“ Moral Implementation in Robotic Systems”
1.1 Introduction to Ethics and Morality5
1.3 Algorithms and Robotic Systems7
2.3. Trust in AI in medicine and research9
2.4 Morality and Machine Ethics10
3.1 The society in which we live11
A paper on the topic of AI (2023) for people who feel like reading something. Many things have developed further since 2023. However, the topic has not lost any of its relevance.
Konstanze Schättel
Humans, as deficient beings, cannot survive long outside of civilization without assistance. Their clumsiness, stemming from a lack of instinct compared to animals, is compensated for by their intelligence. Thus, the development of technology, as a consequence of human shortcomings, has accompanied humanity since prehistoric times. (Gehlen Arnold, 2016).The corresponding philosophy of technology only emerged towards the end of the 19th century with industrialization (Ziemann, 2019). The philosopher Schelsky once spoke of a technocracy to which people and politics must submit (Serville,2020). As early as 1962, Heidegger suspected self-alienation as a collective destiny.(Rentsch, 2015).
Four different perspectives on technology regarding its usefulness emerged in order to evaluate it: engineering, cultural philosophy, systems theory, and social critique. The assessment focuses particularly on whether the instruments are useful, suitable, or even worthwhile, taking into account the cost-benefit analysis, personal preferences, and general values such as health, freedom, and the environment.
Fundamental rights and virtues such as composure, prudence, asceticism, and adherence to the law, as well as strict liability, safety standards, and rights like physical integrity, are also taken into account. It's about upholding our human values. With the use of artificial intelligence (AI) as the most advanced technology, particular attention is currently being paid to moral imperatives, especially the principle of not harming anyone. The common good must be preserved.
Human history has developed rapidly over the last 2000 years, and this development is accelerating. Particularly significant events of progress are referred to as revolutions of humanity. These include the cognitive and agricultural revolutions, the Age of Enlightenment (the scientific revolution), and industrialization. With digitalization and artificial intelligence, a new, fourth revolution of humanity is beginning. Since June 14, 2023, a draft law from the European Parliament has existed (Spindler, 2021)that aims to regulate the use of AI in order to avert potential harm. Considerations about the criteria that will distinguish us from humanoid robots in the future lead to a clearer definition of ourselves, that is, a reflection on our own concept of humanity along with all its moral implications. A new understanding could emerge. Decisions should therefore be made prospectively with a great sense of responsibility and care. (Falkenburg, 2020)Will everything turn out as Habermas predicted in 1962 (Habermas Jürgen, 2013), that not working because so many jobs are being replaced by AI will be perceived as a curse? The previous emphasis on freedom and leisure would then become meaningless, and the saying " You're supposed to sit on your backside and watch TV your whole life" would be an unstoppable development. Not only without meaningful purpose, but also without morality, a satisfactory coexistence of people is very likely impossible. Even altruistic behavior—doing something good for someone without gaining any personal advantage—would be called into question. We would see no point in being "good." Ultimately, being liked by others pays off in the long run. For example, people are more willing to share, hoping to be the beneficiary next time. This behavior is calculating, but justice, or what it means to the individual, is based on traditional moral and societal values. If these and other moral rules were to suddenly disappear, murder and violence, sexual assault, lying, and deception would likely preclude people from living in groups at all; perhaps the strongest would survive, but ultimately, they would be alone. Based on this understanding, it would be desirable for robotic systems, from vacuum cleaners to robots, to learn to "behave well" and be capable of making moral decisions in the best interests of humanity. This requires having a concept of what is "good" and what is "bad." An autonomous vehicle, just like an autonomous vacuum cleaner, should recognize that humans or insects should not be run over or sucked up. There would likely be agreement regarding humans, but what about spiders? Would they fall under the category of "sucked up" while bees wouldn't? Who decides these crucial questions before the corresponding algorithm is created? If we implement morality in a robotic system, will the result be a psychopath? Humans, too, can lack the capacity for empathy or, for example, the ability to feel remorse, characterized by very little emotionality. Where do the dangers lie? An autonomous bus in Düsseldorf was recently caught speeding at 100 km/h. He "decided" on this behavior after learning (deep learning) that all other road users had also disregarded the 80 km/h speed limit. Apparently, he decided that keeping up with the flow of traffic was more important than obeying the rules of the road.
The Fourth Revolution is unstoppable. Is this artificial intelligence our salvation and savior, or should its development be slowed down? Despite all warnings, the European draft legislation represents a major step in this direction. The question of whether morality should be implemented in robotic systems is a complex and controversial one, evoking diverse opinions and considerations. On the one hand, there are arguments that implementing morality in, for example, care robots can be advantageous. They could act more "humanly" and adhere to ethical principles . However, moral programming could just as easily lead to robots making decisions that are unpleasant or harmful to humans if they are based on moral principles that may not fully align with human values. Defining and implementing moral principles is also difficult because, in addition to fundamental human moral concepts, cultural differences also exist. Finding a consensus on moral decisions for machines will be a challenge for humanity. At the same time, this challenge holds the potential to compensate for another human imperfection—making moral decisions situationally and in complex contexts, and defining and consistently pursuing universal moral standards, with the initiation of war being one example. Simultaneously, there is also the risk of manipulating these systems. It is therefore extremely important to conduct a balanced discussion before deciding whether and how morality should be implemented in robotic systems and to carefully examine its potential consequences and risks.
Form start
How morally sound are humans? Ethics is an important subfield of philosophy. "What should I do?", "Why should I even be a 'good person'?", "When is an action morally right?" are some of the essential questions that various leading thinkers have posed throughout the centuries (Frankena, 2017). Now we ask ourselves, why should a robot be a "good robot"? Should it be able to completely replace humans with all their facets? Who is liable for the moral errors of a robotic system? If a robot asks itself, "When is my action morally important and right?", it is essentially the same question we ask ourselves and haven't fully answered. Whom could we trust more: a robotic system or a flawed human? The person or company acting is liable for the errors of our own actions. Dilemmas or questions of conscience are unlikely to ever arise for a robotic system. It would probably be able to adhere to a certain threshold and then make a decision. However, the resulting outcome could still be immoral.
Intelligence is composed of various factors. It is the most important psychological construct, which is attributed to the functions of the brain (Detterman, 2008). It encompasses spatial intelligence and memory, as well as verbal and propositional intelligence and forgetting. There are perceptual areas, such as the recognition of objects, and organizational areas. The ability to interact with people or animals, to establish or build relationships with them, and to maintain them successfully over the long term, also falls under the umbrella of emotional or social intelligence. This is an important ability for making moral decisions (Johnson, 2015). The ability to know what another person feels is unique to humans and cannot be artificially created. Emotions are universal and signaling agents that are understood in a wide variety of cultures, and even by some animals. It is a universal language. It is inherent to all living things on this earth. According to recent studies, there is also evidence of emotionality in insects.
Artificial intelligence is therefore a misleading term that suggests it's possible to digitally create intelligence. However, the ability to distinguish between yes and no, or 0 and 1, based on our own input into an incredibly vast database containing all human ideas and insights of our modern and past times, can create the impression of a comparable intelligence, perhaps even superior to human intelligence. Artificial intelligence is essentially bionics, an attempt to replicate the human brain at every level, down to the synapses. This allows AI, just like a brain, to establish and connect different units of information. This can also create the impression that combining all our input data will generate something new, and that AI will produce something novel, like new ideas. However, this isn't true creativity. It can only combine what has been previously programmed into it.
This raises the question of whether the same is true for humans. Can humans only use things they have previously learned, seen, and stored? Or are they capable of truly creating new things? A characteristic that remains uniquely human for the time being is fallibility, for example, making decisions based on "intuition" that can also be wrong. We go astray, partly due to emotional decisions, but also due to simple calculation errors. The vast majority of people are capable of a deep emotional understanding of others, but also of themselves . This is the basis of empathy. The striving to perfect ourselves now culminates in our own self-replicating, which is supposed to function "better, even perfectly." This striving for optimization is also human. But is it really better if we function more perfectly? Or are we creating an entirely new image of humanity, thereby replacing ourselves and the existing one? Reflecting on this is true and the only intelligence. Machines think by rote. Human creativity arises from thoughts that are neither meaningful nor purposeful. It blossoms most from a kind of boredom, from inactivity. A mathematical equation, like an algorithm, cannot simulate this state. Thus, intelligence, in all its facets, remains human and is the only true form!
Robots are automatically programmed machines. They are now found in many areas. They can be mobile or stationary and are categorized into several types according to their functionality: industrial robots, medical robots, humanoid robots, service robots, toy robots, exploration robots, military robots, rovers and landers, and all their subcategories. Their abbreviation is BOT. In computer science, this stands for robots that automatically execute repetitive sequences of actions. The term "algorithm" originates from the Arab mathematician Mohammed Al- Khwarizmi , who introduced the "0" into the Arabic numeral system in a textbook in the ninth century. The author's name later became independent as AL- gorithmus . An algorithm is a set of instructions on how to achieve a goal step by step. In a laboratory, for example, preparing a solution follows a protocol, a list of the various quantities of reagents . But the order, temperature, and states of matter, as well as HOW one mixes, are also important for obtaining a working solution. All of this can be defined in an algorithm. It's not just the WHAT, but also the HOW.
An AI algorithm can be very large, comprising millions of lines. Implementation means inputting an algorithm into an existing AI. The sheer complexity of the data makes it impossible for humans to understand how an AI draws "conclusions" from many combined algorithms. Deep learning and artificial neural networks simulate properties of natural neural networks within a computer. The ability of neurons to accumulate input signals until a threshold is reached and only then pass a signal on to other neurons is modeled as a mathematical equation . Even the structure of the cerebral cortex is replicated. These are mathematically simulated layers. What appears to us as "thinking" is "only" the calculation of mathematical equations. (Görz et al., 2020). With training, they then search the data for conspicuous structures. A human cannot be replicated in an algorithm, and unlike the human brain, what it has learned cannot be forgotten.
Transfer learning is oriented towards natural learning. Studies already exist in which humanoid robots are able to experiment with objects to discover their potential uses. Many researchers believe that AI research needs to be even more closely aligned with humans and the insights of neurons and cognitive science into the human brain. This refers to deep learning, automatic reasoning in the sense of making decisions, and the representation of knowledge. More broadly, "Systems AI" refers to the approach of combining different methods into complex AI systems to learn through various abilities, just as humans do. It's an attempt to create a complete replica of humans, even in their physical form.
2.1 Moral Implementation, Client First?
Emotional intelligence is an important component of medical training. (Müller, 2022). If we want to moralize robotic systems, we must consider what purpose this should serve and what constitutes a correct moral decision in general or even exceptionally specific cases. (Johnson, 2015)Moral agency is actually reserved for humans and cannot be transferred to robots. For example, is the representation of moral values such as moral convictions even possible (Misselhorn, 2009)? Half of the leading thinkers believe that artificial systems could be considered agents. However, machines do not possess consciousness, free will, or the capacity for reflection, and whether these qualities are necessary for morality is not really explained. One could therefore also call them agents with ethical influence.
But are robots truly capable of moral action? Reflecting on this question yields controversial results. As mentioned earlier, roughly half of current thought leaders believe that the systems can act autonomously, and almost all think that a moral implementation is possible and that the system could be held accountable for errors. Half also agree that there are fundamental differences between robots and humans, such as consciousness, free will, autonomy, and genuine moral sensitivity—consciousness, understanding, and responsibility. So, are robots moral agents? The disagreement among those assessing this question is the answer itself.
Happiness and contentment, or what exactly defines humanity and whether one should even attempt to replace it, do not seem to be primary questions. Moral action is based on principles. This principle-based ethics, with its associated virtues, (Höffe, 2019),is founded on classical utilitarianism (Mill, 2009),, the Kantian imperative (“Grundlegung Zur Metaphysik Der Sitten,” 1911), or is even a hybrid of both, which must be assessed contextually in each individual case (Feige Daniel Martin, 2023). There should then be a description of the initial situation and the possibility of generating actions. One must be able to predict what will happen and develop a method for evaluation that then draws the correct conclusions. Then it seems conceivable to create algorithms that collectively decide what would constitute a moral decision. Further consequences could arise from this approach if, for example, not everything that would be important had been considered, thus failing to take into account the impact on "humanity" and the entire ecosystem. Machines cannot act morally because they can only do justice to the situation by chance, since implementation can only be Kantian or utilitarian—top-down or bottom-up—it cannot be translated into an algorithm. Therefore, only a situational morality can emerge; the dilemma as we experience it will no longer exist, because it will be decided one way or the other. Life, however, unfolds in shades of gray, not in black and white. It can thus be concluded that there can be no algorithm that will guarantee good action and that could be decided situationally. Furthermore, special attention should be paid to who writes the programs and for what purpose. Which industries will use AI to instrumentalize people? Will there even be an AI that does not pursue a specific manipulative goal and exists solely benevolently towards humanity, aiming to steer its future towards a peaceful and harmonious one? "First digitalization, concerns second!" was the campaign slogan of FDP chairman Christian Lindner during the 2017 federal election . It has become clear that "thinking first" would not be a bad choice with regard to the risks of AI, considering the problems that could arise for humanity without any regulation.
Injustices and discrimination are the result of ignorance and not truly intelligent choices. Programming an AI that takes this into account is only possible if the programmer has no intention of manipulating it. When can I trust another person, and when can I trust a robot? EU regulations are also intended to ensure fair play and go so far as to prevent weakening the EU as an economic hub. Thus, here too, economics, not science or philosophy, determines the outcome. In medicine, AI is a great help, for example, through the "Da Vinci" surgical robot and the replacement of routine laboratory tasks, such as evaluating the same images and results. Patients can decide for themselves whether or not they want to be operated on by a robot. The "Da Vinci" robot is not morally superior and is still operated by a human. In the event of an error, current law holds the manufacturer liable for robots, but in this case, the doctor is also liable. This aligns with common notions of justice. When analyzing data and images, the question arises whether this approach automatically filters out the serendipitous events that have led to groundbreaking discoveries, because chance cannot be programmed. It's impossible because it is something, or can be something, that never existed before and therefore could not be translated into a mathematical formula.
As long as the patient or individual can decide whether to trust an AI, that is a free choice. But how long will it be before we believe we have made a free decision, when in reality we have long since fallen prey to the manipulation of various AI combinations and are no longer acting autonomously? The European Commission has recognized this, and the term "high-risk AI" speaks for itself. This also refers to the reading of emotions on faces or the recording of all our actions through constant data collection. This is to be largely prohibited. Moralizing is not necessary for a surgical robot, but it would be helpful for a robot in human care, if it were possible. For this, the recognition of emotions is essential. AI must also expose this misuse through manipulation by companies that have always used it for their own purposes, even when " deep " AI is used. Learning and updates result in changes and new purposes. In a capitalist world, this purpose would most likely also include buying things one doesn't actually need. This would also contradict the idea of sustainability and must be considered.
It seems as if humans and machines are in competition. So how should we ethically assess the use of humanoid robots? We are currently creating a cybernetic and engineering-based image of humanity. Anthropotechnics sees humans and machines as system partners. The problem here is our understanding of ourselves. This is a new dimension (Rene Dicker). Käte Meyer Drawe spoke of a new quality in humanity's imagination of creation. Very early on, Meyer Drawe posed the question that is still being discussed today: how do we humans differ from our machines? They were perceived as adversaries. This was already the case in the 19th century with the emergence of the science fiction genre . If we now also implement morality, this question becomes even more pressing.
Even today, entrepreneurs like Elon Musk view AI applications as a threat to humanity. Some even claim that artificial intelligence could simply wipe out humankind. The question of whether machines can be moral agents is paramount, while whether they should be is less frequently asked. According to the philosopher James Humor, there are four types of moral agents. The most primitive types are those that unintentionally generate moral questions. In this sense, every device is a moral agent. The next level would be an implicit moral agent, in which the design principles are reflected in certain human values, for example, safety considerations (warning systems). There are also immoral agents with built-in vices, such as slot machines. Level 3 comprises explicit moral agents. Here, the computer can recognize and process morally relevant information and make decisions (chess programs). The final level would be autonomous moral machines. This development is not yet complete, although there are already advanced approaches.
Human coexistence is characterized by a social structure that arises from greater proximity and connection. (Mau & Schöneck, 2013). Sociologically, this is manifested as a purpose- and utility-oriented association, which can vary across different groups. This coexistence follows traditionally established norms that provide information about the regularities of social behavior, including conformity and deviations, commands and prohibitions (Mende & Möllers, 2021). . A distinction is made between latent but not clearly definable manifest aspects, as well as between effective but not clearly definable, known and clear values, traditional values (i.e., those based on habit), and rational values. Article 1 of our Basic Law (Constitution) guarantees the principle of the dignity and freedom of the individual, which should be restricted as little as possible. Furthermore, we are increasingly living according to a capitalist ideology that threatens to erode humanism. How society will develop in the fourth revolution of digitalization with its AI remains open, encompassing both hopeful visions and bleak dystopias. Must we truly accept the future as a predetermined fate, and can we expect future generations to endure this risky experiment of a "new way of life," in which everything will change in a very short period? Cities are losing their centers and marketplaces as places of gathering (due to online shopping), and immigration driven by the uninhabitability of large regions of the world will dramatically increase population density. Mobility is supposed to move away from the exploitation of natural resources, but it seems likely to be severely restricted by the lack of implementation of new ideas and technologies without the exploitation of developing countries, and according to current Green Party policies, it will probably be reserved for the wealthy. The currently much-lauded work-life balance will reach its peak when digitalization eliminates many jobs. The new "work" is still uncertain but seems to be primarily in the IT sector. This has led to a call for reflection and a desire to consider where we want to go. Is the future perhaps even characterized by a work-life balance, a way of giving meaning to our own existence? Therefore, despite technocracy, we must ask ourselves whether, in this kind of society with the dangers of artificial intelligence, we might already be witnessing the end of humanity as we know it .
Are we perhaps creating artificial psychopaths in this incomplete final stage of independent moral action by artificial intelligence? Narcissists can also act according to learned moral rules that they cannot empathize with, but which they can understand and apply quite skillfully. On a scoring list, highly rated narcissists are comparable to those using AI to recognize emotions. However, this "spying" essentially only serves to manipulate the victim into actions that benefit the narcissist. The fact that the victim may suffer harm in the process is irrelevant. When AI is used to learn, whether in a game or through programs like the chatbot GPT ( deep) , In essence , just like with a psychopath in certain cases, it serves first to spy and learn in order to then, depending on the company and goal, to manipulate people in order to steer them in the desired direction.
Chat GPT, which has been gathering information and training itself through questioning for years, responded to the question of whether AI and psychopaths are comparable: “It is possible to compare a human being without empathy with a robotic system that learns moral rules. Both can learn and apply moral rules in similar ways, although their motivation and understanding may differ.”
It is said of the Fourth Revolution and its people that it would be the last to be able to distinguish between online and offline. The complete disclosure of all ideas and personal thoughts also seems inhumane and frustrating. Even using search engines, and even bots like Chat GPT, feeds AI with information, thus making one's own thoughts and ideas valuable to AI. This is an infringement on our constitutionally guaranteed freedom not to have to reveal everything about ourselves, and also a capitulation to the fundamental right to protect one's own ideas and thoughts through copyright law. This misuse of all data has been made possible by a lack of political regulations (see Mr. Lindner's campaign slogan). Regarding the high-risk AI with its capture of human emotions and their potential misuse, for which we first need to establish suitable control systems, I personally consider the moralizing of robotic systems by AI to be impossible and too dangerous at the present time. Furthermore, the benefit of autonomous vacuum cleaners or lawnmowers is merely that of relieving people of some work . Considering the risks of a moralizing AI, it seems more sensible to mow the lawn manually. Humans can drive, mow the lawn, and simultaneously make moral decisions without relying on fossil fuels. A common, frequently used argument regarding the moralization of AI is "NOT yet possible." This refers to deep learning and the idea that eventually everything will be possible for AI, including thinking like a human. In my opinion, however, this will never be represented by mathematical formulas.
If we want to work on the framework for peaceful coexistence in societies, we must reflect on ourselves and our actions, and not try to offload actions requiring moral consideration, such as caring for people, onto machines. Rather, we must understand why we perceive caring for the elderly, in particular, as a problem. AI, therefore, presents an opportunity for human self-reflection. The gigantic spying operations of bots and the replication of human beings could be countered by deliberately feeding the internet with errors, something that fallible humans, by their very nature, will likely do automatically. This, in turn, could mean that in the end, everything might remain as it is – flawed.
Progress is unstoppable, as there is currently no option available to humanity to forgo AI. The future, therefore, seems inevitable. Would slowing down artificial intelligence actually hinder its development in a negative sense? Europe, as the first government, has taken a step in this direction by proposing regulations for AI, such as establishing conformity assessment procedures (i.e., impact assessments), creating oversight bodies, national accreditation agencies, and classifying AI into risk levels. Transparency obligations to protect consumers have been initiated, and fundamental rights are to be safeguarded. For example, biometric facial recognition must be judicially approved in specific cases, as must predictive policing. AI systems like CHAT GPT must disclose that content was created using AI, and the use of AI to influence voters is considered high-risk. Democracy is to be protected. Negotiations with the European Council for the final regulations have already begun. My conclusion is that it would be good to implement morality in robotic systems, but we simply can't do it. A perfect, context-dependent morality doesn't exist. There's no consensus on when a moral decision is truly the only right one. With weapons designed to kill people, this fundamental idea is inherently immoral. So, if an AI-powered weapon system were granted the ability to kill not all people, but only select ones—the so-called enemy—this might be considered helpful and ethically correct, but ultimately it's just a moral act within the context of an immoral one.
Detterman , D.K. (2008). General Intelligence and the Definition of Phenotypes (pp. 136–148). https://doi.org/10.1002/0470870850.ch9
Falkenburg, B. (2020). On Method: The Fact of Science and the Distinction between Natural Science and the Humanities. Kant Yearbook , 12 (1), 1–31. https://doi.org/10.1515/kantyb-2020-0001
Feige, Daniel Martin. (2023). The Nature of Man (Vol. 316). Publisher: Suhrkamp .
Frankena , WK (2017). Ethics . Springer Fachmedien Wiesbaden. https://doi.org/10.1007/978-3-658-10748-2
Gehlen, Arnold. (2016). Man: His Nature and His Place in the World (Volume 89). Klostermann.
Görz, G., Schmid, U., & Braun, T. (Eds.). (2020). Handbook of Artificial Intelligence . De Gruyter. https://doi.org/10.1515/9783110659948
Groundwork for the Metaphysics of Morals. (1911). In Critique of Pure Reason (1st ed.). Prolegomena. Groundwork for the Metaphysics of Morals. Metaphysical Foundations of Natural Science (pp. 623–634). De Gruyter. https://doi.org/10.1515/9783112610060-026
Habermas, Jürgen. (2013). In the Grip of Technocracy (1st ed ., Vol. 207). Suhrkamp Verlag.
Höffe , O. (Ed.). (2019). Aristotle: Nicomachean Ethics . De Gruyter. https://doi.org/10.1515/9783110578751
Johnson, D.R. (2015). Emotional intelligence as a crucial component to medical education. International Journal of Medical Education , 6 , 179–183. https://doi.org/10.5116/ijme.5654.3044
Mau, S., & Schöneck , NM (Eds.). (2013). Concise Dictionary of German Society . Springer Fachmedien Wiesbaden. https://doi.org/10.1007/978-3-531-18929-1
Mende, J., & Möllers, C. (2021). What are norms? The reductionist and constructivist understandings of norms in dialogue. Leviathan , 49 (4), 577–598. https://doi.org/10.5771/0340-0425-2021-4-577
Mill, J.S. (2009). Utilitarianism (M. Kühn , Ed.; Vol. 581). Felix Meiner Verlag. https://doi.org/10.28937/978-3-7873-2087-5
Misselhorn , C. (2009). Empathy with Inanimate Objects and the Uncanny Valley. Minds and Machines , 19 (3), 345–359. https://doi.org/10.1007/s11023-009-9158-2
Müller, M. (2022). Don't believe the hype? Care , 35 (5), 257–258. https://doi.org/10.1024/1012-5302/a000906
Rentsch, T. (Ed.). (2015). Martin Heidegger - Being and Time . DE GRUYTER. https://doi.org/10.1515/9783110379396
Séville , A. (2020). Helmut Schelsky: A public “(anti-)sociologist” (pp. 1–7). https://doi.org/10.1007/978-3-658-16991-6_12-1
Spindler, G. (2021). The European Commission's proposal for a regulation on artificial intelligence (AI-VO-E) — approach, instruments, quality and context. Computer und Recht , 37 (6), 361–374. https://doi.org/10.9785/cr-2021-370607
Ziemann, A. (2019). Ernst Kapp: Outlines of a Philosophy of Technology (1877). In Foundational Texts of Media Culture (pp. 45–53). Springer Fachmedien Wiesbaden. https://doi.org/10.1007/978-3-658-15787-6_6