HUMAINT · HUman behaviour and MAchine INTelligence

The goal of the HUMAINT project is to provide a multi-disciplinary under­standing of the impact that artificial intelligence (AI) systems have on hu­man behaviour. The project has three main goals: to advance the scientific understanding of machine and human intelligence; to study the impact of algo­rithms on human behaviour, especially on decision-making and cognitive and socio-emotional development; and finally, to provide insights to policy makers with respect to the previous issues.

HUMAINT research has three main characteristics: it is inter-disciplinary, combining methodologies from machine learning, economics and cognitive sci­ence; it is reproducible: generating open publications, datasets, code and re­search protocols; and it is collaborative: enabling the team to work with re­searchers from different institutions worldwide. Through research, the project enables the promotion of value-centred and ethical approaches for the develop­ment and application of AI systems and it enables the team to actively engage in global initiatives on AI & Human Rights.

In order to have a broad understanding of the impact of AI on human be­haviour, the project focuses on four different scenarios with high social impact and then it establishes commonalities between them to extract scientific and policy-relevant conclusions.

The impact of algorithms on decision making in criminal justice

This scenario studies biases and discrimination in human and algorithm decision making. In particular, human and algorithm decisions for discrimination as well as the design of algorithms with performance and fairness as target outcomes are evaluated, and research is done on explaining and empirically proving which sources of algorithmic discrimination there are. The project has looked at criminological assess­ments of defendants in Catalonia, and open frameworks for fairness evalua­tion have been created.

The impact of AI systems on the tasks we do at work

A framework for the analysis of the impact of AI on the labour mar­ket is developed by extending the task-based approach to the definition of work by cognitive abilities that are required to perform said tasks. Data from AI initiatives is used to identify AI fields with more research activity and where most progress is expected in the future, and to map these fields to the dimension of cognitive abili­ties. This framework is then validated on three international databases on tasks: (1) PIAAC (Programme for the International Assessment of Adult Competencies); (2) EWCS (European Working Conditions Survey), and (3) O*NET (The Occupational Information Network). Through this approach we can identify which occupations are most likely to be transformed by AI over sectors and countries.

The impact of social robots on child development

The project investigates the im­pact of embodied AI on the ways chil­dren develop in formal (school) and informal (play) settings. Towards this direction, various existing robotic plat­forms are used to conduct real-life user studies with children by design­ing specific robot behaviours and in­teraction scenarios. In the first year of the project, the impact of child-robot voluntary interaction on a child’s prob­lem-solving process and performance was investigated. In the second-year user study, the project aims to under­stand a child’s trust development in collaborative settings in terms of ro­bots’ prior cognitive and social behav­iour. In the future, the project aspires to create open-ended scenarios to in­vestigate the role of social robots on a child’s collaborative creative think­ing as one of the current societal de­mands.

The impact of AI on music and culture

The application of artificial intel­ligence (AI) to music stretches back many decades, and presents numer­ous unique opportunities for a variety of uses, such as the recommendation of recorded music from massive com­mercial archives, or the (semi-) auto­mated creation of music. The HUMAINT project addresses these two scenarios where AI can produce outcomes in a domain fully entrenched in human cre­ativity. It intends to better understand how AI is changing our relationship with music and culture.

Finally, with regard to transversal activities, the HUMAINT project organ­ises a yearly interdisciplinary school on AI and its ethical, legal, social and economic impact and it contributes to external research initiatives with inter­national organisations and academic institutions worldwide.