Human-Centred Artificial Intelligence (HCAI) is a transformative research agenda focused on creating intelligent systems that are embedded in complex human social and organisational environments. HCAI leverages a decade of remarkable research success into AI capabilities of language understanding, visual interpretation and autonomous devices. It brings a new spotlight onto the social, psychological and cognitive impacts of AI in order to produce a new category of systems that support humanity intelligently. HCAI is a multidisciplinary effort that seeks to shape, guide, and inform the development of future AI systems based on an understanding of the nature of machine intelligence and a foundational concern with the future of humanity.
The Electronics and Computer Science (ECS) Web and Internet Science (WAIS) research group, studies the interaction between the computational and human constituents of very large-scale information systems. Southampton’s Centre for Machine Intelligence (CMI) brings together researchers and practitioners in AI and Autonomous Systems to deliver the impact of machine intelligence to society, while the Centre for Health Technologies (CHT) works closely with clinical professionals to develop theoretical and practical AI solutions that deal with complex health data and improve diagnostic results.
Southampton is also home to the Web Science Institute (WSI), which leads cross-university research and training activities in data science, web science and AI. The WSI manages Centres for Doctoral Training in Web Science and Human Centred AI, provides online CPD courses, fosters enterprise and impact, influences public policy and coordinates with the UK Alan Turing Institute.
The area of Natural Language Processing (NLP) has marked tremendous progress in the last decade. I am George Konstantinidis and I am looking into the question of what we can learn from the domain of social sciences to improve the processing, detection, and explainability of language and narratives, and in particular hostile narratives. Our research into declarations of war and terrorist manifestos verifies that this kind of language is highly structured, commonly discusses two distinct groups of people (us against them), and uses arguments related to economy, religion, health, family and other recurring disciplines in order to elevate one (the self) and alienate the other.
Utilising theories from multiple Social Science disciplines including psychology, peace studies, group theory, and social identity theory, we devise new NLP techniques, and build algorithmic approaches to detect, analyse and understand hostile narratives. At the same, we measure the digital spread of hostile narratives through the social web and the Internet, and evaluate the sociological impact of hostile narratives by correlating against public health, political and economic data.