THEMIS 5.0. Deliverable 2.1. - THEMIS 5.0 Methodological Framework and Requirements Analysis
Year: 2024
Type: report
Abstract: In WP2 of the THEMIS 5.0, the conceptual basis for the innovative approach of the project is developed, comprising the theoretical groundwork and the human-centric requirements that are summarized in this deliverable, as well as a comprehensive conceptual model to be presented in the forthcoming deliverable D2.2. ‘THEMIS 5.0 Approaches and Conceptual Modelling’. The conceptual basis is established in dialogue with use-case partners as well as with users and domain experts through co-creation sessions (WP3) and are used to guide the technical architecture and development (WP4) as well as the piloting and evaluation (WP5). This deliverable presents the theoretical groundwork for the THEMIS 5.0 approach to trustworthiness optimization of AI systems which details the concept of trustworthiness for AI systems, considers specific trustworthiness characteristics, expands on the involvement of human users for the adaptation of AI systems to human values and business targets, and discusses the use of risk management approaches for assessing AI trustworthiness. The concept of trustworthy AI is grounded in the relevant knowledge base, including scientific literature, technical reports, standards, policy documents, and legal material. This concept is based on the notion that trustworthiness may be assessed and optimized through detectable trustworthiness criteria associated with the AI systems under test within their socio-technical environment. Furthermore, relevant background on human-centred AI promoting the involvement of human users in optimization of AI trustworthiness is analysed, risk management approaches as means to address AI trustworthiness are discussed, and technologies supporting the analysis of AI trustworthiness characteristics and management are detailed. Finally, an overview of the ethical and legal basis for the THEMIS 5.0 approach is documented. Furthermore, this deliverable presents the user requirements, the user journeys and the mock-ups for the THEMIS 5.0 ecosystem. Based on input from users and experts in the three case domains of THEMIS 5.0. the user requirements are identified, analysed and summarized on the basis of extensive user involvement. The requirements are instantiated in mock-ups illustrating user journeys to guide the technical development in the project. The human-centric aspect of the conceptual work is further addressed in collaboration between WP2, and the co-creation activities performed within WP3, and includes the user requirements for the THEMIS ecosystem, the expectations of users from trustworthy AI in their line of work and us-case specific contextual information and trustworthiness considerations that will guide the development of the THEMIS 5.0 services. User journeys and mock-ups describing the different ways that users can use the functionalities of the THEMIS 5.0 ecosystem for the optimisation of AI models in the project use cases were developed to support the upcoming co-creation activities (Living Labs and Piloting).
Source: Lirias
Authors: Doulgerakis, Adam, Tsalapati, Eleni, Giotis, Giorgos, Følstad, Asbjørn, Remonato, Filippo, Vegard Johnsen, Pål, Fikardos, Mattheos, Lepenioti, Katerina, Apostolou, Dimitris, Mentzas, Gregoris, Stamatellos, Giannis, Psaltidou, Alexandra, Garsia, Sara, Corrêa Harcus, Ana Maria, Taylor, Steve, Senior, Samuel, Kioskli, Kitty, Fotis, Theo, Billion, Arnaud, Karavokiros, George
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7053099980
Cites:
Cited by:
Related to: