@inproceedings{, author = {Sellami, Mahdi and Bueno Momčilović, Tomas and Kuhn, Peter and Balta, Dian}, title = {Interaction Patterns for Regulatory Compliance in Federated Learning}, booktitle = {CIISR 2023: 3rd International Workshop on Current Information Security and Compliance Issues in Information Systems Research, co-located with the 18th International Conference on Wirtschaftsinformatik (WI 2023), September 18, 2023, Paderborn, Germany}, publisher = {CEUR Workshop Proceedings}, pages = {6-18}, year = {2023}, month = sep, abstract = {Organizations in highly regulated domains often struggle to build well-performing machine learning (ML) models due to restrictions from data protection regulation. Federated learning (FL) has recently been introduced as a potential remedy, whereby organizations share local models while keeping data on premise. Still, regulatory compliance remains challenging in FL settings: training data needs to be shared to some extent, and models can be reverse engineered or misused towards violation of data privacy by each participating organization. Guided by design science methodology, we introduce four interaction patterns that allow for compliance-by-design and trust-context-sensitive analysis of an FL system by combining different approaches to privacy preservation. We match the patterns to privacy principles and exemplify how verifiable claims about compliance at design- and operation-time FL can be generated to make all participating organizations accountable.}, keywords = {Federated Learning, Privacy, Compliance, Design Patterns}, url = {https://ceur-ws.org/Vol-3512/fullpaper1.pdf}, } @conference{, author = {K{\"{u}}mpel, Michaela and Buchinger, Matthias and Sellami, Mahdi and Balta, Dian and Beetz, Michael}, title = {Trust, But Verify: Towards Trustworthiness in Digital Assistants Based on Verifiable Claims in Knowledge Graphs}, publisher = {AIC’23: 9th workshop on Artificial Intelligence and Cognition}, year = {2023}, month = sep, abstract = {Along the process of digital transformation, retailers are increasingly using digital shopping assistants to support their customers with additional services from electronic shopping carts to click & collect services. Such assistants need large amounts of data that needs to be imported and linked from decentralised sources such as from the Web. Unfortunately, this leads to trustworthiness challenges. While enabling a digital assistant to appraise information reliability would allow consumers to decide about the trust level of the queried data, there is a lack of practical approaches with verifiable appraisal. In this work, we propose a concept and system architecture for realising an assessment of trustworthiness of knowledge graph content and underpinning it by verifiable claims. We represent entity specific data sources in a knowledge graph that connects data from multiple sources and use a trust service based on formal verification and immutable logging. We further demonstrate the architecture through a prototypical implementation of a digital shopping assistant with verifiable claims about the trustworthiness of the queried data.}, keywords = {Trustworthiness, Verifiable Claims, Semantic Web, Knowledge Graph, Digital Assistant}, } @misc{https://doi.org/10.48550/arxiv.2202.12443, author = {Baracaldo, Nathalie and Anwar, Ali and Purcell, Mark and Rawat, Ambrish and Sinn, Mathieu and Altakrouri, Bashar and Balta, Dian and Sellami, Mahdi and Kuhn, Peter and Buchinger, Matthias}, title = {Towards an Accountable and Reproducible Federated Learning: A FactSheets Approach}, publisher = {arXiv}, year = {2022}, month = feb, owner = {arXiv.org perpetual, non-exclusive license}, abstract = {Federated Learning (FL) is a novel paradigm for the shared training of models based on decentralized and private data. With respect to ethical guidelines, FL is promising regarding privacy, but needs to excel vis-{\`{a}}-vis transparency and trustworthiness. In particular, FL has to address the accountability of the parties involved and their adherence to rules, law and principles. We introduce AF^2 Framework, where we instrument FL with accountability by fusing verifiable claims with tamper-evident facts, into reproducible arguments. We build on AI FactSheets for instilling transparency and trustworthiness into the AI lifecycle and expand it to incorporate dynamic and nested facts, as well as complex model compositions in FL. Based on our approach, an auditor can validate, reproduce and certify a FL process. This can be directly applied in practice to address the challenges of AI engineering and ethics.}, doi = {10.48550/ARXIV.2202.12443}, keywords = {Artificial Intelligence (cs.AI), Machine Learning (cs.LG), FOS: Computer and information sciences, FOS: Computer and information sciences}, url = {https://arxiv.org/abs/2202.12443}, } @inproceedings{, author = {Balta, Dian and Sellami, Mahdi and Kuhn, Peter and Sch{\"{o}}pp, Ulrich and Buchinger, Matthias and Baracaldo, Nathalie and Anwar, Ali and Ludwig, Heiko and Sinn, Mathieu and Purcell, Mark and Altakrouri, Bashar}, title = {Accountable Federated Machine Learning in Government: Engineering and Management Insights}, booktitle = {Electronic Participation - 13th IFIP WG 8.5 International Conference, ePart 2021}, publisher = {Springer}, series = {Lecture Notes in Computer Science}, pages = {125-138}, year = {2021}, month = sep, abstract = {Machine learning offers promising capabilities to improve administrative procedures. At the same time, adequate training of models using traditional learning techniques requires the collection and storage of enough training data in a central place. Unfortunately, due to legislative and jurisdictional constraints, data in a central place is scarce and training a model becomes unfeasible. Against this backdrop, federated machine learning, a technique to collaboratively train models without transferring data to a centralized location, has been recently proposed. With each government entity keeping their data private, new applications that previously were impossible now can be a reality. In this paper, we demonstrate that accountability for the federated machine learning process becomes paramount to fully overcoming legislative and jurisdictional constraints. In particular, it ensures that all government entities' data are adequately included in the model and that evidence on fairness and reproducibility is curated towards trustworthiness. We also present an analysis framework suitable for governmental scenarios and illustrate its exemplary application for online citizen participation scenarios. We discuss our findings in terms of engineering and management implications: feasibility evaluation, general architecture, involved actors as well as verifiable claims for trustworthy machine learning.}, isbn = {978-3-030-82824-0}, doi = {10.1007/978-3-030-82824-0_10}, url = {https://www.doi.org/10.1007/978-3-030-82824-0_10}, } @inproceedings{, author = {Balta, Dian and Kuhn, Peter and Sellami, Mahdi and Kalogeropoulos, Anastasios and Krcmar, Helmut}, title = {Blackbox AI: What is in the Box?}, booktitle = {Proceedings of ongoing research, practitioners, posters, workshops, and projects of the international conference egov-cedem-epart}, publisher = {Shefali Virkar, Olivier Glassey, Marijn Janssen, Peter Parycek, Andrea Polini, Barbara Re, Peter Reichst{\"{a}}dter, Hans Jochen Scholl, Efthimios Tambouris}, volume = {2019}, pages = {245–247}, year = {2019}, month = sep, location = {San Benedetto Del Tronto, Italy}, abstract = {We present insights from a chatbot prototype for online citizen participation and discuss particular benefits and caveats of artificial intelligence (AI) application in the government domain. We present an argument that AI represents a blackbox not only in terms the reasoning process itself, but also in terms of applying different building blocks “out-of-the-box”. Our research shows that customizing an AI application involves a hardly manageable combination of buildings blocks: various machine learning techniques, data sources and user interaction designs. Since those building blocks might not fit the requirements of a particular use case, the required effort to configure the expected behaviour should not be underestimated.}, url = {https://biblio.ugent.be/publication/8626904/file/8626906.pdf}, } @inproceedings{, author = {Balta, Dian and Kuhn, Peter and Sellami, Mahdi and Kulus, Daniel and Lieven, Claudius and Krcmar, Helmut}, title = {How to Streamline AI Application in Government? A Case Study on Citizen Participation in Germany}, booktitle = {Electronic Government}, publisher = {Lindgren, Ida and Janssen, Marijn and Lee, Habin and Polini, Andrea and Rodr{\'{i}}guez Bol{\'{i}}var, Manuel Pedro and Scholl, Hans Jochen and Tambouris, Efthimios}, pages = {233--247}, year = {2019}, month = sep, owner = {Springer International Publishing}, address = {Cham}, abstract = {Artificial intelligence (AI) technologies are on the rise in almost every aspect of society, business and government. Especially in government, it is of interest how the application of AI can be streamlined: at least, in a controlled environment, in order to be able to evaluate potential (positive and negative) impact. Unfortunately, reuse in development of AI applications and their evaluation results lack interoperability and transferability. One potential remedy to this challenge would be to apply standardized artefacts: not only on a technical level, but also on an organization or semantic level. This paper presents findings from a qualitative explorative case study on online citizen participation in Germany that reveal insights on the current standardization level of AI applications. In order to provide an in-depth analysis, the research involves evaluation of two particular AI approaches to natural language processing. Our findings suggest that standardization artefacts for streamlining AI application exist predominantly on a technical level and are still limited.}, isbn = {978-3-030-27325-5}, doi = {10.1007/978-3-030-27325-5_18}, keywords = {Natural language processing, Standardization, Government}, url = {https://link.springer.com/chapter/10.1007%2F978-3-030-27325-5_18}, } @inproceedings{, author = {Balta, Dian and Sellami, Mahdi and Kuhn, Peter and Krcmar, Helmut}, title = {Insights from Natural Language Processing}, booktitle = {Proceedings of ongoing research, practitioners, posters, workshops, and projects of the international conference egov-cedem-epart}, publisher = {Shefali Virkar, Olivier Glassey, Marijn Janssen, Peter Parycek, Andrea Polini, Barbara Re, Peter Reichst{\"{a}}dter, Hans Jochen Scholl, Efthimios Tambouris}, volume = {2019}, pages = {241–243}, year = {2019}, month = sep, location = {San Benedetto Del Tronto, Italy}, abstract = {We present an exemplary text categorization pipeline for online citizen participation and aim at discussing our ongoing research in terms of insights from natural language processing (NLP) application. For each of the steps in the categorization pipeline, we share our experience in terms of challenges and potential measures to address these challenges.}, url = {https://biblio.ugent.be/publication/8626904/file/8626906.pdf}, } @article{, author = {Balta, Dian and Krcmar, Helmut and Kuhn, Peter and Kulus, Daniel and Sellami, Mahdi}, title = {Digitalgest{\"{u}}tzte B{\"{u}}rgerbeteiligung & KI}, journal = {PLANERIN}, volume = {2019}, number = {1}, pages = {19--22}, year = {2019}, month = feb, timestamp = 2019.10.24, abstract = {Digitalisierung und k{\"{u}}nstliche Intelligenz (KI) gelten als pr{\"{a}}gende Themen und wecken eine hohe Aufmerksamkeit in Gesellschaft, Wirtschaft, Politik und Verwaltung. Im Bereich B{\"{u}}rgerbeteiligung bergen Digitalisierung und in letzter Zeit auch KI eine Reihe von Chancen und Herausforderungen. F{\"{u}}r die Beteiligungspraxis stellen sich daher u. a. folgende Fragen: Welche Aufgaben kann KI in Beteiligungsverfahren {\"{u}}bernehmen? Wie gut sind solche Systeme heute schon und welche Qualit{\"{a}}ten haben die erzielten Ergebnisse? Welche Fortschritte sind zu erwarten? Der Artikel gibt erste Antworten basierend auf Untersuchungen und dem Einsatz von Prototypen im Rahmen von B{\"{u}}rgerbeteiligungsverfahren in Hamburg und schlie{\ss}t mit einem Ausblick in die Zukunft.}, issn = {0936-9465}, url = {https://www.srl.de/publikationen/planerin/%C3%A4ltere-ausgaben/product/view/2/150.html}, }