GECKO (Grant number 955422) will bring together the latest scientific advances, establishing a unique synergy between the social sciences, information sciences, and engineering. Based on a deep understanding of how people interact with technology, in their homes and outside, GECKO will develop a unique end-to-end methodology for designing ART AI that goes far beyond the state of the art. By designing AI from a user perspective, GECKO will prevent the limitation of current technologies where AI reasoning and outcomes could easily be harmful, emotionally offensive, and damaging to the wider society and economy.
GECKO will target ART AI and explore alternative methods to build ML models drawing on the latest developments in information and social sciences using sustainable living labs and creating ‘virtual labs’ via crowdsourcing. For example, GECKO will leverage on explainable artificial intelligence (XAI), preventing bias outcomes, learning to explain, exploring ambiguity in coding6, analysing co-scripting technology and embedding causal inference in social systems through causal modelling. An inter-disciplinary methodology will be adopted to tackle the most prominent application example driver, where ML & social science must be considered together: addressing urgent sustainability and energy efficiency needs, where a successful responsible AI decision support technology must embed social science understanding of people’s actions and how they interact with technology.
Motivation and Opportunity
EU GDPR’s Article 12 in current regulations allows individuals to enquire about AI system decisions. Thus, providers must be in a position to explain the outcomes of their technology. Responsible and explainable AI is much more than ticking ethical ‘boxes’ as it goes deep into how much autonomy is given to a machine and how the machine will ossify and reinforce the initial values of the designers, during the evolving process of ML. Therefore, AI design should be robust, from both technical and social perspectives, since even with good intentions, AI systems can cause unintentional harm. With this in mind, GECKO will focus on sustainability in relation to households, where the decisions made by AI can significantly and directly affect people.
GECKO will embed low-energy outcomes into connected home technology (CHT) designs, through inter-disciplinary co-creation, working directly with technology users through sustainable living labs and virtual labs. GECKO capitalises on significant EC efforts in developing a local ecosystem of AI skillsets but augmenting AI practices towards strong evaluation, transferability, and accountability.
GECKO will bridge the discipline gap and explore advanced AI designs that consider technological, ethical, and social science angles, embedding non-quantifiable data into ML models, making them accountable, responsible, interpretable, and bias-free. With the new European Green Deal targets, GECKO’s unique ART AI design approach will serve as a major driver towards the first climate-neutral continent. Since GECKO addresses a topic that will continue to be groundbreaking in the long term, the aim is to establish a sustainable, European-wide network for interdisciplinary ART AI research.