On 15 and 16 July 2021, the DigForASP COST Action organised an online conference on Automated Decision Making with more than 250 specialists in Artificial Intelligence, Digital Forensics, Data Privacy, Law, and Ethics.
Automated decision-making (ADM) systems, i.e., systems used to take a decision by automated means without any human intervention, are increasingly widespread.
The development of new Artificial Intelligence (AI) techniques is contributing to making those systems even more common and accessible, and they are now integrated in everyday life. Although they can lead to quicker and more consistent decisions, by helping humans in reaching a better performance, there may be significant risks for citizens’ rights and freedom. Indeed, they are often used to affect citizens’ rights without them being fully aware of the consequences on their lives, as in the case of insurance contracts, verification of credit worthiness, hiring processes, granting of social benefits, and many other tools. In some cases, negative effects may occur even when used responsibly and lawfully. In addition, the increasing availability of personal data on the Internet and from interconnected devices has made the linking between multiple sources easier, allowing the automated and large-scale analysis and prediction of individuals’ preferences, habits, personality, and behaviour.
To address these problems, the European Commission recently published a proposal for a new Regulation on AI, providing specific compliance procedures for those systems that are considered more dangerous, requiring humans to be aware that they are interacting with an AI, and banning some technologies, such as facial recognition in public accessible places.
“Automated Decision Making, AI systems and individual rights” was organized by Dr. Chiara Gallese Nobile from Eindhoven University of Technology, NL, Prof. David Billard from the University of Applied Sciences in Geneva, CH, Dr. Elena Falletti from Carlo Cattaneo University LIUC, IT, and Prof. Francesca A. Lisi from the University of Bari “Aldo Moro”, IT. The conference was part of the initiatives of the COST Action “DIGital FORensics: evidence Analysis via intelligent Systems and Practices” (DigForASP), a cooperation network which explores the potential of the application of Artificial Intelligence and Automated Reasoning in the Digital Forensics field to create synergies between these fields. The conference was hosted by Eindhoven Artificial Intelligence Systems Institute and Data Management and Library Department at Eindhoven University of Technology (TU/e). It attracted more than 250 registered participants with a peak of 70 online attendees in the morning of the first day.
The conference explored the ethical, legal, and technical issues of ADM systems, such as the need for transparent and explainable AI, the difficulties in eliminating biases in data sets, and the consequences of using ADM systems in the legal and medical field. The issue of trust in AI systems has also been extensively discussed. The presentations showed how the use of ADM systems can have a wide range of implications on individuals and society, perpetuating discrimination, and societal biases, and can be sometimes incompatible with human rights: it is therefore important to apply the appropriate techniques and technical measures to minimize as much as possible the negative consequences towards citizens and in particular vulnerable groups. To tackle those issues, it is crucial to have an interdisciplinary vision of the subject, with a contribution from different disciplines.
The event opened in the morning of July 15th, 2021, with the welcome and the introduction by Dr. Gallese Nobile, followed by the salutations of Dr. Carlo van de Weijer (General Manager EAISI, TU/e) and Dr. Iza Witkowksa (Head of Data Management and Privacy, TU/e).
The session, chaired by Prof. Billard, started with the vice-chair of DigForASP, Prof. Stefania Costantini (University of L’Aquila, IT), who illustrated the main objectives of the Action.
Prof. Marta Poblet Balcell (RMIT University, AU) focused on the design of state-centric top-down governance systems within the extra-territorial scope of the future AI regulation.
Prof. Vincent Müller (Eindhoven University of Technology, NL) then addressed the ‘black box problem’ in a context of justification for judgments and actions, especially in privacy.
Prof. Cesar Hidalgo (University of Toulouse, FR) compared people’s reactions to actions performed by humans and machines, thus revealing the biases in human-machine interactions.
The afternoon session of July 15th was chaired by Dr. Gallese Nobile, who summarized the morning talks and introduced the main points of European Commission’s AI Act proposal.
The first talk of the afternoon was given by Prof. Alexandros Kalousis (University of Applied Sciences, CH) who discussed AI’s blind spots and societal challenges.
Prof. Francisco de Elizalde (IE University, ES) introduced the survey conducted among legal tech companies that use automated tools in the legal domain.
Prof. Andrea Pin (University of Padua, IT) discussed the recent AI Act proposal, explaining the difference between the US and EU approaches, and highlighted the difficulties of identifying what you can do and cannot do within a single legal system.
On 16 July 2021, the last session of the conference, chaired by Dr. Falletti, started with Dr. Angela Muller, who provided AlgorithmWatch’s perspective on the implications of ADM systems.
Prof. Sicco Verwer (Delf University of Technology, NL) presented useful techniques for avoiding bias and discrimination in data mining applications.
Prof. Susana Vieira (University of Lisbon, PT) urged multidisciplinary research teams in healthcare to appropriately analyse the data and understand the knowledge by means of AI.
The conference ended with a summary and some final remarks by Prof. Billard.
Each session of the conference was followed by a very active discussion among panellists and with the public, who had many questions and comments. Overall, the conference was very participated and stimulating.
Additional information:
The detailed programme of the event is available here.