Researchers: Stanley Greenstein (SU), Max Gordon (KI/Danderyd Hospital)

Background: As digitization within our society increases so too are emerging technologies such as artificial intelligence (AI) being used in order to make all domains more effective. The implementation of AI into the health care domain is especially complex as the risks associated with this domain are magnified. It is for this reason that the implementation of AI in the Swedish health care system has been a lot slower compared to other fields. Any implementation of AI must invariably start with a mapping of the analogue processes, in this case the patient diagnostic and treatment process. Only then can a determination be made regarding how AI can most effectively be applied, considering relevant ethical and legal values.

Aim: To create a schematic blueprint for the development and application for an orthopedic medicine AI system that will suggest the optimal course of action in relation to acute injury treatment, thereby assisting specialists with the decision-making processes. The project addresses four research questions: what factors influence diagnostic and treatment decisions, what does the effective introduction of AI entail, what are the relevant ethical and legal considerations, and how should these factors be balanced against each other?

Methods: The research questions will be addressed by providing a description of the typical ‘medical processes timeline’ and focus on how this can be made more effective while at the same time adhering to current and future ethical and legal frameworks.

Significance: This project is unique in that it takes a human-centered approach, reflecting not only clinical factors but also non-medical factors relevant to patient autonomy, fear of pain, loss of function, and motivation. A multi-stakeholder perspective, including input from medical practitioners, patients, legal scholars, and data scientists, will facilitate a comprehensive exploration of risks and challenges. The anticipated impact includes promoting the acceptance of AI systems within orthopedic medicine, enhancing data-driven precision medicine and diagnostics, strengthening patient rights, promoting legal science, and advancing legal analysis of technological solutions.

Researchers: Harald Hammarström (UU), Tobias Andermann (UU)

Background: There are around 7,000 spoken languages in the world and around half of those are threatened by extinction. This linguistic diversity serves as an invaluable resource, offering insights into our species’ unique communication system, as well as understanding how our brains process languages. Interestingly, the diversity of languages is not evenly distributed across the globe, but rather distributed in a spatial pattern reminiscing biodiversity patterns. Recent advances in data availability, including detailed demographic, biotic, climatic, and remote sensing data, have paved the way for a quantitative analysis of these spatial patterns.

Aim: A series of papers from the last two decades has partially succeeded to explain linguistic diversity through geographical conditions but fails to consider that such diversity could easily be obliterated by local expansions, especially since colonial times. New AI models developed by one of the Pis (Andermann et al. 2022) have the potential to predict the language diversity potential of a given site, which will allow the researchers to identify areas with recent language diversity loss by subtracting the actual language diversity at a site from the potential diversity at that site.

Methods: To track endangered languages and assess their potential extinction risk, the researchers have identified a unique suitability of computational models originally developed for biodiversity research. This is expected to work since the mechanisms driving biodiversity and linguistic diversity are similar in nature. The project seeks to reassess the drivers of linguistic diversity, model the evolution of speaker communities, and situate linguistic diversity within cultural diversity and biodiversity.

Significance: The project could benefit society through an improved recognition of linguistic diversity components and their ecological limits, a state-of-the-art database featuring endangerment, speaker numbers and future projections, and a contribution to the wider understanding of biodiversity with one of the most important human cultural features. Specifically, databases with speaker numbers are in high demand due to their numerous linguistic and societal applications, and will be made open-access and long-term archived through integration in Glottolog.

Researchers: Fabian Lorig (MaU), Heidi Howard (LU)

Background: Donated organs might be the difference between life and death for many patients but even if the organs are available, many patients die due to ineffective donor systems. Only in Sweden, 33 patients died in 2021 while being on the waiting list and Europe-wide this number is a staggering 10 deaths each day. Even when organs reach their recipient, there may still be issues of poor organ matches or organs being allocated to less needy recipients. Efforts to optimize the donor program have been ongoing for decades and even if progress is made, there are still important gaps in making the system fair and efficient while also adhering to ethical, legal, and social frameworks. Social simulations have proven to be promising tools for supporting policy making. They allow us to carry out harmless simulations on artificial populations and provide valuable insights through in silico experiments. So far, simulations of organ donation only focused on a limited number of aspects, such as waiting list prioritization strategies, registration in different donor service areas, and strategies for organ offering, thus failing to provide a framework for assessing the global effects of policies.

Aim: The project aims to investigate the potential of using Agent-based Social Simulations (ABSS) to support organ donation policy making and will address both the technical (DDLS) and ethical, legal, and social (WASP-HS) aspects.

Methods: European countries often use different donor policies, i.e. opt-out laws, to optimize the number of successful donations but changing strategy might pose a big risk if unsuccessful. A strategy that is successful in one country might not be in another and vice versa. To test which policies might work, ABSS, where an artificial population is used to generate synthetic (simulated) data, will be used.

Significance: The project will serve as an important pre-study to identify facilitators, barriers, and requirements for the development of an ABSS model of organ donation policies. The results will have an impact on all four DDLS research areas as a tool for generating and communicating data and particularly on the Data-driven Precision medicine and diagnostics research area, where it provides a new approach for assessing the effectiveness of different treatment policies.

Researchers: Sverker Sikström (LU), Loïs Vanhée (UmU)

Background: Mental illness is one of the largest causes of suffering in society and of social costs. An alternative to the classic therapies, characterized by high treatment costs, is the online Cognitive Behavioral Therapy (CBT). It provides a scalable, cost-effective, and accessible approach, building on modules that patients can carry out independently. CBT therapists are required to assess patients’ therapeutic context and recommend appropriate therapeutic plans. Currently, these activities are carried out manually, using scales and therapists’ intuitions and memories, which are known to be time-consuming and difficult to carry as the number of patients or sessions increase.

Aim: To establish foundations for developing AI-enhanced CBT therapy methods for supporting therapists both during the assessment and therapy planning phases.

Methods: Systematic literature reviews for developing a specific vocabulary, ontology and theory through interdisciplinary crossing of AI assessment and planning methods with psychology and therapeutic assessment and planning methods and praxis. Data collection for enabling the validation and assessment of future tools.

Significance:

WASP-HS perspective: The project will bring forwards new approaches for AI methods to be used for mental health and psychology research praxis, in particular in the context of CBT applied for depression and anxiety.

DDLS perspective: Data collection will support assessing the relations between the key variables for training, calibrating and validating assessment and CBT planning models.

Together, the project provides a concrete case for enabling a scalable precision personalized mental healthcare and well being.

Researchers: Andreas Kotsios (UU), Robin Strand (UU)

Collaborator: Mikael Laaksoharju (UU)

Background: AI has been regarded as one of the most promising technologies for diagnosis and treatment of cancer. For example, thanks to deep learning, we can generate real-time clear images of tumours and aid clinicians in their decision-making process on the best treatment plan. However, there is still some reluctance when it comes to deploying AI-based therapy in real-world treatments, not least because of the various ethical and legal questions that arise in this context.

Aim: This project aims to create a better understanding of ethical and legal considerations when AI technology is used for personalized treatment planning, in order to promote the uptake of AI-based solutions for cancer treatment.

Methods: Existing methods for advanced 3D deep learning motion models based on the stream of images coming from the MR during the radiation and real-time adaptation strategies will be examined in order to map the current ethical and legal problems and then to recommend alternative ways for the development of such methods based on the standard of trustworthy AI.

Significance:

WASP-HS perspective: This project aims at providing with insights from within a specific technological domain (data-driven diagnosis and treatment of cancer) in order to examine how and whether legislation and the ethics behind specific legal interpretations can effectively achieve the goals they claim as desired.

DDLS perspective: This project will add to the existing research on data-driven cancer diagnosis and treatment by recommending a sustainable model for such design and development.

Researchers: Sonja Aits (LU) and Emily Boyd (LU)

Background: The loss of biodiversity and the threats to human health and societies due to pollution, habitat damage, and climate change require urgent solutions. However, the large amount of fragmented information across millions of scientific publications determines a lack of understanding of the complex interconnections between these issues that hinders progress.

Aim: The project aims at creating a large knowledge graph that links between climate change, biodiversity, human health and social sustainability and that can be used to pinpoint potential mitigation strategies.

Methods: The project is based on the implementation of a text mining pipeline based on large language models that can process all accessible scientific literature on biodiversity, climate change, medicine, and social sustainability. Further computational analysis will allow to identify strategies to mitigate effect of major drivers on the system.

Significance:

WASP-HS perspective: the knowledge graph will help in clarifying sustainability-related challenges.

DDLS perspective: The project will develop a method to make text mining technologies accessible to a much larger audience of life science, social science and humanities researchers and even non-academics and will establish new knowledge about the link between biodiversity, social science and humanities research.