Programme

The multidimensional effects of the COVID-19 pandemic and the post-pandemic scenario highlights the need to adopt a systemic vision and an increasing technologization of health system management. The convergence of Artificial Intelligence (AI) and Healthcare is giving life to a transformative era in Public Health practices as well as Medical Intelligence processes and decision-making in Medicine. In such a disrupting scenario, the Malicious Use of Artificial Intelligence (MUAI) represents a vital threat for patient safety, public trust, and the overall integrity of healthcare systems. This also implies important impacts in terms of both public and national security. Artificial Intellience, as involuntary distorsions and biased Machine Learning (ML) as well as the manipulation of AI algorithms, leading to skewed medical insights, misdiagnoses, and suboptimal and/or inadequate treatment plans. The consequences of such malicious interventions extend beyond individual patient harm, permeating the foundation of medical intelligence and challenging the reliability of AI-driven decision support systems. Furthermore, the compromised security of medical data poses a tangible risk to patient privacy, potentially leading to unauthorized access, identity theft, or the dissemination of sensitive health information to target people.

Therefore, Public Health Security, increasingly reliant on AI for predictive modeling, outbreak detection, and resource allocation, faces entirely new risks. Adversarial actors such as Organized Crime Groups), Violent Extremists, Terrorists, and hostile foreign actors can exploit multidimensional vulnerabilities to attack, interfere and influence also the accuracy of epidemiological predictions for criminal purposes. Such one-shot and/or long-term operation can seriously compromise the effectiveness of disease control measures, ultimately impacting the health and well-being of both entire populations and specific communities.

The presentation underscores the imperative for a proactive approach to mitigate the evolving threats posed by MUAI in Medical Intelligence, Public Health, and decision-making in Medicine, as essential pillars of National Security domain.

Internet and hyper-connectivity have radically transformed communication among individuals, facilitating new channels of dissemination and methods to access information. With Artificial Intelligence, we are experiencing a second revolution, with a technology that will transform sectors such as healthcare, financial services, and manufacturing in the coming years. However, these tools can be used by numerous attackers with malicious objectives. This will frame the antisocial drifts of these means, explaining the methods, types, and most frequent behaviors carried out by criminal actors: from the abuse of generative AI in creating more sophisticated phishing attacks to impersonation attacks involving violations of online privacy and personal data. Finally, possible prevention and counteraction interventions to address these dynamics will be explained.

The wrongful prosecution and miscarriage of justice, known in Chinese legal community by the term of “unjust, false and wrong cases”, have always been an important social issue in China. In recent years, with the judiciary’s resolution to curb miscarriage of justice, an increasing number of old cases are rediscovered, retried and overturned. It has brought the attention of citizens and legal scholars, again, to a long-standing question: How did miscarriage of justice happen? This paper aims to synthesize the research on the psychological mechanism of miscarriage of justice, and then integrate technologies from the perspective applicable to the Chinese context to prevent worongful prosecution. We first analyzes the concept and typology. Then, through the systematic literature review based on articles published in the CNKI database, combined with in-depth case analysis, we identified a few factors as the major predictors of miscarriage of justice, most significant among which are investigators and judicial authentication personnel contextual bias and subjectivity bias. In addition, decision-making psychological mechanism are also examined in those cases. Moreover, we also identified the influence of traditional cultural factors, judicial procedure as well as legal institutions on these psychological process. This review provides useful insights into the prevention of miscarriage of justice from a systemic and institutional point of view.

Over the last few decades, the development of behavioral neurosciences has fostered the study of the biological correlates that subtend the mental processes involved in moral choices and social behavior. For the sake of simplicity, human behavior can be considered as the resultant of instinctual/emotional factors on one hand and critical reasoning on the other. When this balance becomes defective, either because of emotional dysfunction and/or impaired critical ability, individuals may express abnormal behaviors. While Free Will is undoubtely a widely debated and complex concept, in the forensic context it can be defined operationally as the ability (to decide) not to act automatically in response to an impulse. Overall, responsibility requires the integrity of multiple emotional, cognitive and behavioral abilities. Individuals with lesions of the anterior portion of the brain, including patients with the behavioral variant of frontotemporal dementia (bvFTD), offer a valid model to investigate the neural bases underlying responsibility and behavior. These patients invariably show a progressive loss of the ability to control their instincts and impulses, to modulate their behavior, to respect moral laws, along with affective flattening. We argue that the co-occurence of peculiar genetic, neurobiological and environmental factors may result in a condition of Functional Frontal Fragility, even in individuals who do not present, and may never present, any bvFTD. Indeed, allelic variants of gene involved in neurotramsmission, by modulating individual susceptibility to childhood aversive environmental factors, may increase their likelyhood to develop behavioral dyscontrol as adults. Moreover, morphological and functional abnormalities often are found in frontal and limbic areas of antisocial individuals. Recently, we showed also that inhibition of automatic behavioral responses - that is, the ability to do otherwise - is compromised after prolonged sleep deprivation due to local sleep phenomena in frontal cortical areas.

In view of evidence coming from neuroscience, the question of the extent to which individuals are free and responsible of their actions has found renewed vigor. The issue reconnects to the millenary debate in the ethical and philosophical realm on free will versus determinism — a debate whose echo resonates in the courtroom. The capability to distinguish good from bad and to decide to act in one way or another is the foundation of the criminal justice system. Indeed, on such a foundation, retributive jurisprudence, typical of all modern societies, bases culpability and imputability.

Nowadays, there are already several legal systems that have implemented the use of Artificial Intelligence (AI) in their judiciary systems. In Portugal, about 25% of the courts use systems to aid the decision-making process of the human judges, but never replacing them. However, there are some legal systems that have already done so or are taking huge steps in this direction.

AI can be defined as a branch of computer sciences that, through the creation of algorithms, allows a software agent to receive external stimuli – data – and respond to a situation, solving a certain problem. Although we are using the designation of “Robot-Judge” for practical reasons, it is not accurate, as it is merely a bot, a software agent with no corporeal form.

Regardless, the implementation of new technologies, such as this, can have a tremendous impact. The use of AI can speed up the judicial process; at the limit, it can even replace the judge, the translator, the clerk, etc., substantially reducing the waiting time for a judgement and the costs of justice. However, these systems can raise a series of issues. We intend to look at some examples of these systems to see if they ensure respect for the fundamental rights of the citizens: the right to an adversarial hearing, the right to know the ratio decidendi, the right to appeal a judgement, etc. In order words, we intend to verify if these systems can guarantee the respect for the principle of human dignity.

The determination of criminal responsibility and the evaluation of social dangerousness in the criminal sentencing of offenders with mental disorders are crucial and complex issues. Forensic psychiatric evaluation practices and regulations vary significantly across different jurisdictions, offering an opportunity for comparative analysis. This paper compares the forensic psychiatric evaluation processes in China and Italy, examining the evaluation of criminal responsibility from both legislative and judicial perspectives and analyzing the evaluation of social dangerousness in terms of legal concepts, evaluation criteria, and treatment measures. The shared language, the potential for harmonizing forensic psychiatry practices, and the pursuit of unified and transparent standards and practices underscore the importance of conducting cross-national comparative research.

Aims. More than 5 years after the OPGs’ clousure, we examined the transformation's state of the art and whether the forensic inpatients received a correct level of security treatment with a specific assessment tool, the DUNDRUM Toolkit.

Methods. The DUNDRUM Toolkit is a set of four different structured professional judgment and assessment tools consisting of four specific scales used with forensic patients for evaluation and treatment purposes (Structured Professional Judgment Tools for Admission, Urgency, Treatment Completion, and Recovery Evaluation Triage). In particular, DUNDRUM has been validated as matching best practice in Ireland, the UK, Belgium, Australia, New Zealand, and Canada, and now it has been validated in Italy. In this study, we used DUNDRUM 1 and 2 to evaluate whether the inpatient programs to which mentally ill offenders were assigned, having been legally determined to be socially dangerous, corresponded to their assessed needs. We also assessed whether legal placements corresponded not to the international criteria provided by the DUNDRUM toolkit. Then, we used DUNDRUM 3 and 4 to evaluate the effectiveness of the treatment and how to discharge the patient from REMS. A sample was recruited throughout the national territory of over 250 offenders affected by mental disorders from different contexts: Italian REMS, prisoners, released offenders, offenders admitted to hospitals and/or forensic facilities, forensic patients in the communities, and patients placed on waiting lists. The REMS and mental health and penitentiary facilities of Piedmont, Lombardy, Veneto, Tuscany, Lazio, Campania, Puglia, Basilicata, Calabria, and Sicily participated in the research. The research was approved by the Ethics Committee of the University Hospital of Bari (N. 66510/AA. GG of 09.16.2020).

Results. DUNDRUM Toolkit is an evaluation tool for allocating forensic patients to higher or lower levels of therapeutic security according to need. In the regions where it has been systematically adopted by all professionals involved in the forensic field (Piedmont, Emilia, Lazio, Puglia, Calabria, Sicily) it has proven to be a valid and effective tool. The DUNDRUM Toolkit has been shown to have excellent internal consistency and inter-rater reliability also for Italian population.

Conclusion. DUNDRUM Toolkit is shown to be a single, composite, adaptable, and easy to use, which can assess many needs and main characteristics identified in the Italian forensic treatment model, which could be easily adopted by all health professionals involved in the various forensic contexts.

This presentation focuses on the role that Large Language Models (LLMs) can play in the legal landscape, particularly in legal decision making, and the role of the jurist in their use.

To this end, it will be necessary to investigate what are the main areas of law where Artificial Intelligence can be most effectively applied and what are the main problems that have been analyzed so far (e.g., bias, explainability). The research aims to highlight in what aspects and to what extent LLMs represent an element of discontinuity from previous legal AI tools (which use supervised and unsupervised learning models). Only in this way will it be possible to discover whether the rules and assumptions of legitimacy applied so far in the use of AI in legal decision-making are still valid. This will allow for a conclusive discussion of the role of humans using LLMs in legal decision-making.

Scientific information and evidence can impact how decision-makers in criminal courts explain or contextualize the behavior and criminality of defendants. Prior empirical research on judges in the U.S. has provided valuable insights into the ways in which they think about behavioral and biological sciences and how this may shape their judgments during punishment considerations.

More than ever, judges may be particularly affected by scientific information in cases involving defendants diagnosed with psychiatric disorders–with judges’ knowledge and views differing on such information and potentially affecting their later perspectives on the sentencing of these defendants and their outcomes. Building upon this prior work, we present preliminary results from a mixed-methods study that examines how judges consider and apply scientific information during decision-making in sentencing contexts.

We interviewed 34 judges in U.S. state courts that preside over criminal cases and asked them to make sentencing decisions for fictional defendants with psychiatric disorders described as experiencing either a genetic or an acquired brain disorder. We also asked judges to rate the importance of various legal concepts and standards in cases where defendants suffer from a genetic disorder, compared to a history of childhood abuse or no brain disorder. The results of this study provide unique insights into how judges assess and weigh scientific information that explains or contextualizes defendants' behavior. We propose several ways forward that may help better integrate science and research findings into criminal justice decision-making processes, especially during sentencing.

In the complex landscape of Italian organised crime, this study explores the minds of mafia members, focusing on the notorious Foggia Society. Common perceptions represent these individuals as self-centred, cold and emotionally detached people who have no remorse and are unable to feel empathy or pity for others, even less for their victims, who they do not hesitate to plunder, break the law and kill. In a few words, this is the profile of a psychopath. However, some recent research on mafia men and women has excluded that members of these criminal organisations, especially at the highest levels, present this psychopathological dimension. Sociological and psychological studies suggest that the sense of identity experienced by mafiosi is inextricably embedded in the families and community to which they belong and that group identification represents a central dimension in explaining their attitudes and behaviors.

The data referring to, i.e. the absence of registration of traumatic childhood events in subjects belonging to mafia contexts, are not in line, however, with the high number of minors involved in mafia contexts: over 100,000 children live in these contexts and are registered and followed by Save the Children. We are faced with an apparent paradox: on the one hand the presence of mental disorders secondary to exposure to traumatic events in minors living in mafia families and on the other adults belonging to the same contexts who do not seem to experience the traumatic experiences they have experienced as such during their childhood. In a recent ruling by the Juvenile Court of the Calabria Region, parental responsibility is revoked in all cases in which serious damage to the child is ascertained due to the involvement of the family unit in the mafia context.

The authors try to provide some psychological and psychopathological hypotheses of this apparent paradox.

The research project presented here aims to investigate which personality traits, as measured by the Shedler-Westen Assessment Procedure-200 (SWAP-200; Westen & Shedler, 1999a, 1999b), are most representative of criminal non-responsibility. Accordingly, the main aim is to lay the groundwork for the development of a non-responsibility prototype on the SWAP-200 that can empirically support expert assessment of criminal responsibility.

A pool of experts (psychiatrists, medical examiners, psychologists, lawyers and judges), working in the forensic field and involved in the assessment of criminal responsibility, was recruited. The experts were asked to rate each item of the SWAP-200 on a Likert scale from 0 to 7, based on the degree of representativeness/descriptiveness of non-responsibility (i.e., lack of intent or will). To facilitate comparisons, the same scoring scale was used as in the original instrument. The questionnaire was available for completion via Qualtrics by accessing it through a designated link on any electronic device. Average expert scores were calculated for each of the 200 SWAP items and then ranked in descending order so that items with higher average scores were considered more representative. The items, ranked from most to least representative, formed a prototype of nonresponsibility (NRP), similar to how the PD and Q factors were constructed (Shedler, Westen, 1999a; 1999b). Finally, correlations were calculated between the resulting NRP prototype.

In the realm of forensic science, accurate disability assessment is crucial for justice and rehabilitation. This study introduces an unprecedentedly vast dataset, encompassing 80,000 entries spanning most cities and regions within Jiangsu Province, making it the largest collection for similar scientific endeavors. Capitalizing on this robust dataset, we have developed an advanced judicial assessment system that automates the detection of disabilities. By harnessing the power of artificial intelligence algorithms and a dual-driven data model approach, our system can automatically determine disability grades based solely on descriptions. Moreover, it offers a degree of detailed analysis regarding the state of the disability. The implications of our findings not only enhance the precision and efficiency of forensic evaluations but also shed light on the potential of AI in reshaping the domain of legal and medical assessments.

Large language models (LLMs) are demonstrating impressive performance on many reasoning and problem-solving tasks from cognitive psychology. When tested, their accuracy is often on par with average neurotypical adults, challenging long-standing critiques of associative models.

Here we analyse recent findings at the intersection of LLMs and cognitive science. Here we discuss how modern LLMs resurrect associationist principles, with abilities like long-distance associations enabling complex reasoning.

While limitations remain in areas like causal cognition and planning, phenomena like emergence suggest room for growth. Providing examples and increasing the dimensions of the network are methods that further improve LLM abilities, mirroring facilitation effects in human cognition.

Analysis of LLMs errors provides insight into human cognitive biases. Overall, we argue LLMs represent a promising development for cognitive modelling, enabling new explorations of the mechanisms underlying intelligence and reasoning from an associationist point of view. Carefully evaluating LLMs with the tools of cognitive psychology will further understand the building blocks of the normal and abnormal human mind.

In the near future, neurotechnology combined with AI may be applied for a variety of societal purposes. One area of possible application is forensic psychiatry. However, potential use of neurotechnology in this area of medicine raises serious concerns, both ethical and legal in nature. A prominent issue is related to the coercive context of forensic psychiatry. More specifically, doubts have been raised about the validity of a person's consent in such a setting. In this presentation, first, I will consider the current state of neurotechnological developments relevant to forensic psychiatry, and next, I will explore some prominent ethical and legal issues that arise regarding possible application in this field.

This talk will focus on how the incorporation of AI into neurotechnology, especially through Brain-Computer Interfaces (BCI) challenges fundamental notions of autonomous decision-making in medicine and beyond. Furthermore, it introduces the neurorights framework to identify whether and how these challenges can be overcome.

People with impaired decision-making capacity enjoy the same rights to access technology as people with full capacity. Our paper looks at realising this right in the specific context of artificial intelligence (AI) and mental capacity legislation.

Ireland’s Assisted Decision-Making (Capacity) Act, 2015 commenced in April 2023 and refers to ‘assistive technology’ within its ‘communication’ criterion for capacity. We explore the potential benefits and risks of AI in assisting communication under this legislation with focus especially on Ireland’s provisions for advance healthcare directives.

One of the ethical risks in this context is that AI may reach beyond communication and start to influence the content of decisions, especially among people with impaired capacity. For example, we asked ChatGPT to ‘Make me an advance healthcare directive’, its response did not explicitly suggest content, but it did suggest topics that might be included, which could be seen as leading. One possibility for circumventing this is to look to foundational models of AI which we will explore in our paper. Bodies such as the United Nations have produced guidance about ethical use of AI and these guide our analysis.

We will present analysis, guidance, and key principles for effective, safe, and progressive use of AI in relation to capacity legislation. We also highlight the need for open discussion about optimising the potential of AI while minimising risks in this population.

Artificial intelligence (AI) is a fast-growing field of computer science that can perform tasks that normally require human intelligence, such as reasoning, learning, decision making, and problem solving. In clinical setting, AI has proven to be an effective tool to diagnose diseases, develop personalized treatment plans, and assist clinicians with decision-making. Therefore, AI can be seen as a set of technologies that can enhance patient care across healthcare settings.

In the context of forensic psychiatry, AI has been suggested as a possible tool for improving violence risk assessment by reducing bias and supporting human decision with data-driven predictions. Applied to forensic psychiatric assessment, AI could create sophisticated models that can combine data (i.e. text, images) in complex ways to produce standardized judgements. Indeed, AI-based evaluation processes offer a means to enhance, amplify, or address the strengths and limitations inherent in current evaluation techniques.

The primary constraint in harnessing Artificial Intelligence systems for forensic psychiatric assessments lies in the absence of a comprehensive repository housing relevant data. Currently, existing knowledge and research in the field of forensic psychiatric evaluations rely on relatively small sample sizes, which presents a notable methodological challenge. Moreover, there is a pressing need for a universally accepted framework to identify the most effective assessment tools in terms of sensitivity and specificity for this context.

Our proposal is to establish a national repository that can aggregate data from all expert assessments, encompassing both structured instruments and unstructured clinical judgements. This foundational step is pivotal in paving the way for the integration of artificial intelligence systems into the domain of forensic evaluations.

The article mainly concerns the context of international human rights law, psychological torture and cybertorture. The article analyses how these ambiguous interpretations of psychological torture allow state actors to stealthily expand the practice of psychological torture on victims without leaving behind trace evidence of physical torture and evasion of the obligation to provide State compensation for victims by examining judicial cases publicly released by Chinese courts. Moreover, the article discussed the fact that the systematic cyberbullying of individuals by non-state actors on digital platforms resulting in new ethical claims arising from the fact that acts of psychological torture are not criminalised within the legal framework in China. Furthermore, the article also deals with the new challenges of technologies (Artificial Intelligence) in relation to cybertorture and human enhancement, critically analysing the legal and medical perspectives and discussing the risks that may be posed by the fact that there is no specific law to protect victim who have been subjected to cybertorture and that there is no medical support available to them.

Evidence visualization means presenting evidence in court using colorful graphics, charts, animations, and videos. Forensic scientific evidence is formed by individuals with specialized knowledge based on scientific principles, detection methods, standards, as well as decision-making and reasoning. This methods not only enhances the understanding and conviction of the participants in the litigation regarding scientific evidence in an intuitive and vivid manner but also promotes the authority and acceptability. With the development of science and technology, it provides more ways and tools for visualization methods, such as graphics, animations, virtual simulations, mind maps, and relationship diagrams. On the other hand, these methods and tools involve issues related to scientific validity, legality, and user experience. Particularly, regarding the discovery and assessment of electronic data, if hierarchical diagrams and relationship diagrams have not been judicially authenticated, the complex analytical results presented in visual charts may introduce information biases, affecting the cognitive experience of the participants in litigation. Systematically organizing the types of visualization methods and tools, as well as the scenarios they are suitable for and the points to be aware of, can guide the use of graphics to speak, pinpoint key issues concisely and clearly. It can also ensure the accuracy, legality, and comprehensibility of the presented evidence.

This article argues for guidelines for the use of GenAI in clinical psychiatry. Currently, clinicians using chatbots without appropriate approval risk undermining legal protections for patients. This could harm the patient and undermine the standards of the profession, undermining trust in an area where human involvement in decision-making is critical. To explore these concerns, this paper is divided into three parts. First, it explores the potential benefits of GenAI in clinical psychiatry and highlights the risks of its use. The examples considered are clinical decision support (CDS), suicide prediction, and psychiatric ward administration. Second, it outlines the general legal distinction between protections for patients in mental health and other healthcare settings. It will be seen that separate legal considerations around medical liability, informed consent, human rights, data accuracy and data privacy arise that require careful application to the area of GenAI as distinct from other hospital settings. The impact of the European Union’s (EU) Artificial Intelligence Act (AI-Act), the Artificial Intelligence Liability Directive (AILD), and the revised Product Liability Directive (PLD) are all considered in this regard. It will be seen that these laws are insufficient in the psychiatric context. Third, the paper makes recommendations for soft-law guidelines to help resolve the existing legal gaps. It is hoped that these shall eventually become codified in law.

New technology is always met with surprise, panic, and – eventually – prudent adoption of certain aspects of technological advances. This pattern is evident in the histories of most technologies: steam power in the 19th century, television in the 20th century, neuroscience and “artificial intelligence” in the 21st century. Each generation believes the technological advances of its era are quantitatively and qualitatively different to those of previous generations, but the underlying phenomenon is the same: the shock of the new, followed by gradual adjustment to (and of) new technology. These concerns are apparent today in relation to neuroscience and “artificial intelligence”, which reflect interesting but incremental advances on existing technologies, rather than stand-alone developments. The usual concern with all technologies (that they will replace certain aspects of human function) are, perhaps, more concerning in fields such as mental health law, which often applies to people with impaired decision-making capacity who might be especially vulnerable to technologies which appear to encroach on decision-making or other core areas of human function. This paper approaches this topic from an historical standpoint, noting previous panics, the limited negative impact of the relevant technologies, their positive contributions, and the possibilities offered by neuroscience and “artificial intelligence”, if they are approached in a proportionate, prudent, and person-centered way.

The advent of Artificial Intelligence revolutionised several practices in Forensic Psychiatry and Criminal Justice, from violence and recidivism risk assessment to treatment design and planning. Nowadays, we are witnessing a new wave of AI, known as Generative artificial intelligence (GenAI), which is influencing several sectors of society, including but not limited to business, healthcare, education, entertainment and the arts.

The power of GenAI to analyse a vast amount of unstructured data from different sources and generate realistic human-like content will have a profound impact on traditional discriminative AI applications, primarily focused on prediction and classification tasks.

This presentation explores the transformative impact of Generative Artificial Intelligence in Forensic Psychiatry and Criminal Justice, offering a comparison to the well-established paradigm of Discriminative AI. Furthermore, it will discuss the possibilities and challenges introduced by this shift, providing an in-depth analysis of longstanding and emerging risks and ethical considerations.

As AI applications continue to progress and permeate society, understanding the influence of Generative AI on the practice of Forensic Psychiatry is crucial for professionals, researchers, and policymakers to ensure a conscious and ethical application of these tools in forensic practices as well as to stimulate reflections and critical enquiries.

Evidence of bias affecting subjective decision-making in forensic disciplines and de-biasing strategies has been reported. Despite the American Academy of Psychiatry and Law's attempt to draw up forensic evaluation and reporting guidelines, several studies have shown a disagreement among forensic psychiatric reports that assessed the same case. In forensic psychiatric practice, multiple sources of cognitive bias were identified, possibly favored by the absence of methodological standardization of expert evaluations. This study aims to describe and synthesize scientific literature on cognitive bias in forensic psychiatrists to improve awareness of this field and de-biasing strategies. A scoping review was conducted by searching PubMed and Scopus for full-text articles in English published until June 2023, including forensic psychiatry and cognitive bias keywords. Another search was carried out among references of the selected articles. Of the 672 articles found, 101 met the inclusion criteria. Finally, 15 main biases were grouped and summarized by reporting the description of bias, the subcategories, and the first mention of each one. Our review highlighted possible de-biasing solutions and illustrates the complexity of bias influence in expert witness activities' decision-making. Finally, a new model will be proposed, and the opportunities and pitfalls of applying artificial intelligence tools in psychiatry forensic evaluation will be discussed.

Moral dilemmas arise when individuals perceive conflicts between various moral principles and rules. Researchers use sacrificial moral dilemmas to understand how people make moral choices and explore influencing factors. In the current study, multiple interactive video scenarios based on the "footbridge dilemma" were employed, and participants were encouraged to choose whether to let five people die or to act by sacrificing another person to save them in each scenario given. The type of action required to cause the individual's death varied among the scenarios, reflecting different interpersonal distances between the agent and the potential victim. We utilized a 3x2 design with Proximity (low, medium, high) and Type of choice (utilitarian or deontological) as within-subjects variables. The dependent variables included choices, reaction times, proximity estimates, and emotions experienced. Our results suggest that harmful acts requiring high proximity are perceived as less morally acceptable than those involving greater distance. These findings contribute to our understanding of this topic and have implications across multiple fields. Understanding how people perceive and respond to morally challenging situations at different levels of proximity, for example, could be useful in forensic science to gain insights into decision-making dynamics, as well as in medical ethics to account for factors that may impact decisions about patient care and treatment options.

The development of new Artificial Intelligence technologies allows the application of decision making methods to the field of clinical management of patients suffering from mental disorders. In particular, these technologies express potential for use in responding to unmet needs in the current clinical situation, such as the search for a diagnosis based on multiple complex factors, which is not achievable in the context of clinical management based on current diagnostic and statistical methodologies, and the consequent search for strictly personalized pharmacological treatments. Some initial data are highly indicative of potential in this direction.

However, the search for the answer to these unmet needs potentially poses some legal problems. However, the current state of initial development of the technologies and their only initial clinical application allow us to anticipate some issues of possible legal interest. In particular, the question could arise in the very near future of how much a doctor is bound to follow the diagnostic or therapeutic indications developed by decision making systems or how much a level of personal autonomy can be maintained in clinical decisions relating to individual patients. In particular, the problem of possible responsibilities may arise in the event of personal decision that are inconsistent with what is indicated by AI-based decision-making tools.

This is naturally a problem shared in every medical discipline, but it appears more critical in the treatment of psychiatric patients, in which the complexity of the factors at the origin of the disorders and of those that maintain the current symptoms, the scarcity of objective biological elements of support for the diagnosis and the current importance of the subjective clinical judgment of the psychiatrist in front of the individual case make the boundary between operations based on objective tools, such as those of the AI, and the individual decision of the doctor more subtle, with a consequent wide field of possible reflections of both a deontological nature and a legal responsibility.