracts made online, as it already exists in the EU, but also on a new ‘right to anonymity’, ie the right to go shopping online under a cloak. Much like a stealth aircraft, consumers should retain the option to navigate the net ‘under the radar’ of big data.

Gerhard Wagner holds the Chair for Private Law, Commercial Law and Law and Economics at the Law Faculty of Humboldt-University, Berlin.