Harnessing benefits and preventing harms of AI cannot be solved alone through technological fixes and regulation. It depends on a complex interplay between technology, societal governance, individual behaviour, organizational and societal dynamics. Enabling people to understand AI and the consequences of its use and design is a crucial element for ensuring responsible use of AI.
In this report we suggest a new framework for the development and use of AI technologies in a way that harnesses the benefits and prevents the harmful effects of AI. We name it Reflective AI. The notion of Reflective AI that we propose calls for adopting a holistic approach in the research and development of AI to investigate both what people need to learn about AI systems to develop better mental models i.e. an experiential knowledge of AI, to be able to use it safely and responsibly, as well as how this can be done and supported.
The work towards this report has been supported by the Volkswagen Foundation within its programme "Artificial Intelligence and the Society of the Future" as part of the project Reflective AI: Understanding and Designing Environments for Reflective Information Practices in a Digital Society (2020 - 2021) conducted by the European Institute for Participatory Media (Berlin), Radboud University (Nijmegen) and Technical University Dortmund.
The report would not have been possible without the interviews and workshop participation from experts and practitioners from diverse fields. We want to thank them for their invaluable contributions that helped us in shaping this report and developing further the notion of reflective AI.
We interviewed following experts for the purpose of this report: Maria Luciana Axente (PwC), Kirsti Elisabeth Berntsen (Norwegian University of Science and Technology), Xavier Brandao (#iamhere France), Randi Cecchine (University of Amsterdam), Dorothee Dersch (PD – Berater der öffentlichen Hand GmbH), Bart Goethals (University of Antwerp), Dietmar Jannach (Universität Klagenfurt), Lorenz Matzat (AlgorithmWatch), Sofia Papavlasopoulou (Norwegian University of Science and Technology), Nathalie Smuha (KU Leuven), Alex Urban (#iamhere Germany), Bob van de Velde (Nederlandse Publieke Omroep), Thabo van Woudenberg (Erasmus University Rotterdam) and Marcus Winter (University of Brighton).
Following researchers and practitioners took part in our interdisciplinary workshop "Reflective AI in a digital society" in May 2020: Mina Dennert (#iamhere International), Dorothee Dersch, (PD – Berater der öffentlichen Hand GmbH), Jana Diesner (University of Illinois at Urbana-Champaign), Virginia Dignum (University of Umeå), Jos Hornikx (Radboud University), Thomas Liebig (TU Dortmund), Lavinia Marin (TU Delft), Judith Möller (University of Amsterdam), Özlem Özgöbek (Norwegian University of Science and Technology), Niccolo Pescetelli (Max Planck Institute for Human Development), Till Plumbaum (Neofonie GmbH), Carina Prunkl (University of Oxford), Martijn Willemsen (Eindhoven University of Technology). Workshop organizers: Jasminko Novak (EIPCM, University of Applied Sciences Stralsund), Ksenia Koroleva (EIPCM), Kalina Drenska (EIPCM), Lukas Pfahler (TU Dortmund), Martha Larson (Radboud University), Enny Das (Radboud University) and Katharina Morik (TU Dortmund). We also thank Anne Janssen for her work in the project during her time at Radboud University.