ELSA AI lab Northern Netherlands (ELSA-NN)

Four different use cases will be used to develop and map ELSA knowledge and investigate the use of different AI applications with different types of data in different decision-making contexts at different periods during the life course.

Genetic data

• Problem: Currently, the Dutch newborn screening program (NBS) detects 27 rare diseases using biochemical techniques in bloodspots, including 18 inborn errors of metabolism (IEM). Since all IEM are genetically based, genetic techniques like next-generation sequencing (NGS) techniques could be used to detect rare diseases in NBS. In doing so, NBS could also be expanded to other genetic diseases with early onset and options for treatment and prevention. However, NGS generates lots of genetic variants that all have to be analyzed for clinical relevance which is difficult and laborious.
• AI-driven solution(s): Given the huge number of samples in the Dutch NBS program (~170,000/year) and the need to report results within days, when genetic techniques would be used in NBS in the future, the use of AI-based variant interpretation programs might benecessary. Using AI-based prediction programs in NGS variant analysis could reduce the manual curation step making it possible to reach the very short turnaround time requested for the large number of samples.
• Collaboration partners: ELSA-NN collaborates for this use case with the ELSI research group, the Genome Diagnostics group and the Genomics Coordination Centre of the department of genetics of the UMCG.
  Use case leader: Imke Christiaans

Monitoring data

• Problem: Despite rising life expectancy, nearly half of the Dutch population has chronic conditions mostly caused by an unhealthy lifestyle, which greatly affects quality of life, and increases health care utilization. Healthy living and a good environment can help prevent chronic diseases. However, this often requires long-term behavioral changes and making healthy choices, which is difficult.
• AI-driven solution(s): Smart sensors can support citizens in behavioural and healthier changes. For example, smart sensors in "smartwatches" and "activity trackers" can measure daily exercise, sleep patterns, perceived stress and heart rate levels. Using AI, all this information can be combined and analysed. The results of these analyses can provide personalised feedback and advice on lifestyle and health via an app on the phone.
• Collaboration partners: ELSA-NN collaborates for this use case with the project ‘healthy living as a service’, which aims to provide citizens with health, technology and data literacy and give them ownership over their health.
  Use case leader: Claudine Lamoth

Personal Health Data

• Problem: In Dutch hospitals, the electronic patient record (EPD) is used to maintain the patient's medical history. This file includes basic patient data, medication history, test results and notes from healthcare providers. The EPD offers a lot of valuable information that can be used in patient care, however, getting the patient information in and out of the EPD goes together with an administrative burden on the level of the physician.
• AI-driven solution(s): Generative AI solutions can be used in the EPD context, for example, to generate concept answers to patient questions and summarise and process relevant patient information.
• Collaboration partners: ELSA-NN collaborates for this use case with the AI EPD project group of the UMCG.
  Use case leaders: Job Doornberg

Synthetic data

• Problem: One of the biggest challenges for deep learning AI models is collecting large enough datasets of high quality. In many cases, there are too few training samples to achieve a reliable and generalisable deep learning network, and retrieving data from the clinical setting is even harder because of privacy and security reasons. Moreover, when datasets are collected, there is always a risk of bias because of an ill-selected population or simply the availability of only skewed data.
• AI-driven solution(s): Data augmentation is a common methodology to extend the number of training samples. A more recent development in deep learning showed that deep learning networks can also be used to generate synthetic data. This technique could also be employed to add new samples to datasets that are generated using deep learning techniques based on the original database. These new samples are fully synthetic and no longer linked to a real patient.
• Collaboration partners: ELSA-NN collaborates for this use case with the machine learning lab community of the UMCG.
  Use case leader: Peter van Ooijen

WPs will investigate the ethical, legal, social and psychological aspects concerning the implementation of AI in healthcare. Besides the research activities carried out in the WPs, ELSA-NN will also:

• ensure patient and public participation in the development of ELSA-NN research activities;
• focus on dissemination of information about responsible development and implementation of AI;
• promote internationalisation through learning from and cooperation with international partners;
• integrate ELSA aspects in a tool intended to guide people through the whole process from AI development to AI implementation.

WP ethical issues

Objectives

• Enhancing ethical competency/literacy of all stakeholders involved in applying AI tools in healthy living, working and ageing decision-making contexts.
• Developing input for ethics by design guidelines and principles by generating and mapping a practical overview of moral understandings of different stakeholders developing and using AI and by identifying stakeholders’ responsibilities to ensure reliable and trustworthy AI in healthcare.

Methodology

• Empirical ethics: empirical ethics input will be provided, using qualitative methods. In order to ensure reliable and trustworthy AI, the moral understandings of a wide variety of stakeholders will be investigated. These moral understandings will be analysed and structured. This empirical input will be used to contribute to AI-ethics literacy, i.e. making stakeholders ethically competent regarding use and development of AI.
  Work package leaders: Els Maeckelberghe and Christoph Jedan

WP legal issues

Objectives

• To perform a data-protection impact assessment (DPIA) for the proposed processing of personal data for each use case.
• To develop input for privacy by design guidelines and principles by generating and mapping the legal frameworks that apply in the different use cases, identify the validity, completeness and quality of datasets, algorithms and AI outcomes, and terms and conditions for decision-making, interventions and self-management by patients versus automated decision making and profiling.

Methodology

• Desk research, interviews with stakeholders, dataflow diagrams and the Dutch Framework for DPIA's
  Work package leaders: Jeanne Mifsud Bonnici and Marie Jose Bonthuis

WP Socio-political issues

Objectives

• To investigate socio-political and regulatory considerations in countering health and digital disinformation, with a specific focus on low SES perspective.
• To provide input for how (different) socio-political factors and regulatory incentives may influence: trust in AI, willingness to participate in health research, and digital health literacy and health disparities at individual and societal levels, therewith contributing to an increase in understanding regarding willingness to share data, trust in AI, digital health literacy and health disparities.

Methodology

• This WP will take broad socio-political and demographic approaches as a starting point to classify public/patients as well as healthcare professionals within the use cases and map the regulatory scenarios and incentives that might be available to these participants through a secondary literature review. Based on the classification from the data and literature review additional surveys, focus group discussions and interviews on digital literacy and health literacy will be conducted amongst stakeholders (with a specific focus on lower SES).
  Work package leaders: Jeanne Mifsud Bonnici and Ritumbra Manuvie

WP Psychological issues

• To examine the public concerns within and across the different AI use cases
• To examine how concerns and drivers of acceptance are associated with acceptance of AI within the different use cases
• To enhance understanding of how and which information should be offered to the general public regarding AI applications in healthcare to lessening the concerns and whether it can be associated with increased acceptance in the use cases

Methodology

• The first and second objective will be answered in the general population, using a case-based design as part of the larger population study in the overall project. Using a survey, hypothetical cases resembling the use cases in this proposal will be described in detail. Participants will randomized to one of the use cases and will answer general questions as well as questions specifically related to the use cases.
• The third objective will be addressed with an experimental pre-post design, in which participants will be offered extended information addressing the concerns that were identified in the previous study.
  Work package leaders: Adelita Ranchor and Maya Schroevers

WP Art

Objectives

• To study the use of art as a means to engage with AI.

Methodology

• To enable, organise and evaluate art-scientific collaboration within the context of different ELSA-NN use cases to explore, understand and extend existing research processes integrating computer-based technologies, especially AI applications.
  Work package leaders: Anke Coumans and Judith van der Elst

WP Participation

Objectives

• To involve the public and patients from the Northern region of the Netherlands in the development of ELSA-NN activities

Methodology

• A patient advisory council (PAC) will-be set-up including public and patient representatives to provide advice on all planned ELSA-NN (research) activities.
• Dialogues with patients and public from the Northern region will be organised to develop awareness and provide information about responsible development and implementation of AI
• The initiative has been taken to set-up the ‘Your technology of tomorrow’ educational program, to increase young people's awareness about and knowledge of AI and to actively involve them in debates about AI development and application.
  Work package leaders: Mirjam Plantinga and Petra Steenbergen

WP Dissemination

Objectives

• To disseminate the ELSA-NN results in order to enhance knowledge about the ELSA issues involved when developing and implementing trustworthy AI to promote healthy living, working and ageing.

Methodology

• Dissemination is done by creating an ELSA-NN website, developing educational materials for healthcare professionals and public and patients, organizing educational meetings for societal partners in the Northern region, and scientific dissemination via open-access scientific publishing and conference presentations.
• Together with the DASH, ELSA-NN has set up the ‘Your technology of tomorrow’ educational program, to increase young people's awareness and knowledge of AI and to actively involve them in debates about AI development and application. By offering the program 'Your Technology of Tomorrow' in the science truck, young people can learn about AI in a playful and interactive way and complex technology is made understandable and accessible.
  Work package leaders: Peter van Ooijen and Michiel Hooiveld

WP Internationalisation

Objectives

• To learn from and integrate knowledge from international use cases in ELSA-NN.
• To stimulate international cooperation between researchers from ELSA-NN, Centre for Personalised Medicine (CPM) and Clinical Ethics, Law and Society (CELS).

Methodology

• International cooperation between researchers from ELSA-NN, CPM and CELS will be stimulated by inviting researchers from CPM and CELS to attend the yearly ELSA-NN meetings and by organising a workshop with participants and presentations from ELSA-NN, CPM and CELS.
  Work package leaders: Lisa Ballard and Anneke Lucassen

WP Integration

Objectives

• The WP integration researches how we can integrate ELSA principles in the Machine Learning / AI development process to integrate the knowledge related to the concepts of availability, use and performance in an online ELSA tool for trustworthy and human-centred AI;
• To test the ELSA tool for trustworthy and human-centred AI within the use cases and among stakeholders in the Northern region to evaluate if the developed ELSA tool sufficiently guides users to the process of AI and ELSA aspects, in different decision-making contexts in the fields of healthy living, working and ageing.
• This process is intended to guide people through the whole process from AI development to AI implementation.

Methodology

• The ELSA tool will be tested iteratively within the use cases and among stakeholders in the Northern region.
  Work package leaders: Claudine Lamoth and Hilbrand Oldenhuis

 

The ELSA-NN consortium consists of researchers from different knowledge institutes in the Northern region (UMCG, University of Groningen (UG) and Hanze University of Applied Sciences) representing diverse age, genders, ethnicities and social status and bringing in different (ELSA) expertise.

• Marie-José Bonthuis: WP legal issues, RUG
• Imke Christiaans: use case genetic data, UMCG
• Anke Coumans: WP art, Hanze
• Judith van der Elst: WP art, Hanze
• Hadi Hasan: WP socio-political issues, UMCG
• Michiel Hooiveld: WP dissemination, UMCG
• Christoph Jedan: WP ethical issues, RUG
• Lennart Johansson: use case genetic data, UMCG
• Rick de Klerk: WP integration, Hanze
• Claudine Lamoth: use case monitoring data and WP integration, UMCG
• Sara Soriano Longaron: WP psychological issues, UMCG
• Els Maeckelberghe: WP ethical issues, UMCG
• Ritumbra Manuvie: WP socio-political issues and WP legal issues, RUG
• Balasz Markos: WP legal issues, RUG
• Jeanne Mifsud Bonnici: WP legal issues and WP socio-political issues RUG
• Jacobien Niebuur: WP psychological issues, UMCG
• Hilbrand Oldenhuis: WP integration, Hanze
• Peter van Ooijen: use case synthetic data and WP dissemination, UMCG
• Mirjam Plantinga: project leader ELSA-NN, UMCG
• Adelita Ranchor: WP psycological issues, UMCG
• Natascha Reich: Project manager ELSA-NN, UMCG
• Maya Schroevers: WP psychological issues, UMCG
• Steven Warmelink: WP integration, Hanze

AI has an important role to play for patients and citizens. After all, it is about their data, and when in used in the context of healthcare, about their health and their care. In the patient and public advisory board of ELSA-NN they therefore provide their perspective toward AI development and everything that happens in the ELSA Lab. The board consists of members with diverse backgrounds, trying to create a broad and diverse representation of different patient and public groups. Current members of the patient and public advisory board are: Erwin Steegen, Tom Hallegraeff, Marian Kuypers, Alex Schepel, Ria van Loon, Ali Drenth, Diana Wolsink, Sereh Simons and Joke Wentholt. The patient advisory board group meets about four times a year and is being coordinated by Zorgbelang Groningen.

ELSA-NN is integrated within the Data Science Center in Health (DASH) and is working together with the AI hub Northern-Netherlands, the Jantina Tammes School and the Health Technology Research & Innovation Cluster (HTRIC) and additional consortium partners, especially patients/citizens, private, governmental and research representatives, to have a quadruple-helix consortium.

Patients and public organisations

 

Societal partners

 

Business partners

 

International partners

 

Knowledge Institutions