WP1: Human-in-the-Loop Machine Learning, Reasoning and Planning
This WP aims to develop the fundamental Learning, Reasoning and Planing methodologies that allow humans to be interactively involved “in the loop”. As outlined in section 220.127.116.11 this goes beyond explainability (which in itself is a challenge) toward methods that allow interactive human input to influence their inner workings.
WP 2: Multimodal Perception and Modeling
Our ambition it to build on recent progress in discriminative and generative networks, to provide integrated multi-modal perception and modeling that combines fast real-time reaction for sensori-motor reflexes, with spatiotemporal and geometric reasoning, prediction of recurrent events and consequences for actions and dynamic processes and linguistic expressions for perceptual concepts to enable communication with and learning from humans. In prticuar we intend to develop systems that can understand complex human actions,
motivations and social settings.
WP 3: Human AI Collaboration and Interaction
This work package aims to establish new methodological and conceptual basis for human-AI collaboration. As described in Section 18.104.22.168, the goal is to develop methodology for social basis for human-AI partnership, especially group cognition and emotional expression. For AI to understand people, it needs to both be able to infer intentions and emotions from observations as well as make its own intentions understandable to human partners via grounding, emotional expression, and explanation. We believe that these capabilities need to be to some extent be engineered into AI, in order to ensure more natural behavior from first interaction and to reach a desirable level of controllability and transparency. However, they need to be made interactive for users to control and understand. The main objective of this WP is machine-learning methods and suitable interaction techniques based on theoretically grounded models of human-human communication, which can drive the inference and planning of an AI agent in a more human way and with less training data. These models include models of multimodal communication, for grounding, theory of mind, and emotion. They work with interaction histories collected over a longer timespan and over a richer set of sensors than previously.
WP 4: Societal AI
This work package aims at shaping the research on the societal dimension of AI, as increasingly complex socio-technical AI systems emerge, made by (explicitly or implicitly) interacting people and intelligent agents as described in section 22.214.171.124. It aims to address the undesired emerging network effects of social AI systems, as well as the design of transparent mechanisms for decentralized collaboration and decentralized personal data ecosystems that help toward desired aggregate outcomes, i.e., toward the realization of the agreed set of values and objectives at collective level, such as accessible and sustainable mobility in cities, diversity and pluralism in the public debate, fair distribution of economic resources, environmental sustainability, a fair and inclusive job market.
WP 5: AI Ethics and Responsible AI
This WP is dedicated to ensuring that AI systems operate within a moral and social framework, in verifiable and justified ways as elaborated in section 126.96.36.199). Theory and methods are needed for the Responsible Design of AI Systems as well as to evaluate and measure the ‘maturity’ of systems in terms of compliance to ethical and societal principles. This concerns legal, ethical, trustworthy aspects but need to be combined with robustness, social and interactivity design. The focus here is the prioritization of ethical, legal, and policy considerations in the development and management of AI systems to ensure responsible design, production and use of trustworthy AI. This requires integration of engineering, policy, law and ethics approaches. This topic is thus about understanding, developing and evaluating ethical agency and reasoning abilities as part of the behavior of artificial autonomous systems (e.g. artificial agents and robots). We will focus on explanation aspects and core data protection principles of fairness, transparency, accountability and responsibility.
WP 6: Applied research with industrial and societal use cases
This WP is dedicated to synchronising the research agenda and network activities with industrial and social needs. The three main objectives are (1) ensuring that the needs of important European industry are adequately then into account within the research agenda, (2) making sure that key results are evaluated in industrially (and socially) relevant use cases and (3) making sure that the knowledge created by the microprojects of WP1-5 reaches key European industrial players.
WP 7: Innovation Ecosystem and Socio-Economic Impact
The objective of this work package is to maximize the socio-economic impact of the research roadmap of the consortium. This is twofold, (1) providing means and mechanisms to transform basic and applied research results into ventures and businesses that are provide value to European citizens, and (2) to ensure that applied research is guided by real world challenges and steered toward domains that are beneficial for society. In this work package we provide research and provide mechanism that supports the creation of start-ups, the transformation of traditional (non-digital) SMEs into high-tech companies, and to push agile innovation in major industries. A range of dedicated mechanisms in envisioned and will be created, that creates leaders in AI technologies and applications.
WP 8: Virtual Center of Excellence, Capacity building and Dissemination
This WP is devoted to the aim of fostering excellence, increasing the efficiency of collaboration, disseminating the latest and most advanced knowledge to all the academic and industrial AI laboratories in Europe and making HumanE AI Net the center of a vibrant AI network in Europe. This includes implementing and operating the the Virtual Laboratory, integration of the Virtua Laboratory with the AI4EU platform, running the industrial Ph.D. postdoc and internship program, running the dissemination events to all relevant target groups (from scientific summer schools to workshops for policy makers and participation in public festivals) and the creation and distribution of relevant dissemination and knowledge spreading materials (from MOOCs, through policy brochures to general public facing YouTube videos).
WP 9: Synergies with AI on demand platform(s) and the Broader European AI Community
The aim of this WP is to embed HumanE AI Net within the landscape of relevant European and national initiatives. This includes, as primary concerns, the interaction with the AI on demand Platform (AI4EU), Digital Innovation Hubs and other ICT 48 networks as specified in the call. However, given the interdisciplinary nature of our work, we believe that HumanE AI is important for a wide range of initiatives around AI such as SoBigData, the European Language Technology community (European Language Grid) and the broader European AI networks (EurAI, CLAIRE, ELLIS). The WP is lead by Barry O’Sullivan from Cork who is the President of EuRAI, the main European AI association and well connected within virtually all relevant European AI communities.