Coping with Moral Dilemmas in German Criminal Law Theory and Justice: Classical Cases and Modern Variants

Abstract

Dilemma scenarios have always been among the most common problems of moral philosophy and criminal law theory. One only has to contemplate the Plank of Carneades, the classic thought experiment whereby two shipwrecked people’s only hope of rescue is a floating board that can only be occupied by one person. Other scenarios are Welzel’s switchman case and the well-known Trolley Problem. In most of the debated cases the death of one or more people is absolutely unavoidable. The protagonists do not cause the situation but are fated to come into conflict. The focus of this article is on one recent and one future variant. First, the prioritization of medical aid (also known as “triage”) is the subject of intense debate, because the COVID-19 pandemic posed a permanent risk of a temporary collapse in the health system in several countries. Situations had arisen whereby some patients can no longer be treated owing to lack of capacity. It can be asked whether a decision to treat may be based on which patients have a better chance of survival, whether reckless previous behaviour may play a role, and whether a treatment, once started, may be discontinued in favour of another. Second, dilemma scenarios are also one of the last remaining (largely unresolved) legal difficulties of autonomous vehicles. Never before has a machine been given the power to determine the life or death of human beings. Even though the automotive industry promises that such situations will hardly ever occur, the problem could prove to be a tangible obstacle to acceptance and innovation. The article offers solutions for those distinct scenarios, but it is also intended to demonstrate the underlying legal concepts of German law: namely, the tripartite analysis of criminal law and the idea of human dignity as a fundamental principle of the German constitution.

Introduction

Substantive criminal law in the legal system of common law still develops quite independently from that in civil law and vice versa.¹ Nevertheless, quite often we deal with similar problems. For example, ethical dilemmas have always posed a special fascination for the public, as well as in moral philosophy and jurisprudence. There is much to suggest a role of the psychological phenomenon of “wanting to be scared”² in this, and even philosophy professors and law teachers are not entirely immune. One only has to contemplate the Plank of Carneades, the well-known thought experiment whereby two shipwrecked people’s only hope of rescue is a floating board that can only be occupied by one individual.³ One castaway kills the other to secure the life-saving plank for himself and be rescued. The case may be found in every German textbook on criminal law.⁴ Real cases from the 19th century have been handed down from US-American⁵ and English⁶ maritime law. The English Mignonette case is a particularly well-known example in Germany.⁷ The case concerned cannibalism among shipwrecked people; the proceedings ended with the offenders being sentenced to death and later pardoned after serving six months in prison.⁸

However, there are much more recent scenarios. The pandemic triggered by the SARS-CoV-2 virus had led to temporary situations (e.g., in Italy and France), where hospital physicians had to decide which patients were allocated an intensive care bed with a ventilator (and thus the prospect of survival) and who had to die because they could no longer be treated owing to lack of capacity. The prioritization of medical assistance in emergency situations is known as “triage” (from the French word trier = to sort). Although the collapse of the German health care system has been averted, the vaccination campaign is showing clear successes and the current omicron variant seems to be quite harmless, the end of the pandemic is hardly foreseeable in view of possible new virus variants. In any case, coronavirus will not be the last pandemic or crisis in our lifetime. Resources are not unlimited, even in the Western world. So the problem of the allocation of intensive care resources is still the subject of intense debate.⁹ It can be asked whether the decision may be based on which patients have the better chance of survival, whether children and young people with greater life expectancy may be given preference over older people and those with previous illnesses, whether the vaccination status or previous reckless behaviour may influence the decision, and whether a treatment, once started, may be discontinued in favour of another.

In traffic, scenarios can also occur in which violations, human error, or natural phenomena lead to risky situations when accidents cannot be avoided. Highly and fully autonomous vehicles can record and process information very quickly. If an accident cannot be avoided, the computer is quite capable of considering all possible courses of action to minimize damage. Of course, the vehicle must always choose material damage over personal injury, but one can also ask whether the system may be programmed to keep the number of personal injuries as low as possible.¹⁰ At some point, the technical systems may be able to determine how many people are in a vehicle or even their age. However, the “right” decision cannot ultimately be determined by a machine or the automotive industry. To promote autonomous driving, the industry is rather dependent on acceptance by the population and a secure legal framework.

For some years now, a worldwide internet survey by US researchers – titled Moral Machine – has attracted attention. The results of a survey on autonomous driving were published in an article in Nature¹¹ magazine. The vast majority of the test participants allowed the quantitative weighing of human life;¹² the majority also showed a strong tendency to spare children over elderly people (although there are cultural differences in this respect)¹³ and judged whether the people concerned had obeyed the rules. There are also even more recent surveys concerning the allocation of ventilators to COVID-19 patients on a smaller basis, the results are similar, but not as clear.¹⁴

These two problems – the prioritization of medical aid (infra III.) and the emergency algorithms for autonomous vehicles (infra IV.) – are the focus of the following article.

Legal Foundations and Classic Thought Experiments

Legal Basics

The German criminal law – as well as many other legal systems that follow the same pattern¹⁵ – distinguishes between three levels of inquiry: satisfaction of all objective and subjective elements of the offense as defined in the statute (Tatbestandsmäßigkeit),¹⁶ the wrongfulness (Rechtswidrigkeit), and the culpability (Schuld) in the sense of “blameworthiness” of the actor.¹⁷ The three-step analysis of criminal liability may be familiar to some Anglo-American scholars, as it also appears in a few US textbooks.¹⁸

Statutory Rules Concerning Dilemma Scenarios

In consequence, the German Criminal Code (Strafgesetzbuch – StGB) contains (contrary to US-American § 3.02 Model Penal Code; English case law is also different¹⁹) two necessity regulations (choice of evils defences), to distinguish between justification (excluding wrongfulness) and excuse (excluding culpability).

a) Justifying Necessity (Rechtfertigender Notstand) is regulated in § 34 StGB; the provision reads as follows.

“Whoever, when faced with a present danger to life, limb, liberty, honor, property or another legal interest which cannot otherwise be averted, commits an act to avert the danger from themselves or another is not deemed to act unlawfully if, upon weighing the conflicting interests, in particular the affected legal interests and the degree of the danger facing them, the protected interest substantially outweighs the one interfered with. However, this only applies to the extent that the act committed is an adequate means to avert the danger.”

This general “lesser evil defence” was the last grounds of justification to be expressly integrated into written law and has been part of the German Criminal Code since 1973.²⁰ However, after a decision by the Imperial Court of Justice (Reichsgericht) in 1927²¹ to legalize abortions performed to save the mother’s life or health, it has been recognized under customary law. Two further provisions from the Civil Code (Bürgerliches Gesetzbuch – BGB), namely §§ 228 and 904 BGB, are much older. They allow interference with the property of others but are not applicable to personal injury.²²

b) Excusing Necessity (Entschuldigender Notstand) is regulated in § 35 StGB; the provision reads as follows.

“(1) Whoever, when faced with a present danger to life, limb or liberty which cannot otherwise be averted, commits an unlawful act to avert the danger from themselves, a relative or close person acts without guilt. This does not apply to the extent that the offender could be expected, under the circumstances, to accept the danger, in particular because said offender caused the danger or because of the existence of a special legal relationship; the penalty may, however, be mitigated pursuant to section 49 (1), unless the offender was required to accept the danger on account of the existence of a special legal relationship…”

The wording makes it clear that the provision only protects the perpetrator, if he tries to avert a danger to his own life, limb and personal liberty or that of a close person. However, “excusing necessity” may even be granted in cases where life stands against life.

Unwritten Legal Rules Concerning Dilemma Scenarios

In addition, in dilemmatic situations, two other institutes have to be taken into account; even so, the written law does not explicitly regulate them:

• a) The “collision of duties” (Pflichtenkollision) is a defence sui generis (infra III.2.2(3.2.2)), defined as a conflict between two or more grounds of obligation of an equal nature that cannot be cumulatively fulfilled.²³ Because a decision of the Imperial Court of Justice (Reichsgericht) in 1890,²⁴ it has been recognized under customary law.²⁵ It is regularly applied by the courts; however, it remains highly controversial if the institute excludes the Tatbestandsmäßigkeit²⁶ or it is used in defence as a justification or excuse.

• b) Furthermore, the existence of a supra-legal excusing necessity (übergesetzlicher entschuldigender Notstand) is discussed in Germany (infra II.2.3.b(2.2.3.b)). It is meant to cover cases beyond the remit of § 34 StGB, because such interests cannot be quantified (such as human lives); nor are they within the remit of § 35 StGB, because the perpetrator is acting in favour of persons other than relatives and close friends.²⁷

Switchman Case (Weichenstellerfall) and the Trolley Problem

Outline of the Problem

Dilemma situations were often the subject of theoretical thought experiments. At the beginning of the 20th century, road and rail traffic often played a role in this. As early as 1915, Kohler presented the case of an “automobile driver” whose car could no longer be brought to a standstill over a short distance. He could only steer to the right or left, but there were people in front of him, who could not step aside.²⁸ Very similar are the German switchman case (Weichenstellerfall) by Hans Welzel (1951)²⁹ and the Trolley problem described by the British philosopher Philippa Foot (1967) and further developed by the recently deceased Judith Thomson (1985). In Welzel’s textbook example,³⁰ a freight train approaches a fully occupied standing passenger train owing to a positioning error. A switchman recognizes the danger and diverts the freight train to a siding, so that it races into a group of track workers, all of whom are killed. In the Anglo-Saxon variant, an out-of-control streetcar threatens to roll over five people. Again, they can only be rescued by pulling a lever, so that the trolley will switch to a different set of tracks (however, there is one person on the side-track)³¹ or by pushing an uninvolved “fat man” off a bridge (his massive body would bring the streetcar to a stop).³² We see that thought experiments do not necessarily have to represent reality, but rather the structure of a problem.

No Justification Under German Law

According to German criminal law, the switchman acts illegally when he diverts the approaching train to another track, so that only a few track workers are killed instead of the many passengers.³³ Also illegal is the killing of a single person (including Thomson’s fat man) to save several other people. Furthermore, in terms of quality, no distinction may be made: the “withered old man” must not be sacrificed at the expense of the powerful young man or the mentally ill person in favour of a Nobel Prize winner,³⁴ because each life represents an absolute maximum value.

For example, after the terrorist attacks of September 11, 2001 on the World Trade Center in New York, the question was discussed whether the shooting down of a fully loaded aircraft hijacked by terrorists and intended to be used as an instrument of attack (against a high-rise building or a nuclear power plant) is legally permitted or conversely would lead to severe criminal consequences for the fighter pilot and his superiors, in view of the numerous innocent passengers involved. A German public-law authorization, which would have allowed the German air force to shoot down the plane, was declared void by the German Federal Constitutional Court (Bundesverfassungsgericht),³⁵ because it was incompatible with the guarantee of human dignity (Article 1, paragraph 1 of the Grundgesetz – GG, i.e., Basic Law) and the right to life (Article 2, paragraph 2, sentence 1 of the GG) guaranteed by the German constitution. Of course, the court clarified that this only applied to people on board who were not involved in the kidnapping, not to the terrorists. Most criminal scholars regard this judgment as decisive for the interpretation of § 34 StGB,³⁶ even though the constitutional court expressly emphasized that it did not have to resolve how a shooting that occurred regardless of this finding was to be assessed under the applicable criminal law.³⁷ Since then, the discussion has arisen repeatedly in literature and the media.³⁸

In US criminal law, where the choice of evils defence is based more on rational expediency, weighing lives against each other does not appear to be excluded from the outset³⁹ – in contrast to the common law tradition.⁴⁰ The same could possibly apply to the criminal law in countries such as Japan, South Korea, and Taiwan, even if these three jurisdictions have traditionally strong connections to German legal thought. If the conflicting interests are equal, at least the statutory law gives neither a positive nor a negative verdict of illegality. Furthermore, the necessity regulations in Article 37 of the Japanese Criminal Code (緊急避難), Article 22 of the Korean Criminal Code (긴급피난), and Article 24 of the Taiwanese Criminal Code (緊急避難), only provide for the exclusion of criminal liability as a legal consequence. Nevertheless, it is controversial in these countries (according to the German tripartite analysis of criminal law) whether the provisions contain a justification or excuse, or both.⁴¹

According to the German understanding, the switchman would only act in the sense of a justifying state of necessity (§ 34 StGB) if he perceives an overriding interest; e.g., if he sets the switch in such a way that the train strikes a herd of sheep.⁴²

Possible Excuses Under German Law

a) The actor can only achieve impunity if he wants to save himself (as with the shipwrecked people) or a close relative (for example, on the train). According to the German understanding, such conflict situations can only be taken into account at the level of culpability: namely, through the necessity defence according to § 35 StGB. Under the terms of the provision, an individual could take a life to save himself or close others from death or serious bodily injury. The guilt is reduced on the one hand because of the special psychic predicament and on the other hand because of a certain reduction of injustice compared with a normal case of homicide (at least one human life is saved).

b) The excusing supra-legal necessity defence (übergesetzlicher entschuldigender Notstand) is based on similar considerations: According to the major academic opinion,⁴³ it is applied in constellations where the perpetrator (as in the case of the switchman) sacrifices a few people for the benefit of many; the latter need not necessarily be relatives of the perpetrator (as in § 35 StGB). Such an excuse has been previously discussed by the courts in connection with various trials concerning the Nazi euthanasia crimes.⁴⁴ The main focus has been on the accusation of aiding and abetting the murder of mentally ill people in the years 1940–1941 in six killing centers of the German Reich that were specially equipped with gas chambers. Hospitals, nursing homes, and sanatoria were required to report on their patients. In their defence, physicians in psychiatric hospitals who cooperated and were later indicted claimed that by extraditing a few patients, they had prevented something worse (namely, the killing of all inmates of the institution). However, in many cases the alleged dilemmas turned out to be refutable lies.⁴⁵ Insofar as an actual conflict for the physicians could not be ruled out, the German higher courts occasionally reached a “personal reason for exclusion from punishment” decision in the immediate post-war period (1945–1949).⁴⁶ However, the Federal Supreme Court (BGH) made no such decisions after 1949.⁴⁷

Therefore, the preconditions for a supra-legal necessity defence are largely unclear. The shooting down of a passenger aircraft used as an instrument of attack (in a “9/11-scenario”, supra II.2.2(2.2.2)) is a possible future precedent that has not been excluded by the Federal Constitutional Court (Bundesverfassungsgericht).⁴⁸

Interim Conclusion

It makes a difference whether the perpetrator acted in accordance with the law or merely had been discharged for lack of personal culpability. Despite the exclusion of culpability (supra II.2.3(2.2.3)), the potentially injured are not subjected to an unreasonable duty to acquiesce in either case constellation (their right of self-defence is not lost),⁴⁹ which would be the case if the perpetrator was justified. Three arguments seem to be decisive for the solution of Welzel’s switchman case and Foot’s Trolley problem according to German criminal law:

Argument 1: Every human life has the same rank and eludes any quantification. No one can be expected to sacrifice himself in favour of one or more others.

Argument 2: Action is more significant than mere omission. The active killing of a person weighs so heavily that even the rescue of a large number of people could never compensate for this. Prohibitions have to be respected over commands.

Argument 3: Secured legal positions (standing on a side-track without danger) enjoy a higher level of protection than mere expectances (the chance to be rescued by changing the switch).

This is almost logically compelling. Otherwise (to take an example from US literature), in extreme cases one could force every healthy person to sacrifice himself to save the lives of at least five seriously ill patients with his transplantable organs (heart, lungs, two kidneys, and one liver).⁵⁰ Such a cruel duty of solidarity⁵¹ cannot exist and is unlikely to be advocated even by the greatest utilitarian. In Germany, it would be incompatible with the guarantee of human dignity of Art. 1 I Grundgesetz (“human dignity shall be inviolable”, supra II.2.2(2.2.2)). The violation of such elementary fundamental rights cannot be justified by advantages for third parties.

Transferability of These Principles to the Triage?

Outline of the Problem

It is now questionable how the aforementioned principles can be applied to the problem of triage in the context of COVID-19. In addition to the numerous statements from criminal law scholars,⁵² the German Ethics Council⁵³ and national medical societies⁵⁴ have also issued recommendations. According to these recommendations, the decision as to who should be given priority treatment should be made according to the “two-man rule,” with the participation of two doctors experienced in intensive care medicine.⁵⁵ In the prioritization process, not the age of the patient (Italy and Switzerland are more outspoken),⁵⁶ but the chances of success of an intensive care therapy should play a role; these are considered poor if no improvement or stabilization is to be expected or survival would be tied to permanent confinement in an intensive care unit.⁵⁷

However, the problem of course remains primarily a legal one.⁵⁸ Recommendations of the Ethics Council and professional associations must be measured against the applicable law: In December 2021 the German Constitutional Court decided, that it would be against Art. 3 (3) 2 Grundgesetz (Basic Law) disadvantaging patients based on disability.⁵⁹ In November 2022 the German legislator made changes in the Infektionsschutzgesetz (Protection against Infection Act)⁶⁰. According to the new § 5c Infektionsschutzgesetz an allocation decision may be made on the basis of the current and short-term survival probability of the patients concerned, but not on the degree of infirmity, age, ethnic origin, religion or belief, gender or sexual orientation. However, age, frailty due to age or disability are overwhelmingly considered to be specific risk factors for a negative short-term prognosis of patients in intensive care. So it is questionable, how physicians should deal with these guidelines.⁶¹ Furthermore, the new regulations only apply in the context of pandemics, but not in the case of natural disasters, war or terrorist attacks.

However, also substantial criminal offences have to be considered: a physician who refuses to provide medical treatment or who abruptly ends intensive therapy to treat another patient may be punished for intentional homicide (§ 212 StGB) or causing bodily harm (§ 223 StGB) by omission (§ 13 StGB) if the patient dies as a result, suffers unnecessary pain, or suffers damage to health. While in the case of the switchman (supra II.2.1(2.2.1)), duties to act and duties to omit collide and the perpetrator is reproached for his actions; in this context, it is primarily a matter of colliding duties to act. However, an accusation of punishable inaction can only be made against the physician if he has the opportunity to provide the treatment. We can distinguish between two different cases:

Case 1: Two ventilation places are available in a hospital. Five patients arrive at the same time, including a 20-year-old vaccine refuser who further ignored all travel warnings, a vaccinated 35-year-old family man with two small children, a 55-year-old doctor who became infected by a new dangerous variant while treating COVID-19 patients, and two elderly people. All five need ventilators. The short-term survival probability of the patients cannot be accurately answered.

Case 2: All ventilation places are occupied by other patients, including an 83-year-old man with significant pre-existing conditions. Now a 35-year-old father of two small children who also needs to be ventilated urgently is admitted to the clinic. His chances of survival would be much better.

Ex ante Competition (Case 1)

Possible Grounds for Exclusion, Justification, and Excuse

In case 1, five equivalent duties to act collide, of which the physician can fulfil only two. First, Argument 1 (supra II.2.4(2.2.4)) that every human life has the same rank also applies here without compromise. In the case of conflicting duties, § 34 StGB (justifying necessity) does not apply, because the life of one patient does not outweigh that of the other (supra II.2.2(2.2.2)), nor does § 35 StGB (excusable necessity) apply unless the responsible physician acts in favour of a relative (supra II.2.3.a (2.2.3.a)). However, the nonstatutory institute of the collision of duties (Pflichtenkollision, supra II.1.2.a (2.1.2.a)) must be considered if one assumes that the respective need for treatment is equally urgent⁶² and all patients want to be ventilated by machines. Especially with older patients this is not a matter of course, because ventilation via a plastic tube inserted into the trachea can be quite painful.

Legal Classification of the Conflict of Duties as a Defence Sui Generis

According to the main academic opinion, the collision of duties (Pflichtenkollision) in German criminal law is not a mere excuse.⁶³ This is absolutely convincing. After all, impunity is not based on the perpetrator’s lack of insight or control. A mere excuse would also have the consequence that the patients or their relatives would have to be granted the right of self-defence or emergency aid (they could force the doctor to come to the aid of their relatives first).⁶⁴ Hospital violence is not a completely unrealistic scenario⁶⁵ and would ultimately block any rescue.⁶⁶

The major opinion⁶⁷ in Germany assumes that the collision of duties is a justification. In my opinion, the nonstatutory Pflichtenkollision should even exclude the objective side of the offense.⁶⁸ Because it is not the principle of predominant interest that is applied here, which is actually typical for justifications, but the principle already known in Roman law, that the impossible cannot be demanded by the legal system – impossibilium nulla est obligatio.⁶⁹ For a lawyer from a common law jurisdiction traditionally used to rather equally ranked defences, this discussion may be weird. Real differences in criminal liability – to be honest – do not result from the one or the other concept (at least at first glance). In each case, it is sufficient that the person responsible fulfils part of the conflicting obligations to be found not guilty. However, by placing the conflict within the objective elements of the offense, it is made clear that the death of three patients cannot be attributed to the doctor as “his work” – the patients died of a fatal illness rather than medical negligence or even malice.⁷⁰ Furthermore, the rescue of two patients with the sacrifice of the other three is not connected to a positive legal judgment, which seems also unreasonable. Nature acts in a morally neutral manner.

Applicability of the Collision of Duties in Case 1 (“Ex Ante Triage”)

In case 1, the physician can only provide necessary treatment to two of five patients. Then it must be up to him – regardless of the aforementioned dispute – to decide which commandment he wants to fulfil.⁷¹ The law does not offer any clear positive selection criteria for this decision.⁷²

The physician may decide on the basis of prospects for short-term clinical success (supra III.1(3.1)). However, if this question cannot be accurately answered (what is quite probable), he maybe uses other criteria, such as individual responsibility, (long-term) life expectancy,⁷³ marital status, and the social value of the profession.⁷⁴ Thus, he could put the 20-year-old vaccine refuser and returning traveller (despite his youth) in second place owing to his irresponsible behaviour,⁷⁵ even though it is debatable whether the denial of basic health and life chances is really appropriate in these cases.⁷⁶ He could give preferential treatment to the family man (because of his parental responsibilities) or to the doctor (on the principle of “save the rescuers”),⁷⁷ because it is expected that after his recovery (and immunization) he will be able to participate particularly effectively in the fight against the pandemic. He could decide by lot also.⁷⁸

As long as the criminal law cannot clearly determine for which patient the guarantor is to decide (the new § 5c Infektionsschutzgesetz does not either)⁷⁹, subsequent disapproval of the behaviour is not possible.⁸⁰ The recommendations of the professional associations, which pursue the ethically convincing goal of saving as many patients as possible with the available resources, have no legal force.⁸¹ A deviation cannot have consequences in terms of criminal law (but may do under labour law).⁸² Also the new § 5c Infektionsschutzgesetz is not meant to change the criminal assessment, at least according to the reasoning in the draft bill.⁸³ In my opinion the uncountable value of every human life (supra II.2.2 (2.2.2) and II.2.4 (2.2.4)) ensures the impunity of the physician even if he saved the life of a patient who had somewhat lower chances of survival than another patient who does not receive treatment.⁸⁴

Ex post Competition (Case 2)

Case 2 is even more difficult to assess. The recommendations of the professional associations probably assume that an ongoing treatment that is still indicated could be terminated, because the ventilation place is needed for a patient with a greater chance of survival. In any case, all patients requiring intensive care should be considered equal when prioritizing.⁸⁵ According to the guidelines of the Swiss Academy of Medical Sciences, the continuation of treatment should also be checked every 48 hours on the basis of a list of clinical criteria.⁸⁶ According to British recommendations, health professionals may be obliged to withdraw treatment from some patients to enable treatment of other patients with a higher chance of survival. They argue that there is no ethically significant difference between decisions to withhold life-sustaining treatment or to withdraw it, other clinically relevant factors being equal, although health professionals may find decisions to withdraw treatment more challenging.⁸⁷ The recommendations of the German Ethics Council are more reserved in this respect. The termination of a life-sustaining clinical procedure may be illegal, even if it follows ethical and transparent criteria.⁸⁸

Possible Reasons for Exclusion and Justification

It is questionable whether it is legal to disconnect the 83-year-old patient who is still ventilator-dependent from the device to treat subsequently the 35-year-old father of a family with two small children, because the latter has better chances of survival owing to his age and loses more years of life if he is not ventilated. The German draft for a new § 5c Infektionsschutzgesetz (Protection against Infection Act) states, that already allocated intensive care treatment capacities should be excluded from the allocation decision.⁸⁹ If one regards the separation as a positive act, then one will have to realize, analogous to Welzel’s switchman case (supra II.2.1(2.2.1)), that the active killing of a human being (§ 212 StGB) for the purpose of saving the life of another can never be justified by § 34 StGB.⁹⁰ Of course, the collision of duties also must be considered (supra III.2.2(3.2.2)). However, according to Argument 2 (supra II.2.4(2.2.4)), duties to provide assistance are generally regarded as less important than the prohibitions of causing injury. The fact that the separation of the patient from the ventilator is associated with the use of energy could be an argument against its application.

Applicability of the Collision of Duties in Case 2 (“Ex Post Triage”)

However, it would also be conceivable that the doctor's behaviour can be reinterpreted as an omission, as the focus of the legal reproach may not be switching off the device, but failure to continue treatment of the 83-year-old patient.⁹¹ In neutralizing Argument 2 (supra II.2.4(2.2.4)) by means of such a dogmatic “trick”, it does not seem impossible from the outset to achieve impunity through the institute of collision of duties (supra III.2.2(3.2.2) and III.2.3(3.2.3)). In fact, one could argue that there is no real qualitative difference from mouth-to-mouth resuscitation in the context of immediate life-saving measures (the termination would undoubtedly have to be considered an omission). In addition, a ventilator could be technically designed in such a way that it would need a positive impulse every hour to continue the ventilation activity.⁹² On the other hand, the act of withdrawing a ventilator may be connected with further actions (e.g., waking up the patient so that his respiratory reflexes can kick in). Nevertheless, the punishability of the physician cannot really depend on the rather arbitrary and situation-dependent evaluation of his behaviour as acting or omitting to act.⁹³

Decisive for the solution of Welzel’s case was probably less the distinction between doing and not doing as such, but rather Argument 3 (supra II.2.4(2.2.4)), that secured legal positions (to stand on a siding without danger) as the given status quo enjoy greater protection than mere expectances (the chance of the train passengers being rescued by changing the switch).

Here, the 83-year-old, who is already being ventilated, has such a secured legal position. With the connection to the ventilator, he has left the group of those competing for a place on the ventilator, and he or his relatives will have ceased looking for an alternative hospital, trusting in the position they have attained (comparable to a passenger who has found a place in a lifeboat after a shipwreck).⁹⁴ An intervention in this secured legal position is only permitted for the purpose of safeguarding much higher-value interests, which is not the case if the maximum value of life is affected on both sides (cf. Argument 1, supra II.2.4(2.2.4)). This thought also prevails if one assumes that the doctor will not continue the treatment in this case.⁹⁵ There is no equivalence of conflicting duties. A different decision can only be made if the condition of the 83-year-old patient receiving initial treatment has deteriorated to such an extent that further treatment appears hopeless (in which case further treatment is no longer indicated), or conversely the risk of death has been reduced to the level of a mere abstract danger to life.⁹⁶

Interim Conclusion

As a result, in case 1, the physician may decide which patients will receive treatment, if one assumes that the respective need for treatment is equally urgent. According to the criminal law, he can, but does not have to, follow the recommendations of the medical societies. It is an unsolved question, if the criteria of § 5c Infektionsschutzgesetz are binding in a criminal law sense. In my opinion, even in the case of (illegal) discriminatory or corrupt allocation decisions, offences against the person (like homicide) seem to be unsuitable for punishing such violations of the law.⁹⁷ Also in this regard the uncountable value of every human being has to be taken into consideration.

In case 2, the physician should not terminate the treatment of the 83-year-old patient (against the recommendations of the professional associations). A defence can only be considered in the case of § 35 StGB, if the physician acts in favour of a relative (supra II.2.3.a(2.2.3.a)). No excusing supra-statutory necessity (supra II.2.3.b(2.2.3.b)) has yet been explicitly recognized by the courts, and even the criteria established by the literature are not met, because it is not a matter of sacrificing a few in favour of saving many. Even an unavoidable error juris according to § 17 sentence 1 StGB (a defence in Germany)⁹⁸ will hardly be accepted,⁹⁹ because the medical societies themselves admit that the termination of intensive care measures in the context of prioritization in Germany may “come up against legal limits”.¹⁰⁰

Transferability of These Principles to Emergency Algorithms for Autonomous Vehicles

Outline of the Problem

1. The aforementioned principles can also be applied to conventional road traffic. For example, a human driver who deliberately steers his vehicle into a crowd of people instead of against a deadly obstacle that surprises him is acting illegally (in the sense of the tripartite analysis of criminal law in II.2.1, above). However, he will not be punished for manslaughter. The conflict situation of the individual to have to sacrifice himself must be taken into account (as with a shipwrecked person) on the level of culpability, namely by § 35 StGB (supra II.2.3.a(2.2.3.a)). Even if many lives could be saved, the danger to certain road users must not be averted in an inadmissible way at the expense of another road user¹⁰¹ – analogous to Welzel’s switchman case (supra II.2.1(2.2.1)).

Therefore, the driver of a fully occupied school bus is not justified in (but may be excused for) preventing a collision with a surprising obstacle that would be fatal for all passengers by intentionally (but regretfully) running over a pedestrian on the sidewalk.¹⁰² However, in most real-life dangerous situations, the human driver has hardly any time for lengthy deliberations; he will often make some decision “reflexively” – with no chance to weigh up the pros and cons. Thus, prosecution is unlikely to occur, unless the driver’s negligence caused the dilemmatic situation.

2. Many German authors¹⁰³ want to transfer the aforementioned principles (unchanged) to autonomous vehicles. Of course, an autonomous vehicle must always favour material damage over personal injury; this is not controversial. However, according to these authors, it would be illegal to program a “utilitarian” steering impulse that saves many people at the expense of an innocent pedestrian on the sidewalk. The reasoning is based on the above principle that a danger to certain road users resulting from fateful circumstances should not be averted in an inadmissible way at the expense of another road user (supra IV.1.1.(4.1.1). However, according to this opinion, even programming a “utilitarian” steering impulse would probably be culpable,¹⁰⁴ as the programmer, developer, management, and other parties are not in an acute mental predicament, when they act. Hence, it does not seem improbable that they could be prosecuted for manslaughter (§ 212 StGB). No owner, manufacturer, or programmer would wish to be exposed to such risks.¹⁰⁵

Car manufacturers could program the vehicles in such a way that in a dilemma, humans would take over. Of course, this would not be a great improvement; quite the contrary. The time would usually be insufficient for a meaningful decision. The installation of a random number generator has been suggested,¹⁰⁶ but the programming of a nondeterministic random number generator is not easy. The computer cannot generate real random numbers at all; it always requires the inclusion of external (for example, physical) processes.¹⁰⁷ Moreover, of course, the decisions are not better.

One could also try to adapt the technology according to the aforementioned academic opinion. However, this will only succeed with great difficulty. How can a machine in a dilemmatic constellation let fate take its course? After all, it was programmed beforehand for every conceivable case. The idea is probably that the autonomous vehicles should brake in an emergency event, but steer straight ahead killing the people in front of them (no matter how many). However, there is no “natural” route that the vehicle would have to take.¹⁰⁸ On a curve, for example, it is not even clear what “straight ahead” means:¹⁰⁹ Should the car continue to follow the so-called “vehicle trajectory” or should centrifugal force carry it forward into the “field”?

3. In contrast, the vast majority of the test participants in the Moral Machine allow the quantitative weighing of human lives.¹¹⁰ The majority also showed a strong tendency to save children rather than older people (although there are cultural differences in this respect) and judged according to whether the persons concerned have followed traffic rules. One could argue that at least US participants tend to orientate themselves more toward their own penal laws (II.2.2(2.2.2)). However, the continental European participants did not achieve a significantly different survey result on this point either. Is there a striking gap between the law and the moral perceptions of citizens?

Novelty of the Dilemma Problem in Autonomous Driving

The legal situation is not as clear as it seems at first glance; moreover, alternative solutions are increasingly being discussed in Germany.¹¹¹ In my opinion, the dilemma problem in the context of autonomous driving is a new one and can only be compared to a very limited extent with the scenarios discussed so far. The actual decision on how the vehicle should behave is not made at the moment of the accident or immediately before it.¹¹² Rather, it is made months or years earlier, in view of a possible (by no means safe) hazardous event with unknown participants.

Of course, one cannot shake the principle behind Argument 1 (supra II.2.4(2.2.4)) that every human life always has the same value and evades any quantification. Nevertheless, even if human life is always equal, the claim that the killing of the track workers (supra II.2.1(2.2.1)) or the passer-by on the sidewalk (supra IV.1.1 (4.1.1)) is worse than the death of many more passengers on the train or bus also requires justification.

In Welzel’s switchman Case, the difference between positive and negative duties played an important role: According to Argument 2 (supra II.2.4(2.2.4)) the duty to act (to save the lives of the train passengers) is classified as less important than the duty to refrain from killing a track worker, although on both sides the maximum value of life is affected. The human driver in an acute traffic situation may decide only to brake and leave the position of the steering wheel unchanged, letting fate take its course. However, when programming a command (braking yes/no; steering impulse yes/no), we are always dealing with an action, regardless of whether the algorithm lets the car continue to roll “straight ahead” (whatever that means; supra IV.1.2 (4.1.2)) or steers it to the right or left in the case of an impending accident.¹¹³ Similar to case 2 of the COVID-19 Triage case (supra III.1 (3.1), III.3.1 and 2 (3.3.1, 3.3.2)), the distinction between doing and not doing as such does not lead to clear results. The option to do nothing, to let fate take its course, is not left to the developer, as long as he continues his work.

In Welzel’s switchman case (supra II.2.1(2.2.1)) and case 2 of the COVID-19 Triage (supra III.3.1 (3.3.1)) problem, there was another important point: According to Argument 3 (supra II.2.4(2.2.4)), a secured legal position (to stand on a siding without danger or to be ventilated) enjoys greater protection than mere expectances (the chance of the train passengers to be rescued by changing the switch or the chance of the 35-year-old father to receive the ventilator of the 83-year-old patient). An intervention in the secured legal position would only be allowed to safeguard much higher-value interests, which is not the case if the maximum value of life is affected on both sides.

However, in the case of the autonomous vehicle, no one has a secure legal position at the time of programming. It would hardly be possible to foresee the role a certain person may find himself in a potential accident (in the role of the lonely pedestrian who is run over, but perhaps also as part of a group of people who are saved by the pre-programmed steering impulse). As a result, there are no compelling legal reasons for privileging one or the other potential victim at the time of programming (similar to case 1 of the COVID-19 Triage, supra III.2.3(3.2.3)).

Conceivable Ethical Objectives

The dilemmatic constellations should ideally be (or be able to be) resolved in such a way that is most acceptable to the population. The results of the Moral Machine are an important indicator (supra IV.1.3 (4.1.3)). The German Federal Ministry of Transport and Digital Infrastructure established a commission for the further clarification of “ethical” questions, under the leadership of the former Federal Constitutional Court judge Udo Di Fabio. The commission definitely excluded the qualification of people according to personal characteristics (age, sex, physical or mental constitution) in unavoidable accident situations.¹¹⁴ A “general programming to reduce the number of personal injuries” could “be acceptable”, whereby (consciously) it remains unclear to which situations it may refer.

In my opinion, regardless of specific participants, it is obvious that the benefits of new technology must be maximized in such a way that the greatest possible number of people have the chance to be spared. Therefore, the degree of imminent danger must also play a role, but not the gender or age of the persons concerned, and certainly not the profession or the social significance. Failure to take self-protection measures (such as the use of a bicycle helmet) could either favour or oppose the person in traffic; the increased risk of injury indicates a greater need for protection, but this might create false incentives to refrain from self-protection measures.¹¹⁵ However, in any case, one must assume that the cause of the dangerous situation (a group of drunk people running wantonly across the street) or contributory negligence by the actors must be weighted negatively. It seems unfair to impose the costs of risky misconduct on a (lonely) uninvolved person who has done nothing wrong himself.¹¹⁶ Only with respect to children – particularly endangered in road traffic (and until the 10th year of life not responsible under German civil law, § 828 II 1 BGB) – should this point not apply.

Car manufacturers will also ask themselves whether they are allowed to give vehicle occupants a preferential position.¹¹⁷ One argument for this could be that human drivers generally do not behave differently; i.e., they instinctively protect themselves and their passengers at the expense of others (supra IV.1.1 (4.1.1)). Therefore, the society does not lose anything. If purely utilitarian programmed vehicles were to compete with those that guarantee increased self-protection (even at the expense of others), the former would have little chance of selling on the free market.¹¹⁸ However, the argument against permissibility is that it is specifically the vehicle’s occupants who will benefit most from the new technology. It would appear unfair for them to pass the residual risks to outsiders or even optimize them. Of course, passive systems that mitigate the consequences of an accident for the occupants are unproblematic.

Applicability of the Collision of Duties

These aforementioned ideas (somewhere between Kant and Bentham)¹¹⁹ must of course be measured against the applicable criminal and constitutional law. In my view, the programmer of an emergency algorithm sees himself confronted by a collision of duties to omit (thus conflicting prohibitions) rather than by a duty to omit and a duty to act (from which the opposite view proceeds, supra IV.1.2 (4.1.2)). A collision of duties to omit has hardly been discussed in jurisprudence so far. This is no coincidence. Real collisions of equivalent prohibitions seldom occur in real life, because prohibitions normally do not completely exhaust the sphere of action. Some academics¹²⁰ even claim that they do not exist from a logical point of view, quoting the jurist, mathematician, and philosopher Christian Wolff (1679–1754):

“Leges prohibitivae nunquam inter se colliduntur”—prohibition laws never collide with each other.¹²¹

I would not follow that argument. In most cases, of course, we face avoidable state failure. Imagine, after a change in traffic routing, four one-way streets suddenly meet. There must be some way out for the driver, who is virtually “trapped” at the crossing. In my opinion, it is not illegal if he leaves the “trap” against the direction of travel via one of the one-way streets. Why should the perpetrator not be granted impunity and freedom of choice (supra III.2.3 (3.2.3)) when he is confronted with equal prohibitions?¹²² It is generally accepted that a legal system must not demand the impossible from anyone (supra III.2.2(3.2.2)).

Two other examples may be more appropriate in this context. In a mass panic (such as during religious pilgrimages or large entertainment events or concerts) a person who is pushed forward by a crowd of people may find himself in the situation of having to step either on A or on B, who have fallen in front of him.¹²³ It would be wrong to regard his behaviour as a deliberate, illegal attack. No matter what the offender decides, the injury to A or B remains an accident. The same must apply to an airplane crash where the pilot has only the choice to steer the plane into one (frequented) or another (less frequented) building. In a situation in which the perpetrator must actually do something, the law must not deny him every opportunity to act (whereby in my opinion the objective side of the offense should be excluded, supra III.2.2(3.2.2)); everything else would amount to an invalid “norm trap”.¹²⁴

Nevertheless, the constellation when programming emergency algorithms deviates on one point from the abovementioned cases: While the perpetrator in the crowd and the pilot must act in some way, the programmer could completely decline to cooperate in the development of vehicle algorithms. Of course, nothing would be gained by this. In 2021, there were 2.562 traffic-related deaths in Germany and 42,915 in the US; misbehaviour by human beings is considered the cause in 88.0% of the accidents.¹²⁵ The technology of autonomous cars has the potential to save a considerable number of lives.

Social Adequacy and Risk Minimization

As we have already seen, the programming of algorithms is not performed for people in actual emergency situations but with a view to a possible dangerous event with unknown participants (supra IV.2 (4.2)). Here, no one presumes to play “destiny” (or even God) for certain people, in stark contrast to Welzel’s switchman case (supra II.2.1(2.2.1)) or to some physicians in the Nazi era who allegedly wanted to prevent worse things by “sacrificing” a few patients (supra II.2.3.b (2.2.3.b)).

Loss of life on the road has always been accepted; this is not an innovation of autonomous driving. According to the doctrine of social adequacy, actions that appear to fulfil all elements of a crime but are entirely within the normal social order do not constitute injustice. The taking of risks is allowed if every average citizen agrees to the extension of freedoms in the hope of being spared from the risks.¹²⁶ In the case of negligence, the permitted risk is the flipside of a breach of due diligence. Under road traffic law, once a vehicle has been registered its distribution and operation are within this framework, even though the manufacturer and the owner can of course never rule out the possibility of an accident involving personal injury. Danger or harm resulting from these risks is not imputable to the actor despite the existing causal link between his conduct and the damage or injury.

Another question is whether this can be transferred specifically to the programming of emergency algorithms. Here we have to think in the categories of an intentional offense. In German criminal law, intention is the label used not only for cases of direct intention (dolus directus I) and almost certain knowledge (dolus directus II), but also includes cases of “conditional” intention (dolus eventualis). German law has no category of mens rea that fully equals the concept of recklessness as understood in common law.¹²⁷Dolus eventualis means that the actor considers the realization of the injury as possible and accepts it if it should occur. If the offender hopes it will not materialize, he only acts with conscious negligence (luxuria). However, even if the probability of an accident is very low, the emergency algorithms are programmed for this very case. It will be hard to deny that the developer is willing to accept that a causal chain is set in motion and causes the death of one or more people. Can one still speak of a permitted risk here?

I think it is possible, as his programming is a step to minimize the abstract danger of the car for everyone, which seems to me to be the crucial point. For the track workers (supra II.2.1(2.2.1)), the decision of the switchman at the time of his act was undoubtedly disadvantageous. On the other hand, from an ex ante point of view, in automated road traffic every single member of society benefits from more people receiving a certain amount of preferential treatment, and behaviour that conforms to the rules is rewarded (supra IV.3 (4.3)). The general reduction in the number of accidents and victims ultimately lowers the probability of becoming a victim.¹²⁸

A programmer who reduces the hitherto accepted risk of an individual becoming the victim of a traffic accident, while at the same time having no better choice as the result of exposure to conflicting duties (supra IV.4 (4.4)), does not create a legally disapproved danger. If the assessment turns out to be wrong to certain persons in retrospect, it does not matter. Wearing a seat belt can also be disadvantageous in about 1% of accidents (for example, in the event of a fire or a fall into water),¹²⁹ but in many more cases, the obligation to wear a seat belt saves lives.¹³⁰ This approach can also be justified from the perspective of constitutional law and deontological moral philosophy. The individual is not only an “object” or “tool,” as Hevelke and Nida-Rümelin correctly describe it, but also the “purpose” of such programming.¹³¹ This is the difference between the accident victim, who (unfortunately) has become the target of the algorithm, and the track workers or aircraft occupants (supra II.2.2(2.2.2)), who are purposefully sacrificed for others, as well as the healthy person asked to give up his life to save five seriously ill people (supra II.2.4(2.2.4)). Thus, the guarantee of human dignity (article 1 paragraph 1 GG) remains untouched (supra II.2.2(2.2.2)). The privileging of specific groups (old or young people, men or women), which would go hand in hand with a general increase in the risk to others, would certainly not be compatible with this approach, but it would also be undesirable from a legal policy standpoint.

Interim Conclusion

An algorithm that reduces the ex ante risk of each individual becoming the victim of an accident without creating false incentives does not impose unreasonable solidarity obligations on anyone, even if it proves to be disadvantageous for individuals in retrospect. However, the issue remains highly controversial in Germany at present. The solution proposed by opponents (supra IV.1.2 (4.1.2)), which is closely based on the classic textbook scenarios with human decision-makers, leads to results that are unsatisfactory and difficult to implement in practice. The principle that duties to help are less important than prohibitions of infringement is correct; however, it is not applicable in this context. The analogy to Welzel’s switchman case fails to recognize that at the time of the accident, the developer has no control options whatsoever, but is active in programming, regardless of whether the preliminary decision he makes leads to a new steering impulse or to remaining on the previous “vehicle trajectory”.

Conclusion

The question of how to deal with dilemma scenarios was discussed in antiquity; it fills libraries. It has also been the subject of the largest cross-cultural study of moral preferences ever since. In most of the debated cases, the death of one or more people is absolutely unavoidable. The protagonists have regularly not caused the situation themselves, but fatefully came into the conflict. Even a person who did everything right in the end will have to struggle with regrets and feelings of guilt (“moral residue”). Some researchers of course, have criticized the use of stylized dilemma scenarios, like Welzel’s switchman case or the Trolley problem, arguing that the scenarios are too extreme and unconnected to real-life moral situations.¹³²

However, the legal problem of prioritizing medical aid, which had previously also been rejected as theoretical, suddenly became real in 2020 in connection with the COVID-19 pandemic (even in the Western world). The ad hoc recommendations of the European professional associations expose doctors to unnecessary criminal liability risks, as withdrawing life support from one patient to provide it to another could be assessed as a homicide. Therefore, a clear warning from the legal professionals was urgently needed. Ethical conflicts also arise in engineering; the solution to the dilemma of who must die in an unavoidable accident seems to be one of the last unresolved legal problems of autonomous vehicles. Even though the automotive industry promises that such situations will hardly ever occur, it could prove to be a tangible obstacle to acceptance and innovation. Nevertheless, the cultural differences in judging such scenarios are not as high as they appeared at first glance. The programming of an algorithm with regard to a possible accident at some point in the future, with participants still unknown, differs significantly from Welzel’s switchman case and Foot’s trolley problem. The goal of saving as many lives as possible to minimize the chances of anyone involved in a grave accident seems to be widely accepted. In Germany, the dilemma problem is also legally solvable in a satisfactory way (between utilitarian and deontological moral principles), although emergency algorithms may lead to a certain shift in the previously accepted risk.

In the end, we may ask ourselves why philosophy professors and law teachers are so “fascinated” by such dilemmas. To be honest, it is not primarily to provide practical legal guidelines for doctors or the automotive industry, who must not be left alone with their “worries and needs.” Of course, the law-abiding citizens want and have the right to know right from wrong. However, the discussions also have an end in itself. Moral and criminal law systems must be free of contradictions. There is no legal or moral question that can remain unanswered. When weighing up life against life, our systems are put to a final test. The debated problems are understood (and can arise equally) in all countries around the world. They prompt a strong plea for a more global criminal law theory; however, this should not be limited to solving distinct stylized cases but should take into account the underlying criminal law concepts.¹³³ For example, the tripartite analysis of German criminal law (the distinction between the statutory elements of the offense, the wrongfulness and the culpability) – its strengths and weaknesses – could perfectly be exemplified by the dilemma scenarios discussed here.