The International AI Conference hosted by GOAL took place on April 8 and 9 2021. Due to the special circumstances caused by the corona crisis, it was held digitally. Both internal and external speakers discussed interesting issues revolving around the governance of algorithms.
The approximately 80 participants were first welcomed by Prof. Dr. Johannes Wessels, Rector of the WWU Münster, and Prof. Dr. Thomas Hoeren. Then Assoc. Prof. Dr. Sandra Wachter of the Oxford Internet Institute gave the opening lecture on how the extensive inequality and bias that characterise Western societies are inevitably embedded in the use of data for machine learning processes. In the Q&A session that followed, Prof. Wachter also emphasised that AI should become more equitable and unbiased than human decision-makers.
The lecture was then followed by Ass. Prof. Dr. Christian Djeffal from the Munich Centre for Technology in Society at the Technical University of Munich. He spoke about how digital technologies would gradually reconfigure our democracy. Afterwards, Prof. Dr. Katharina Zweig from the AAL at the University of Kaiserslautern spoke on the topic “Is Fairness by Unawareness fair?”. In addition to the question of whether protected characteristics should be included in decision-making processes, she used various practical examples to show how human discrimination is reflected in training data sets and affects ADM systems. The topics presented by Prof. Zweig and Prof. Djeffal were more comprehensively addressed by questions and suggestions from the participants.
After a short lunch break, Reuben Binns, Assoc. Professor of Human Centred Computing at Oxford University, spoke about the challenge it poses for regulators to undertake comprehensive investigations and enforcement actions against the harmful use of algorithms. Professor Binns also shed light on the use of algorithmic methods to guide and support regulatory efforts. Next, Hanna Hoffmann and Johannes Kevekordes from ITM argued in their presentation that the growing influence of algorithms on society makes it essential that citizens are able to understand and evaluate decisions that algorithms make for and about them. To this end, they shed light on the existence of a possible Right to Explanation in the GDPR and the possibilities and challenges for implementing such a right.
An exciting discussion was followed by Catharina Rudschies and Mattis Jacobs from the University of Hamburg, who reported on their recent research findings that ADM systems are best understood as dynamic and evolving socio-technical ecosystems. The final talk on the first day of the conference was given by Frank Pasquale, professor at Brooklyn Law School of the law of artificial intelligence, algorithms, and machine learning. He presented how algorithmic accountability has become a major concern for social scientists, computer scientists, journalists and lawyers over the last decade and gave some recent examples from the public debate on socially harmful behaviour of algorithmic systems.
At the end of the first day, Professor Frank Pasquale, Professor Sandra Wachter, Professor Joanna Bryson (Hertie School of Governance) and Dr. Fabian Niemann (Partner at Bird & Bird LLP) discussed the topic: “Transparency in algorithmic decisions” in a panel discussion. The discussion was moderated by Dr Nikolas Guggenberger (Yale Law School). The panelists debated, among other things, the opportunities and risks of AI, as well as the progress and hurdles that have existed for market participants from a data protection perspective since the GDPR came into force.
The second day started with a presentation by Dr. Christina Timko from the Ruhr University Bochum. She reported on their research into how app developers’ values and views harmonised with client expectations and market business models. For this purpose, they conducted interviews with software developers and explored how app developers perceive needs for governance, consumer protection and risks from applied behaviour management, including discrimination risks. Finally, they presented possible protection measures and alternative incentives and business models.
This presentation was followed by Assoc. Prof. Dr. Linnet Taylor from the Tilburg Institute for Law, Technology and Society, who spoke about the concept of fairness in relation to AI, which could be applied in a formalised way through guidelines and technology regulation. This presentation triggered a lively question and discussion session among the participants. The main controversy was whether trustworthy AI can be created more effectively through normative or human-centred concepts (human-centred AI).
Afterwards, Prof. Dr. Joanna Bryson from the Hertie School in Berlin spoke in her lecture about what she sees as an untruth in the political discourse on AI regulation: that AI is necessarily opaque. In fact, she said, artificial intelligence is not necessarily more opaque than natural intelligence; in fact, AI can be made far more transparent by design. She shed light on the technological, sociological and economic barriers to transparency, how these are influenced by AI and the digital revolution, and what governance policies should be used to counter it. Dr Carsten Orwat from Karlsruhe University – ITAS then addressed challenges associated with risk regulation of AI and ADM in his presentation.
The lunch break was followed by the presentation of Andrew Burt, Managing Partner at bnh.ai, a boutique law firm focusing on AI and analytics, and Chief Legal Officer at Immuta. He spoke about the profound privacy and security gaps in AI systems and the three implications of this trend that are not immediately obvious, which he summarised as follows: “1. Privacy is dead. 2. so is trust. 3. and you are not who you think you are.”. In the concluding discussion round, the participants mainly discussed the proposals presented to counteract the effects of the data protection and security gaps and also discussed the advantages and disadvantages of the GDPR.
The contributions from the project partners and external speakers, as well as the controversial discussions, gave the participants some new impulses and detailed insights into the opportunities and risks of artificial intelligence. Above all, the differentiated perspectives of the participants in the course of the international dimension of the conference opened up new perspectives and broadened horizons. Already during the conference, some of the experts agreed on further exchanges in the future.
The GOAL team would like to take this opportunity to thank: the BMBF and DLR for making the event possible; all speakers for their exciting presentations; the participants in the panel discussion for the lively debate; and all conference participants who enriched the event with their interested questions and contributions to the discussion.
A detailed conference report can be found here.
Video recordings of individual lectures can be found here on the GOAL Youtube channel.
You can listen to a report by Deutschlandfunk in the programme Computer und Kommunikation in the programme of 10 April 2021 (topic: “Machtfrage: Wenn künstliche Intelligenz über Menschen entscheiden”) here.