Janna Hastings

Background: The integration of artificial intelligence (AI) into health care has become a crucial element in the digital transformation of health systems worldwide. Despite the potential benefits across diverse medical domains, a significant barrier to the successful adoption of AI systems in health care applications remains the prevailing low user trust in these technologies. Crucially, this challenge is exacerbated by the lack of consensus among experts from different disciplines on the definition of trust in AI within the health care sector. Objective: We aimed to provide the first consensus-based analysis of trust in AI in health care based on an interdisciplinary panel of experts from different domains. Our findings can be used to address the problem of defining trust in AI in health care applications, fostering the discussion of concrete real-world health care scenarios in which humans interact with AI systems explicitly. Methods: We used a combination of framework analysis and a 3-step consensus process involving 18 international experts from the fields of computer science, medicine, philosophy of technology, ethics, and social sciences. Our process consisted of a synchronous phase during an expert workshop where we discussed the notion of trust in AI in health care applications, defined an initial framework of important elements of trust to guide our analysis, and agreed on 5 case studies. This was followed by a 2-step iterative, asynchronous process in which the authors further developed, discussed, and refined notions of trust with respect to these specific cases. Results: Our consensus process identified key contextual factors of trust, namely, an AI system’s environment, the actors involved, and framing factors, and analyzed causes and effects of trust in AI in health care. Our findings revealed that certain factors were applicable across all discussed cases yet also pointed to the need for a fine-grained, multidisciplinary analysis bridging human-centered and technology-centered approaches. While regulatory boundaries and technological design features are critical to successful AI implementation in health care, ultimately, communication and positive lived experiences with AI systems will be at the forefront of user trust. Our expert consensus allowed us to formulate concrete recommendations for future research on trust in AI in health care applications. Conclusions: This paper advocates for a more refined and nuanced conceptual understanding of trust in the context of AI in health care. By synthesizing insights into commonalities and differences among specific case studies, this paper establishes a foundational basis for future debates and discussions on trusting AI in health care.

We discuss recent progress in the development of cognitive ontologies and summarize three challenges in the coordinated development and application of these resources. Challenge 1 is to adopt a standardized definition for cognitive processes. We describe three possibilities and recommend one that is consistent with the standard view in cognitive and biomedical sciences. Challenge 2 is harmonization. Gaps and conflicts in representation must be resolved so that these resources can be combined for mark-up and interpretation of multi-modal data. Finally, Challenge 3 is to test the utility of these resources for large-scale annotation of data, search and query, and knowledge discovery and integration. As term definitions are tested and revised, harmonization should enable coordinated updates across ontologies. However, the true test of these definitions will be in their community-wide adoption which will test whether they support valid inferences about psychological and neuroscientific data.

We discuss recent progress in the development of cognitive ontologies and summarize three challenges in the coordinated development and application of these resources. Challenge 1 is to adopt a standardized definition for cognitive processes. We describe three possibilities and recommend one that is consistent with the standard view in cognitive and biomedical sciences. Challenge 2 is harmonization. Gaps and conflicts in representation must be resolved so that these resources can be combined for mark-up and interpretation of multi-modal data. Finally, Challenge 3 is to test the utility of these resources for large-scale annotation of data, search and query, and knowledge discovery and integration. As term definitions are tested and revised, harmonization should enable coordinated updates across ontologies. However, the true test of these definitions will be in their community-wide adoption which will test whether they support valid inferences about psychological and neuroscientific data.

Mental and behavioral disorders represent a significant portion of the public health burden in all countries. The human cost of these disorders is immense, yet treatment options for sufferers are currently limited, with many patients failing to respond sufficiently to available interventions and drugs. High quality ontologies facilitate data aggregation and comparison across different disciplines, and may therefore speed up the translation of primary research into novel therapeutics. Realism-based ontologies describe entities in reality and the relationships between them in such a way that – once formulated in a suitable formal language – the ontologies can be used for sophisticated automated reasoning applications. Reference ontologies can be applied across different contexts in which different, and often mutually incompatible, domain-specific vocabularies have traditionally been used. In this contribution we describe the Mental Functioning Ontology (MF) and Mental Disease Ontology (MD), two realism-based ontologies currently under development for the description of humanmental functioning and disease. We describe the structure and upper levels of the ontologies and preliminary application scenarios, and identify some open questions.

While genomic and proteomic information describe the overall cellular machinery available to an organism, the metabolic profile of an individual at a given time provides a canvas as to the current physiological state. Concentration levels of relevant metabolites vary under different conditions, in particular, in the presence or absence of different disorders. Metabolite concentrations thus mediate an important link between chemistry and biology, contributing to a systems-wide understanding of biological processes and pathways. However, there are a number of challenges in the ontological representation of such information. Firstly, concentration information is numeric and ranges over continuous values, while ontologies consist of discrete classes. Secondly, ontologies usually model only what is certain, and their logical formalisms are adapted to reasoning from certain axioms to logical deductions, however, the link between chemicals and diseases via concentration levels, like many threshold phenomena, is both uncertain and vague. In this paper we evaluate the representation of this knowledge using a combination of concrete domains and probabilistic reasoning. We parse concentration values from HMDB and create an ontology able to distinguish normal from abnormal concentrations and able to evaluate a probabilistic risk category for the presence of an associated disorder.

Progress in addiction science is hampered by disagreements and ambiguity around its core construct: addiction. Addiction Ontology (AddictO) offers a path to a solution of the kind that has addressed similar problems in other areas of science: a set of clearly and uniquely defined entities to which terms such as ‘addiction’, addictive disorder’ and ‘substance dependence ’can be applied for ease of reference while recognizing that it is the construct definitions and their unique IDs that are central, not the terms.

Advances in emotion and affective science have yet to translate routinely into psychiatric research and practice. This is unfortunate since emotion and affect are fundamental components of many psychiatric conditions. Rectifying this lack of interdisciplinary integration could thus be a potential avenue for improving psychiatric diagnosis and treatment. In this contribution, we propose and discuss an ontological framework for explicitly capturing the complex interrelations between affective entities and psychiatric disorders, in order to facilitate mapping and integration between affective science and psychiatric diagnostics. We build on and enhance the categorisation of emotion, affect and mood within the previously developed Emotion Ontology, and that of psychiatric disorders in the Mental Disease Ontology. This effort further draws on developments in formal ontology regarding the distinction between normal and abnormal in order to formalize the interconnections. This operational semantic framework is relevant for applications including clarifying psychiatric diagnostic categories, clinical information systems, and the integration and translation of research results across disciplines.

Ontologies aim to represent what is general, by means of universal statements. In contrast, dispositional predications capture knowledge about what is likely to happen if a certain set of circumstances obtain, which is crucial in investigative research such as in drug discovery and systems biology, where entities which are constitutionally dissimilar can nevertheless have similar behavior in a biological context. While such dispositional properties are increasingly included in biomedical ontologies, the circumstances under which the dispositions are realized are seldom explicitly modeled, and doing so is problematic due to the necessary restriction to binary relations in OWL ontologies. In this paper we address this shortcoming, focusing on the bioactivity of small molecules at varying levels of concentration within a living organism as our problem domain, although our approach is generalizable to other problems. We discuss the ontological nature and representation of dispositions and their realization; consider the nature of concentrations and their representation; and finally we detail an approach to linking dispositions to the conditions for their realization which regards conditions as triggers for the process in which the disposition is realized.

Landgrebe and Smith provide an argument for the impossibility of Artificial General Intelligence based on the limits of simulating complex systems. However, their argument presupposes a very contemporary vision of artificial intelligence as a model trained on data to produce an algorithm executable in a modern digital computing system. The present contribution explores what it means to be artificial. Current artificial intelligence approaches on modern computing systems are not the only conceivable way in which artificial intelligence technology might be created. If there are conceivable routes by which an artificial intelligence might be developed that are not constrained in the same way as the current generation of artificial neural networks, then these might be plausible routes towards the engineered generation of artificial general intelligence that are not precluded by the Landgrebe and Smith argument against the possibility of artificial general intelligence.

In August 2022, only a few months before ChatGPT was released, Barry Smith, well-known contemporary philosopher, together with Jobst Landgrebe, artificial intelligence entrepreneur, published a book entitled Why Machines will Never Rule the World: Artificial Intelligence without Fear (Landgrebe and Smith 2022). In this important, dense and far-reaching work, Landgrebe and Smith argue from the mathematical theory of complex systems, and a sophisticated analysis of the capabilities of human intelligence, that AGI— at the level of human intelligence—will never be possible. In broad outline, Landgrebe and Smith claim that intelligent human behaviour is a facility that develops from a complex world and for a complex world; that the complexities of the real world and our behaviour in it can never be fully specified or simulated mathematically; that digital systems can only execute what can be simulated mathematically; and that therefore human-level intelligence will never be manifested by a digital system. The present collection of articles is in response to and inspired by Landgrebe and Smith. It provides reflections from a variety of perspectives and disciplines on the possibility of AGI, as well as on the transformations that AI technology is already bringing even in its current incarnation.

Affective science conducts interdisciplinary research into the emotions and other affective phenomena. Currently, such research is hampered by the lack of common definitions of te rms used to describe, categorise and report both individual emotional experiences and the results of scientific investigations of such experiences. High quality ontologies provide formal definitions for types of entities in reality and for the relationships between such entities, definitions which can be used to disambiguate and unify data across different disciplines. Heretofore, there has been little effort directed towards such formal representation for affective phenomena, in part because of widespread debates within the affective science community on matters of definition and categorization. We describe our efforts towards developing an Emotion Ontology (EMO) to serve the affective science community. We here focus on conformity to the BFO upper ontology and disambiguation of polysemous terminology.

Ontologies are being developed throughout the biomedical sciences to address standardization, integration, classification and reasoning needs against the background of an increasingly data-driven research paradigm. In particular, ontologies facilitate the translation of basic research into benefits for the patient by making research results more discoverable and by facilitating knowledge transfer across disciplinary boundaries. Addressing and adequately treating mental illness is one of our most pressing public health challenges. Primary research across multiple disciplines such as psychology, psychiatry, biology, neuroscience and pharmacology needs to be integrated in order to promote a more comprehensive understanding of underlying processes and mechanisms, and this need for integration only becomes more pressing with our increase in understanding of differences among individuals and populations at the molecular level concerning susceptibility to specific illnesses. Substance addiction is a particularly relevant public health challenge in the developed world, affecting a substantial percentage of the population, often co-morbid with other illnesses such as mood disorders. Currently, however, there is no straightforward automated method to combine data of relevance to the study of substance addiction across multiple disciplines and populations. In this contribution, we describe a framework of interlinked, interoperable bio-ontologies for the annotation of primary research data relating to substance addiction, and discuss how this framework enables easy integration of results across disciplinary boundaries. We describe entities and relationships relevant for the description of addiction within the Mental Functioning Ontology, Chemical Entities of Biological Interest Ontology, Protein Ontology, Gene Ontology and the Neuroscience Information Framework ontologies.

Disagreement about the nature of ‘addiction’, such as whether it is a brain disease, arises in part because the label is applied to a wide range of phenomena. This creates conceptual and definitional confusions and misunderstandings, often leading to researchers talking past one another. Ontologies have been successfully implemented in other fields to help solve these problems by creating unifying frameworks that can accommodate divergence while clarifying the basis for it. We argue that ontologies can help transform the way we think and communicate about addiction as well, providing a basis for a shared understanding of the domain and a shared means for communicating and achieving that understanding. We introduce the use of ontologies in science and medicine and show how they can unify differing perspectives on addiction. We also show how ontologies provide a practical tool for organizing our collective knowledge in this field, even, and especially, when that knowledge is garnered through distinct disciplinary approaches adopting disparate methodologies or perspectives (e.g., neurobiological, psychological, sociological, etc.). In the end, we believe that ontologies can help to avoid letting questions like “Is addiction a brain disease?” divide the field, hamper communication, or otherwise disrupt the ability to make meaningful progress.

Mental health research faces a suite of unresolved challenges that have contributed to a stagnation of research efforts and treatment innovation. One such challenge is how to reliably and validly account for the subjective side of patient symptomatology, that is, the patient’s inner experiences or patient phenomenology. Providing a structured, standardised semantics for patient phenomenology would enable future research in novel directions. In this contribution, we aim at initiating a standardized approach to patient phenomenology by sketching a tentative formalisation within the framework of an applied ontology, in the broader context of existing open-source Open Biomedical Ontologies resources such as the Mental Functioning Ontology. We further discuss a number of prevailing challenges and observations bearing on this task.

There have been renewed calls to use phenomenology in psychiatry to improve knowledge about causation, diagnostics, and treatment of mental health conditions. A phenomenological approach aims to elucidate the subjective experiences of mental health, which its advocates claim have been largely neglected by current diagnostic frameworks in psychiatry (eg, DSM-5). The consequence of neglecting rich phenomenological information is a comparatively more constrained approach to theory development, empirical research, and care programmes. Although calls for more phenomenology in psychiatry have been met with enthusiasm, there is still relatively little information on how to practically facilitate this integration. In this Personal View, we argue that phenomenological approaches need a shared semantic framework to drive their innovative potential, thus enabling consistent data capture, exchange, and interoperability with current mental health data and informatics approaches (eg, the Research Domain Criteria project). We show how an applied ontology of phenomenological psychopathology offers a suitable method to address these challenges.

Ontology development methodologies emphasise knowledge gathering from domain experts and documentary resources, and knowledge representation using an ontology language such as OWL or FOL. However, working ontologists are often surprised by how challenging and slow it can be to develop ontologies. Here, with a particular emphasis on the sorts of ontologies that are content-heavy and intended to be shared across a community of users (reference ontologies), we propose that a significant and heretofore under-emphasised contributor of challenges during ontology development is the need to create, or bring about, consensus in the face of disagreement. For this reason reference ontology development cannot be automated, at least within the limitations of existing AI approaches. Further, for the same reason ontologists are required to have specific social-negotiating skills which are currently lacking in most technical curricula.

Affective science conducts interdisciplinary research into the emotions and other affective phenomena. Currently, such research is hampered by the lack of common definitions of terms used to describe, categorise and report both individual emotional experiences and the results of scientific investigations of such experiences. High quality ontologies provide formal definitions for types of entities in reality and for the relationships between such entities, definitions which can be used to disambiguate and unify data across different disciplines. Heretofore, there has been little effort directed towards such formal representation for affective phenomena, in part because of widespread debates within the affective science community on matters of definition and categorization. To address this requirement, we are developing an Emotion Ontology (EMO).

The Emotion Ontology is an ontology covering all aspects of emotional and affective mental functioning. It is being developed following the principles of the OBO Foundry and Ontological Realism. This means that in compiling the ontology, we emphasize the importance of the nature of the entities in reality that the ontology is describing. One of the ways in which realism-based ontologies are being successfully used within biomedical science is in the annotation of scientific research results in publicly available databases. Such annotation enables several objectives, including searching, browsing and cross-database data integration. A key benefit conferred by realismbased ontology is that suitably annotated research results are able to be aggregated and compared in a fashion that is based on the underlying reality that the science is studying. This has the potential of increasing the power of statistical analysis and meta-analysis in data-driven science. This aspect has been fruitfully exploited in the investigation of the functions of genes in molecular biology. Cognitive neuroscience uses functional neuroimaging to investigate the brain correlates of areas of mental functioning such as memory, planning and emotion. The use of functional neuroimaging to study affective phenomena such as the emotions is called ‘affective neuroscience’. BrainMap is the largest curated database of coordinates and metadata for studies in cognitive neuroscience, including affective neuroscience (Laird et al., 2005). BrainMap data is already classified and indexed using a terminology for classification, called the ‘Cognitive Paradigm Ontology’ (CogPO), that has been developed to facilitate searching and browsing. However, CogPO has been developed specifically for the BrainMap database, and the data are thus far not annotated to a realism-based ontology which would allow the discovery of interrelationships between research results across different databases on the basis of what the research is about. In this contribution, we describe ongoing work that aims to annotate affective neuroscience data, starting with the BrainMap database, using the Emotion Ontology. We describe our objectives and technical approach to the annotation, and mention some of the challenges.