Skip to main content

Interactive Natural Language Technology for Explainable Artificial Intelligence

Periodic Reporting for period 1 - NL4XAI (Interactive Natural Language Technology for Explainable Artificial Intelligence)

Reporting period: 2019-10-01 to 2021-09-30

The aim of this four-year European project is to use natural language to generate explanations for decisions made by an Artificial Intelligence (AI) system, which are understandable to non-expert users.

According to Polanyi's paradox, humans know more than they can explain, mainly due to the huge amount of implicit knowledge they unconsciously acquire through culture, heritage, etc. The same applies for AI systems. According to EU legislation, humans have a right to explanations of those decisions affecting them, even if AI systems make such decisions; however, AI systems (which mainly learn automatically from data), often lack the required transparency. Therefore, NL4XAI, the first European Training Network (ETN) on Natural Language (NL) and Explainable AI (XAI), is aimed for researching on how to develop more transparent systems. The main goal of this initiative is to train 11 creative, entrepreneurial and innovative early-stage-researchers who will face the challenge of designing and implementing a new generation of self-explanatory AI systems. The development of XAI systems requires addressing technical issues (e.g. designing explainable algorithms and human-machine interfaces) but also legal and ethical issues. XAI validation must be done in compliance with the assessment list for trustworthy AI which was developed by the High-Level Expert Group on Artificial Intelligence set up by the European Commission to help assess whether AI systems are in agreement with the requirements (Human Agency and Oversight; Technical Robustness and Safety; Privacy and Data Governance; Transparency; Diversity, Non-discrimination and Fairness; Societal and Environmental Well-being; Accountability) in their Ethics Guidelines for Trustworthy AI. Moreover, XAI validation must be also carried out in line with the EU General Data Protection Regulation (GDPR) and the EU regulation for AI.

NL4XAI includes a broad program of training events and opportunities, ranging from network-wide events, to courses covering technical and scientific domains and transferable skills. Each ESR will work on an individual research project at one of the network’s host-organizations and take part in network wide training events and meetings, as well as in secondments to other beneficiaries or partners.

As a result, the ESRs will be well prepared to design and build XAI models that generate interactive explanations on the basis of NL and visual tools which are intuitively understandable even by non-expert users, validated by humans in specific use cases and accessible to all European citizens. Main outcomes are to be publicly reported and integrated into a common open source software framework for XAI. In addition, those results to be exploited commercially will be protected through licenses or patents.
In the first 2 years of implementation, the work carried out mainly consists of the agreed establishment of the project management system to ensure a good implementation of the project (with especial emphasis on the data management plan and ethical issues), the recruitment of ESRs and the starting of the scientific work and trainings. ESRs have been trained in fundamentals of XAI as well as in complementary skills. They have thoroughly revised the state of the art of the underlying research areas in the four technical work packages: XAI models, end-to-end Natural Language Generation (NLG) systems, argumentation technology (ARGTECH) for XAI and interactive interfaces for XAI.

In addition to having published some papers in international conferences, we have organized the first two workshops on Interactive Natural Language Technology for Explainable Artificial Intelligence (XAI@INLG2019 and XAI@INLG2020). These workshops are part of a series of workshops to be organized in the context of NL4XAI, with the aim of establishing a forum of discussion regarding automatic generation of interactive explanations in natural language, as humans naturally do, and as a complement to visualization tools. The audience attending these workshops were early-stage and senior researchers as well as practitioners interested in taking Natural Language Generation further to enable the next generation of XAI systems. This is the reason why the workshops were collocated with the International Conference on Natural Language Generation.
NL4XAI ESRs will become a new generation of scientists ready to develop a more human-centred set of AI systems, facing XAI challenges through an interdisciplinary approach which merges AI, NL technology (NLP, NLG and argumentation) and HCI. NL4XAI will bring together a set of leading international actors able to foster research and technological development around XAI systems.

Relevance of boosting the AI strategy is of major importance, and even the relevance of the topic of XAI for the European strategy has been explicitly identified to ensure an appropriate ethical and legal framework for AI systems, which could facilitate the development of an AI that benefits people. In a new economic and societal model where AI systems will play a fundamental role, training of professionals is one of the most important factors for adapting Europe to a new model pervasive in the use of intelligent systems. The NL4XAI network will enhance the perspectives of trainees as it fulfils the following conditions:
• Uniqueness: NL4XAI offers a unique opportunity to be involved in one of the most promising research challenge for the future of AI. Cooperation will enable top research groups and companies to work together to speed up breakthroughs and tackle the challenges in XAI, which will give ESRs a deep vision of leading research and innovation in AI, NL technology and HCI, i.e. the key research areas for the development of XAI.
• Triple-i: Involvement of trainees, researchers and industry in a series of research and training activities will enhance the transfer of technologies, methodologies and scientific results among all related research areas. It will help the trainees to work in different professional and interdisciplinary environments. Moreover, the international and inter-sectoral mobility will give ESRs a deep vision of leading research and innovation in the key research topics for XAI.
• Enhanced training: NL4XAI training programme will improve and complement traditional PhD training. ESRs will benefit from the exposure to different research disciplines, inter-sectoral collaboration, geographical mobility and training in research and transversal skills.
• Future career opportunities: The quality of the involved research groups and industrial partners will provide future career opportunities for trainees. AI research is in the top of national and international agendas, which assures a high employability for trainees in coming years. Participation in top international conferences, and collaboration with top researchers will permit the trainees to establish connections with leading research groups. Secondments in industry will be an excellent opportunity to establish connections with the private sector.
• Embryo for XAI industry: NL4XAIwill work as an embryo for the development of an XAI industry in Europe. NL4XAI objectives face from an integrative point of view the main technologies that will be necessary for the development of XAI. Technological outputs, contact with industry should act as a catalyst for the development of new products and businesses.
NL4XAI logo