Francesca Rossi

As AI systems make decisions that affect our lives, we must ensure that these systems operate according to the same constraints, guidelines, and ethical principles that a human would follow. Humans make many complex decisions that rely on their subjective preferences, but their decisions are usually also constrained by these ethical priorities. Hence it is essential to equip AI systems with the tools to evaluate whether or not preferences are compatible with these other priorities. In computer science, the Conditional Preference networks (CP-nets), which graphically represent conditional and qualitative preference relations, can be used to model, combine, and compare subjective preferences and ethical priorities. This chapter proposes that one can use CP-nets to measure the distance between an agent’s subjective preference and the ethical principles of the agent’s community in order to ensure that the decisions of an AI system are aligned with a given set of ethical priorities.

This article reports the findings of AI4People, a year-long initiative designed to lay the foundations for a “Good AI Society”. We introduce the core opportunities and risks of AI for society; present a synthesis of five ethical principles that should undergird its development and adoption; and offer 20 concrete recommendations – to assess, to develop, to incentivise, and to support good AI – which in some cases may be undertaken directly by national or supranational policy makers, while in others may be led by other stakeholders. If adopted, these recommendations would serve as a firm foundation for the establishment of a Good AI Society.

The concept of adaptation is employed in many fields such as biology, psychology, cognitive sciences, robotics, social sciences, even literacy and art,1 and its meaning varies quite evidently according to the particular research context in which it is applied. We expect to find a particularly rich catalogue of meanings within evolutionary biology, where adaptation has held a particularly central role since Darwin’s The Origin of Species (1859) throughout important epistemological shifts and scientific findings that enriched and diversified the concept. Accordingly, a conceptual taxonomy of adaptation in evolutionary biology may help to disambiguate it. Interdisciplinary researches focused on adaptation would benefit from such a result. In the present work we recognize and define seven different meanings of adaptation: (1) individual fitness; (2) adaptation of a population; (3) adaptation as the process of natural selection; (4) adaptive traits; (5) molecular adaptation; (6) adaptation as structural tinkering; (7) plasticity. For convenience here, we refer to them as W-, P-, NS-, T-, M-, S- and PL-ADAPTATION. We present the seven meanings in some detail, hinting at their respective origins and conceptual developments in the history of evolutionary thought (references are offered for further deepening). However, it is important to point out that evolution researchers seldom if ever refer to a single meaning purified from the others. This applies also to the authors we cite as representatives of one of the seven meanings. In Discussion and Conclusion draw from our work some future perspectives for adaptation within evolutionary biology.

Constraint programming is the fruit of several decades of research carried out in mathematical logic, automated deduction, operations research and arti?cial intelligence. The tools and programming languages arising from this research?eldhaveenjoyedrealsuccessintheindustrialworldastheycontributetosolving hard combinatorial problems in diverse domains such as production planning, communication networks, robotics and bioinformatics. This volume contains the extended and reviewed versions of a selection of papers presented at the Joint ERCIM/CoLogNET International Workshop on Constraint Solving and Constraint Logic Programming, which was held from June 30 to July 2, 2003. The venue chosen for the seventh edition of this annual workshop was the Computer and Automation Research Institute of the Hungarian Academy of Sciences in Budapest, Hungary. This institute is one of the 20 members of the Working Group on Constraints of the European Research Consortium for Informatics and Mathematics. For many participants this workshop provided the?rst opportunity to visit their ERCIM partner in Budapest. CoLogNET is the European-funded network of excellence dedicated to s- porting and enhancing cooperation and research on all areas of computational logic, and continues the work done previously by the Compulog Net. In part- ular, the aim of the logic and constraint logic programming area of CoLogNET is to foster and support all research activities related to logic programming and constraint logic programming. The editors would like to take the opportunity and thank all the authors who submitted papers to this volume, as well as the reviewers for their helpful work.

The main purpose of this paper is to investigate the role of social and emotional learning skills and resilience in explaining mental health in male and female adolescents, during the COVID-19 pandemic. Three self-report questionnaires were administered to 778 participants aged between 11 and 16 years and recruited from 18 schools in Northern Italy. The SSIS-SELb-S and the CD-RISC 10 assessed SEL and resilience skills respectively, while the Strengths and Difficulties Questionnaire was used to measure mental health in terms of internalizing problems, externalizing problems, and prosocial behavior. We found that SEL and resilience skills were positively and significantly associated with each other, negatively associated with internalizing and externalizing problems, and positively related to prosocial behavior. Three linear regression analyses showed the significant role of resilience, age, and gender in explaining the variance of internalizing problems; the significant role of SEL skills, resilience, age, and gender in explaining the variance of externalizing problems; and the role of SEL skills, age, and gender in explaining prosocial behavior. Importantly, we found that resilience fully mediated the relationship between SEL skills and internalizing problems, partially mediated the relationship between SEL skills and externalizing problems and didn't mediate the relationship between SEL skills and prosocial behavior. The paper concludes with a discussion of the limitations of the study as well as its practical implications.

[Multiple authors] In this paper, we propose a general architecture that is based on fast/slow solvers and a metacognitive component. We then present experimental results on the behavior of an instance of this architecture, for AI systems that make decisions about navigating in a constrained environment. We show how combining the fast and slow decision modalities allows the system to evolve over time and gradually pass from slow to fast thinking with enough experience, and that this greatly helps in decision quality, resource consumption, and efficiency.

This article reports the findings of AI4People, an Atomium—EISMD initiative designed to lay the foundations for a “Good AI Society”. We introduce the core opportunities and risks of AI for society; present a synthesis of five ethical principles that should undergird its development and adoption; and offer 20 concrete recommendations—to assess, to develop, to incentivise, and to support good AI—which in some cases may be undertaken directly by national or supranational policy makers, while in others may be led by other stakeholders. If adopted, these recommendations would serve as a firm foundation for the establishment of a Good AI Society.

While the search for models and explanations of cognitive phenomena is a growing area of research, there is no consensus on what counts as a good explanation in cognitive science. This Ph.D. thesis offers a philosophical exploration of the different frameworks adopted to explain cognitive behaviour. It then builds on this systematic exploration to offer a new understanding of the explanatory standards employed in the construction and justification of models and modelling frameworks in cognitive science. Sub-goals of the project include a better understanding of some theoretical terms adopted in cognitive science and a deep analysis of the role of representation in explanations of cognitive phenomena. Results of this project can advance the debate on issues in general philosophy of cognitive science and be valuable for guiding future scientific and cognitive research. In particular, the goals of the thesis are twofold: (i) to provide some necessary desiderata that genuine explanations in cognitive science need to meet; (ii) to identify the framework that is most apt to generate such good explanations. With reference to the first goal, I claim that a good explanation needs to provide predictions and descriptions of mechanisms. With regards to the second goal, I argue that the neurocomputational framework can meet these two desiderata. In order to articulate the first claim, I discuss various possible desiderata of good explanations and I motivate why the ability to predict and to identify mechanisms are necessary features of good explanations in cognitive science. In particular, I claim that a good explanation should advance our understanding of the cognitive phenomenon under study, together with providing a clear specification of the components and their interactions that regularly bring the phenomenon about. I motivate the second claim by examining various frameworks employed to explain cognitive phenomena: the folk-psychological, the anti-representational, the solely subpersonal and the neurocomputational frameworks. I criticise the folk-psychological framework for meeting only the predictive criterion and I stress the inadequacy of its account of cause and causal explanation by engaging with James Woodward’s manipulationist theory of causation and causal explanation. By examining the anti-representational framework, I claim that the notion of representation is necessary to predict and to generalise cognitive phenomena. I reach the same conclusion by engaging with William Ramsey (2007) and Jose Luis Bermudez (2003). I then analyse the solely subpersonal framework and I argue that certain personal-level concepts are indeed required to successfully explain cognitive behaviour. Finally, I introduce the neurocomputational framework as more promising than the alternatives in explaining cognitive behaviour. I support this claim by assessing the framework’s ability to: (i) meet the two necessary criteria for good explanations; (ii) overcome some of the other frameworks’ explanatory limits. In particular, via an analysis of one of its family of models — Bayesian models — I argue that the neurocomputational framework can suggest a more adequate notion of representation, shed new light on the problem of how to bridge personal and subpersonal explanations, successfully meet the prediction criterion (it values predictions as a means to evaluate the goodness of an explanation) and can meet the mechanistic criterion (its model-based methodology opens up the possibility to study the nature of internal and unobservable components of cognitive phenomena).