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. (shrink)

The European Medical Information Framework project, funded through the IMI programme, has designed and implemented a federated platform to connect health data from a variety of sources across Europe, to facilitate large scale clinical and life sciences research. It enables approved users to analyse securely multiple, diverse, data via a single portal, thereby mediating research opportunities across a large quantity of research data. EMIF developed a code of practice to ensure the privacy protection of data subjects, protect the interests of (...) data sharing parties, comply with legislation and various organisational policies on data protection, uphold best practices in the protection of personal privacy and information governance, and eventually promote these best practices more widely. EMIF convened an Ethics Advisory Board, to provide feedback on its approach, platform, and the EcoP. The most important challenges the ECoP team faced were: how to define, control and monitor the purposes for which federated health data are used; the kinds of organisation that should be permitted to conduct permitted research; and how to monitor this. This manuscript explores those issues, offering the combined insights of the EAB and EMIF core ECoP team. For some issues, a consensus on how to approach them is proposed. For other issues, a singular approach may be premature but the challenges are summarised to help the community to debate the topic further. Arguably, the issues and their analyses have application beyond EMIF, to many research infrastructures connected to health data sources. (shrink)

Ex aequo et bono compensations refer to tribunal’s compensations that cannot be determined exactly according to the rule of law, in which case the judge relies on an estimate that seems fair for the case at hand. Such cases are prone to legal uncertainty, given the subjectivity that is inherent to the concept of fairness. We show how basic principles from statistics and machine learning may be used to reduce legal uncertainty in ex aequo et bono judicial decisions. For a (...) given type of ex aequo et bono dispute, we consider two general stages in estimating the compensation. First, the stage where there is significant disagreement among judges as to which compensation is fair. In that case, we let judges rule on such disputes, while a machine tracks a certain measure of the relative differences of the granted compensations. In the second stage that measure, which expresses the degree of legal uncertainty, has dropped below a predefined threshold. From then on legal decisions on the quantity of the ex aequo et bono compensation for the considered type of dispute may be replaced by the average of previous compensations. The main consequence is that this type of dispute is, from this stage on, free of legal uncertainty. (shrink)