“The use of AI must foster human flourishing, not diminish it. AI must not be used to replace people. Those are baseline rules for its ethical evaluation,” states German Ethics Council Chair Alena Buyx.
AI systems have found their way today into practically every field of public and private life, whether used for cancer diagnosis in the medical field, intelligent tutoring systems in education, recommendation systems in online platforms, or software intended to assist decision-making in the social welfare and criminal justice systems or the police.
“Artificial intelligence cannot substitute human intelligence, responsibility, or judgment,” stresses Julian Nida-Rümelin, Vice-Chair of the German Ethics Council and also the Vice-Chair for the Council’s working group on humans and machines. This is the conclusion that arises in the Council’s Opinion after considering intelligence, reason, action, and responsibility – key philosophical and anthropological categories in the relationship between humans and machines.
This is important for the ethical evaluation of AI, as it is not enough simply to understand the technologies; the complex interactions between people and technology and the overall impact on society must also be taken into account. It also explains what the Ethics Council considers to be the central question when making an ethical assessment of AI: Will human authorship or the conditions for responsible action be expanded or diminished with the use of AI?
Guided by this question, the German Ethics Council considers AI’s usage within four fields chosen by way of example: medicine, education in schools, public communication and opinion-making, and public administration. In doing so, the Opinion shows that any assessment of AI must remain specific to its context, intended application, and the people involved. “When human activities are delegated to machines, it can impact different groups of people, actors, and affected parties in entirely different ways,” says Judith Simon, Chair of the working group. “This makes it important to pinpoint those for whom [the use of AI] is linked to an expanded field of action, and those whose opportunities may be diminished.”
This appeal finds reflection in the AI usage recommendations that the German Ethics Council develops for each of the four areas considered in the Opinion. In the field of medicine, the Council’s recommendations look, among other goals, to assure quality in the development and application of AI products, avoid a loss in doctors’ professional competence, and reconcile patients’ right to privacy with intensive data use for medical research.
Throughout, it is essential to preserve trust between all involved parties and avoid the complete replacement of human medical professionals.
The implementation of AI in school education should not be guided by technological visions, the Ethics Council advises, but rather take its cues from foundational concepts of education, and remain restricted to elements that demonstrably expand learners’ skillsets and social interactions, protect their privacy, and foster their development as people.
In the area of public communication and opinion-making, recommendations include continuing to develop rules for online platforms regarding content selection and moderation, as well as for personalised advertising and data trading. The Council also calls for improved access to platform data for research purposes and recommends considering the establishment of a public digital communications infrastructure.
In public administration, the Ethics Council advises approaches to AI that guard against discrimination and prevent the blind following of machine recommendations. It further insists on guaranteeing individual case examinations as well as inspection and objection rights for affected persons. The application of AI by law enforcement agencies should be subject to a process of social deliberation regarding the appropriate balance of risks and opportunities associated with such methods.
In addition, the German Ethics Council identifies ten overarching topics and recommendations that cut across subjects in their importance for the ethical assessment of AI. They include the broader notion of using AI to support human decision-making rather than replace it; preventing the diffusion of responsibility; not impairing options for human control; and guaranteeing access to the basis for machine decisions, especially in cases of greater intervention. Further recommendations seek to avoid bias and prevent people from becoming dependent on or misusing technology, or losing important skills and capacities. Across all areas of application, it is essential to keep the interests of the people whose data is being used by AI front and center and prevent excessive intrusions into people’s privacy by means of effective legal and technological precautions, while at the same time allowing for data usage that is oriented toward the common good.
The Opinion is available in German on the German Ethics Council website under www.ethikrat.org/fileadmin/Publikationen/Stellungnahmen/deutsch/stellungnahme-mensch-und-maschine.pdf. An English translation will be available in due course.