There are reasons why digital transformation is tantamount to a “slow” revolution for psychiatry. Brunn et al. were able to identify challenges that have an influence on the integration of AI applications: skeptical attitudes of psychiatrists toward AI, potential obsolescence of psychiatrists, and potential loss of definitional authority through AI . User acceptance has a pivotal impact on the implementation of AI. Furthermore, technologies reflect and influence social structures—e.g. they shape communication and interpersonal relationships . At the same time, they can give rise to developments that in turn can become drivers of insecurities and pathologies . This continues to raise questions about the interdependence of technologies and society, thus also for psychiatry and which changes it will be subject to . A survey of psychiatrists on the impact of AI and machine learning (ML) addressed this question. The study found that one in two psychiatrists predicted that their professional field will change significantly in the future.The majority of respondents do not believe that AI/ML could or will ever replace their work as psychiatrists, but that time-consuming work (e.g. documenting) will be transferred to AI/ML systems .
In addition to the physical, psychiatry focuses on the psyche and the brain of humans. In diagnostics and therapy, it thus faces the challenge of identifying and taking into account factors that ultimately influence the human psyche and brain. This is a challenge that has not been met primarily by technology. Nonetheless, or precisely because psychiatry is directly intertwined with the social matrix, AI-powered technology has found its way into it. In particular, this has been spurred by phenomena that psychiatry has faced in recent years, leading to calls for supportive or transformative technologies; e.g. regarding the COVID-19 pandemic, natural disasters, or war conflicts . Digitalization has a stake in the crisis-ridden social matrix and at the same time embodies the tool as part of the coping process. This has led to increased engagement in research and clinical implementation of innovative technologies, which come with challenges, e.g. regarding research ethics standards such as transparency or reproducibility of information .
The emergence of technological innovations has thus triggered a dynamic in medicine in which the assessment of the use of such systems constantly oscillates between opposites: opportunities and hope on the one hand, risks and skepticism on the other . A conclusive evaluation, e.g. with regard to possible risks or benefits, is often not possible, but rather a constant evaluation of the technology used is required. This is essential due to the rapid technological progress, which has also led to an acceleration and complexity of knowledge in medicine: today, medicine has a half-life of about 1–2 years, in the future it will certainly be even shorter .
Ethical considerations on AI
Technological upheavals, such as the introduction of AI in society, simultaneously generate ethical challenges, which the European Union addressed in 2019 by introducing general ethical guidelines for the development, deployment and use of AI : “Its central concern is to identify how AI can advance or raise concerns to the good life of individuals, whether in terms of quality of life, or human autonomy and freedom necessary for a democratic society” . These guidelines concern the society as a whole, which is why they are formulated in an open manner, with the indication that they can be adapted and evaluated depending on the scope of application of AI. In addition to fundamental rights (such as respect for human dignity), the guidelines specify four non-hierarchical ethical principles that should be considered: respect for human autonomy, prevention of harm, fairness, and explicability . These principles, which serve to protect humans interacting with AI, reflect ethical values that are also relevant in medicine when dealing with patients and can be found in the “principlism” established by Beauchamp and Childress. Their principles include (1) respect for human autonomy, (2) nonmaleficence, (3) beneficence, and (4) justice . Unlike Beauchamp and Childress, the European Commission uses the principles mentioned to be taken into account as fixed values and not to be weighed against each other. When considering and evaluating AI applications in psychiatry, it makes sense to consult not only general but also medical-specific ethical guidelines—especially when AI is used with vulnerable groups. To this end, various ethical criteria in the use of technology in medicine provide guidance for conducting an ethical evaluation, e.g. with regard to self-determination, safety, privacy, or fairness [15, 16].
Ethical challenges in psychiatry: doctor–patient interaction and AI
But what ethical challenges arise from the use of AI in psychiatry? This question aims at the ethical acceptability of using AI systems in this context. Various stakeholders (such as patients, relatives, or medical, nursing, and technical staff) involved in psychiatry and in the digitization process play a role. To illustrate the changes in the interpersonal interaction of these stakeholders, the doctor–patient relationship is considered as an example.
Physicians have always had sovereign power in medical diagnosis and treatment. This expertise will undoubtedly be strengthened by the use of AI-based systems for the time being in terms of optimization. Thus, it is already possible to provide more objectified and more complex diagnostics as well as personalized prognosis —for example, referring to biomarkers (e.g. clinical, imaging, genetics), psycho-markers (e.g. personality traits, cognitive functioning), and social markers (type of social media use) in classifying certain mental disorders [17,18,19]. In the near future, psychiatrists will consciously and transparently shape their mediating role between the AI-generated expertise and the ethical decision-making process in the sense of patient autonomy.
In recent years, patients have matured into medical “lay experts” who use digital tools and the Internet in particular to acquire knowledge and derive actions or treatments. For example, AI-powered apps that are easily accessible to smartphone users expand patient empowerment in this regard and shape trust by making physicians’ actions verifiable . How the free will to decide can be guaranteed, however, remains a central topic of the situational as well as the developing doctor–patient relationship. Not only physicians and patients grow and learn, but also ML or even Deep Learning (DL) are trainable technologies and, like humans, must be continuously subjected to the learning process . As a consequence, this can also improve the interaction and trust relationship with AI.
For psychiatry, various ethical challenges (Table 1) to which the doctor–patient relationship is subject arise or intensify not only in the areas of prevention, diagnosis/prognosis, and therapy, but also in the areas of education and research.