DFS2026MSCA DIGITAL Network Workshop at the European Central BankFrankfurt am Main
European Central Bank, Frankfurt am Main
Monday 30 November 202610:00–16:10 CETOnsiteFree (grant-funded)
| 10:00–10:15 | Registration & Welcome Coffee |
|---|---|
| 10:15–10:30 | Opening RemarksTheodoros Mastrokostopoulos (ECB) and Jörg Osterrieder (University of Twente) Welcome by ECB and DIGITAL Network representatives |
| 10:30–11:15 | Keynote 1: The Role of Reference Datasets and Biases of LLMs for Banking and InvestmentProf. Charles-Albert Lehalle, CMAP, École Polytechnique, Institut Polytechnique de Paris Invited keynote (45 min) Abstract and biography |
| 11:15–12:05 | Research Session 1: AI/ML for Bank Supervision and Nature-Related Risk Chair: Codruta Mare, Babes-Bolyai University. Two invited talks, 30 minutes and 20 minutes
|
| 12:05–13:00 | Lunch Break & Networking |
| 13:00–13:20 | Invited Talk: AI vs Humans: what changes when machines make the financial decisionsDr Roman Matkovskyy, Associate Professor, Rennes School of Business Invited talk (20 min) Abstract and biography |
| 13:20–15:10 | Research Session 2: NLP, SupTech and Industry Perspectives Chair: Maria Iannario, University of Naples Federico II. Three contributed talks (20 minutes each) and two invited talks (30 and 20 minutes)
|
| 15:10–15:40 | Coffee Break |
| 15:40–16:00 | Invited Talk: Do Banks Respond to Their Friends’ Markets? Social Spillovers in Deposit PricingSofia Anyfantaki, European Central Bank, Financial Research Division; Bank of Greece Invited talk (20 min) Abstract and biography |
| 16:00–16:10 | Closing RemarksJörg Osterrieder, University of Twente |
Extended abstract, 2–4 pages (excluding references), PDF format. Use single-column layout, 11pt font, standard margins. Include author names, affiliations, and a short abstract (max 200 words).
All submissions will be reviewed by the organizing committee based on relevance, novelty, and potential impact on supervisory practice. Authors of accepted papers will be invited to give a 15-minute presentation. We particularly encourage submissions from PhD students and early-career researchers.
Submit your extended abstract as a PDF attachment via email to joerg.osterrieder@utwente.nl with the subject line 'DFS2026 Submission'. PhD students should indicate their PhD program and expected completion date.
| Abstract submission closes | 1 October 2026 | in 28 days |
|---|---|---|
| Notification to authors | 15 October 2026 | in 42 days |
| Camera-ready abstract | 1 November 2026 | in 59 days |
| Registration closes | 1 November 2026 | in 59 days |
| Workshop | 30 November 2026 | in 88 days |
Application of artificial intelligence and machine learning techniques to enhance banking supervision and regulatory oversight.
Natural language processing applied to regulatory documents, compliance monitoring, and supervisory communication.
Large language models and generative AI for supervisory processes and regulatory technology.
The environmental footprint of AI infrastructure, and the measurement of nature-related financial risk that supervisors are beginning to ask about.
Crypto markets, digital-asset interdependencies and the systemic risk they carry into the regulated financial system.
Dear colleagues and friends, it is our great pleasure to welcome you to the Digital Finance for Supervision workshop, hosted at the European Central Bank in Frankfurt am Main on 30 November 2026. The day is held on site, with speakers and participants together in the room, and brings together ECB supervision experts with researchers from the MSCA DIGITAL network to explore how artificial intelligence, machine learning and natural language processing are transforming the landscape of banking supervision.
The DIGITAL network, funded by the European Union's Horizon Europe programme under the Marie Sklodowska-Curie Actions, unites 18 partner institutions from across Europe, including universities, central banks, financial institutions and technology companies. Our shared mission is to advance the frontiers of digital finance research while training the next generation of researchers who can navigate both the technical and regulatory dimensions of financial innovation.
We are particularly grateful to the ECB for hosting this event and for the active engagement of ECB staff in both the organisation and the intellectual programme. We encourage you to engage actively in the discussions, connect with fellow researchers and practitioners, and explore opportunities for collaboration.
Digital Finance for Supervision is an onsite half-day workshop organised by the MSCA DIGITAL network at the European Central Bank in Frankfurt am Main. Speakers and participants attend in person for the whole day. The workshop brings together researchers and practitioners to explore how AI, machine learning and NLP can transform banking supervision.
The event features one keynote presentation, six invited talks, and three contributed research talks. We particularly welcome contributions from PhD students and early-career researchers.
Attendance is Free (grant-funded), capacity 30-50 participants. Registration closes 1 November 2026.
For questions about submissions, registration or the workshop, write to joerg.osterrieder@utwente.nl. The network site is at digital-finance-msca.com.
European Central Bank
Sonnemannstrasse 20, 60314 Frankfurt am Main
Frankfurt am Main, Germany
The workshop runs 10:00 to 16:10 CET. Onsite at the European Central Bank in Frankfurt am Main; speakers and participants attend in person.
10:30–11:15
Prof. Charles-Albert Lehalle, CMAP, École Polytechnique, Institut Polytechnique de Paris
About the speaker. Member of the Scientific Committee of the AMF, the French financial markets regulator. Co-director, with Prof. Vianney Perchet, of the MScT programme AI for Markets and Quantitative Investment. Leads the Finance Vertical of the AI Factory project (European Commission / France / GENCI) together with Prof. Damien Challet, where they are launching a research project on biases in large language models.
11:15, 30 minutes
Prof. Ania Zalewska, University of Leicester School of Business
Artificial intelligence and nature conservation are both becoming defining priorities, and their relationship is far from straightforward. AI promises productivity, innovation and growth, and can help fight climate change through remote sensing, digital monitoring and other data-intensive technologies that make ecological change more observable, biodiversity risks more measurable and environmental outcomes more verifiable. Yet AI’s expansion requires substantial physical infrastructure, particularly data centres, with growing demands for land, energy and water. At the same time, biodiversity loss and climate change are increasing the pressure to protect land, restore ecosystems and reduce demands on natural resources. The pursuit of a digital future could therefore come into direct competition with the pursuit of a nature-positive one. A key question is where data centres should be built: on greenfield sites, or should brownfield land be prioritised? What happens when brownfield sites themselves have significant biodiversity value? And how should environmental costs be weighed against regeneration, employment and other socioeconomic benefits? These questions ultimately become local decisions. National and European governments can set ambitious biodiversity, planning and sustainability policies, but implementation depends on regional authorities and politicians. Does the quality of regional authorities and politicians really matter, and if so, which qualities?
11:45, 20 minutes
Prof. Huei-Wen Teng, National Yang Ming Chiao Tung University, Taiwan
About the speaker. Associated partner of the MSCA DIGITAL network. Her work applies machine learning to credit card default prediction and evaluates tail risk under Basel standards.
13:00–13:20
Dr Roman Matkovskyy, Associate Professor, Rennes School of Business
About the speaker. Associate Professor of Finance and Geopolitics at Rennes School of Business, and a Management Committee member of the CA19130 FinAI COST Action. His research covers digital finance, cryptocurrency markets, systemic risk and financial market interdependencies.
13:40, 30 minutes
Ólöf Jónsdóttir, Managing Director of Personal Banking, Íslandsbanki
About the speaker. Managing Director of Personal Banking at Íslandsbanki, bringing extensive senior leadership experience from the Icelandic financial sector.
14:10, 20 minutes
Prof. Tian-Shyr Dai, National Yang Ming Chiao Tung University, Taiwan
Catastrophe bond markets display persistent regularities in contract design. Earthquake bonds almost always adopt indemnity triggers while windstorm bonds favor industry loss indices; nearly every private-sponsor issue is intermediated by a bankruptcy-remote special purpose vehicle; and principal is typically only partially at risk. Whether these conventions are economically rational has not been quantified. This paper develops a bilateral valuation framework that prices a catastrophe bond from both sides of the transaction. Using contingent-claim analysis, we compute the minimum spread at which investors will purchase at par and the maximum spread at which the cedent's shareholders remain no worse off than without hedging. The interval between the two, the trading region, measures the scope for mutually beneficial trade, and its width serves as a proxy for bilateral welfare. The model couples a structural credit-risk specification with compound Poisson catastrophe losses and stochastic interest rates, is solved on a trinomial tree, and is calibrated from disclosed attachment points, attachment probabilities, and expected losses rather than from sparse historical loss data. Applying the framework to trigger mechanisms, loss calculation methods, SPV intermediation, and funding structure, we find that industry-index triggers widen the trading region by roughly eight percent relative to indemnity triggers, and that SPV isolation widens it by twenty to sixty percent depending on cedent leverage. Full funding generates wider regions in the baseline model yet is rare in practice, a gap we reconcile through regulatory capital efficiency and asymmetric capital costs. Across 145 bonds issued between 2020 and 2025, provisions associated with wider trading regions are those the market most frequently adopts, suggesting that observed designs reflect bilateral welfare optimization rather than arbitrary convention.
About the speaker. Tian-Shyr Dai received his Ph.D. from the Department of Computer Science, National Taiwan University. He chaired the Department of Information Management and Finance from 2016 to 2019 and directed the Taiwan Association of Business School from 2018 to 2020. He is currently Full Professor in the Department of Information Management and Finance at National Yang Ming Chiao Tung University, and a Research Member of the Risk and Insurance Research Center at NCCU. He has been a Senior Fellow of AdvanceHE and a Faculty Member of Beta Gamma Sigma since 2021. His research interests include financial engineering and financial technology.
14:30, 20 minutes
Dennis Hoffmann, University of Twente
About the speaker. PhD candidate at the University of Twente, Faculty of Behavioural, Management and Social Sciences, in collaboration with Quoniam Asset Management.
14:50, 20 minutes
Dr Lennart Baals, University of Applied Sciences of the Grisons (FHGR)
About the speaker. Researcher at the University of Applied Sciences of the Grisons, and a member of Working Groups 1, 2 and 3 of the CA19130 FinAI COST Action. His work concerns market microstructure and the behaviour of ultra-fast traders.
15:40–16:00
Sofia Anyfantaki, European Central Bank, Financial Research Division; Bank of Greece
We study how deposit rate shocks transmit across banking markets through digital social ties. Depositors’ inattention implies that households react to outside rate changes only when social networks make these changes salient, inducing connected banks to raise their own rates. Using merger-driven shocks to local deposit rates and county-level social connectedness, we show that small banks increase rates in response to shocks occurring in socially linked but geographically distant counties. Spillovers are economically meaningful, persistent, and stronger in competitive markets and in counties with more financially sophisticated households. Digital social ties therefore activate depositor search and integrate deposit markets across space.