Vital IntelliGence to Investigate ILlegAl DisiNformaTion

Projektinformationen

VIGILANT

ID Finanzhilfevereinbarung: 101073921

DOI

10.3030/101073921

Projekt abgeschlossen

EK-Unterschriftsdatum 21 September 2022

Startdatum 1 November 2022

Enddatum 31 Oktober 2025

Finanziert unter

• Civil Security for Society

Gesamtkosten

€ 3 725 130,00

EU-Beitrag

€ 3 376 604,50

Koordiniert durch

THE PROVOST, FELLOWS, FOUNDATION SCHOLARS & THE OTHER MEMBERS OF BOARD, OF THE COLLEGE OF THE HOLY & UNDIVIDED TRINITY OF QUEEN ELIZABETH NEAR DUBLIN
Ireland

CORDIS bietet Links zu öffentlichen Ergebnissen und Veröffentlichungen von HORIZONT-Projekten.

Links zu Ergebnissen und Veröffentlichungen von RP7-Projekten sowie Links zu einigen Typen spezifischer Ergebnisse wie Datensätzen und Software werden dynamisch von OpenAIRE abgerufen.

Leistungen

Documents, reports (8)

Project Management Documentation v2

Production of v2 of Project Handbook, Risk Management Plan, Quality Assurance Plan, Legal and Security Compliance Plan to serve as guidance and ensure consistency across consortium partners. The updated version is to reflect the evolution of the project.

DC&E Evaluation Report v1

The DC&E evaluation report is a written report evaluating the success and failures of the various actions undertaken in the DC&E plan v1.

Ethics guidelines for PAs

This deliverable will provide guidance to decision-makers in PAs for the adoption of VIGILANT. It will present the results of the ethics impact assessment based on the framework developed in T2.1 and generate suggestions for mitigating the residual ethics impact of VIGILANT, focusing specifically on technical aspects. The results of the ethics assessment will also inform the Disinformation Response Manual, which will take into account the broader societal and legal context and suggest possible normative accompanying measures to ensure that the use of VIGILANT is in line with democracy and the rule of law.

Causes, contents and consequences model

This deliverable will produce insights from various fields of the social and behavioural sciences that will be assimilated so that VIGILANT is designed to reflect a whole-of-society (rather than merely technological) understanding of disinformation campaigns. This includes, amongst other factors, analysis of online misogyny and gendered disinformation (National Democratic Institute, 2020), as part of online violence against women. Through literature review, conceptual models will be developed to represent key elements of disinformation campaigns, such as creation, intent, dissemination, content characteristics, audience characteristics, subsequent effects, contextual influences and relevant interventions. The elements in these models will be used to support development of the analysis tools in WP4 and WP5, especially the Impact analysis tool (T5.1) and the Intervention support tool (T5.2).

Project Management Documentation v1

Production of v1 of Project Handbook, Risk Management Plan, Quality Assurance Plan, Legal and Security Compliance Plan to serve as guidance and ensure consistency across consortium partners.

Ethics framework

The Ethics Assessment Framework will be developed to be tailored to the specific VIGILANT needs to enable the consortium members to develop the system in accordance with the most up-to-date ethics standards and to ensure that the developed platform has the highest possible positive ethical impact with minimal ethical risks. The Ethics framework will follow the principles set out by the High-Level Expert Group on AI (of which Maria Bielikova of KInIT is a member). It will especially focus on: 1) Autonomy and human agency, considering that the VIGILANT platform provides automated tools to support human decision making; 2) Privacy and data protection issues; 3) Transparency and interpretability of the automated analysis results; 4) Fairness, diversity and non-discrimination related to profiling activities; 5) Societal well-being and democracy, especially considering potential chilling effects on basic freedoms such as freedom of information and expression. Finally, dual use considerations will play an important role in the assessment of the VIGILANT platform, including potential issues regarding export control restrictions in line with the most recent EU normative developments.

DC&E Plan v1

DC&E Plan: The Dissemination, Communication and Exploitation Plan v1 is a written report detailing of a series of activities that the VIGILANT consortium will undertake to maximise the impact and longevity of the VIGILANT Innovation Action. It will include the intended actions we plan to take as a consortium to disseminate the research undertaken in the project among specific communities of security researchers and PAs to advocate their joining the Community of Early Adopters. It will also include specific actions to communicate what VIGILANT is to members of the public. Lastly it will include specific actions which we will take to leverage the many technologies in VIGILANT for commercialisation and future research projects. This report will cover M1 - M36 of the project.

DC&E Plan v2

DC&E Plan: The Dissemination, Communication and Exploitation Plan v2 is an update to the v1 report. It is designed to reflect any changes in the DC&E plan as the project evolves. This report will focus on M18 - M36 of the project.

Data Management Plan (2)

Data Management Plan v2

Second version of Data Management Plan which will have evolved over the project and will reflect the current status of the project's handling of data (M18).

Data Management Plan v1

First version of data management plan.

Veröffentlichungen

Conference proceedings (18)

IMGTB: A Framework for Machine-Generated Text Detection Benchmarking

Autoren: Michal Spiegel, Dominik Macko
Veröffentlicht in: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations), 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.ACL-DEMOS.17

MULTITuDE: Large-Scale Multilingual Machine-Generated Text Detection Benchmark

Autoren: Macko, Dominik; Moro, Robert; Uchendu, Adaku; Lucas, Jason Samuel; Yamashita, Michiharu; Pikuliak, Matúš; Srba, Ivan; Le, Thai; Lee, Dongwon; Simko, Jakub; Bielikova, Maria
Veröffentlicht in: Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2023.EMNLP-MAIN.616

Effects of diversity incentives on sample diversity and downstream model performance in LLM-based text augmentation

Autoren: Jan Cegin, Branislav Pecher, Jakub Simko, Ivan Srba, Maria Bielikova, Peter Brusilovsky
Veröffentlicht in: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.ACL-LONG.710

MultiSocial: Multilingual Benchmark of Machine-Generated Text Detection of Social-Media Texts

Autoren: Dominik Macko, Jakub Kopál, Robert Moro, Ivan Srba
Veröffentlicht in: Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2025.ACL-LONG.36

A Large-Scale Comparative Study of Accurate COVID-19 Information versus Misinformation

Autoren: Mu, Yida; Jiang, Ye; Heppell, Freddy; Singh, Iknoor; Scarton, Carolina; Bontcheva, Kalina; Song, Xingyi
Veröffentlicht in: 2023
Herausgeber: ICWSM TrueHealth 2023
DOI: 10.48550/ARXIV.2304.04811

Fighting Randomness with Randomness: Mitigating Optimisation Instability of Fine-Tuning using Delayed Ensemble and Noisy Interpolation

Autoren: Branislav Pecher, Jan Cegin, Robert Belanec, Jakub Simko, Ivan Srba, Maria Bielikova
Veröffentlicht in: Findings of the Association for Computational Linguistics: EMNLP 2024, 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.FINDINGS-EMNLP.644

NEXT: An Event Schema Extension Approach for Closed-Domain Event Extraction Models

Autoren: Elena Tuparova, Petar Ivanov, Andrey Tagarev, Svetla Boytcheva, Ivan Koychev
Veröffentlicht in: 2023
Herausgeber: INCOMA Ltd., Shoumen, Bulgaria
DOI: 10.26615/978-954-452-089-2_016

KInIT at SemEval-2024 Task 8: Fine-tuned LLMs for Multilingual Machine-Generated Text Detection

Autoren: Michal Spiegel, Dominik Macko
Veröffentlicht in: Proceedings of the 18th International Workshop on Semantic Evaluation (SemEval-2024), 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.SEMEVAL-1.84

A Ship of Theseus: Curious Cases of Paraphrasing in LLM-Generated Texts

Autoren: Nafis Irtiza Tripto, Saranya Venkatraman, Dominik Macko, Robert Moro, Ivan Srba, Adaku Uchendu, Thai Le, Dongwon Lee
Veröffentlicht in: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.ACL-LONG.357

Disinformation Capabilities of Large Language Models

Autoren: Ivan Vykopal, Matúš Pikuliak, Ivan Srba, Robert Moro, Dominik Macko, Maria Bielikova
Veröffentlicht in: Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.ACL-LONG.793

It’s All About In-Context Learning! Teaching Extremely Low-Resource Languages to LLMs

Autoren: Yue Li, Zhixue Zhao, Carolina Scarton
Veröffentlicht in: Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2025.EMNLP-MAIN.1502

Authorship Obfuscation in Multilingual Machine-Generated Text Detection

Autoren: Dominik Macko, Robert Moro, Adaku Uchendu, Ivan Srba, Jason S Lucas, Michiharu Yamashita, Nafis Irtiza Tripto, Dongwon Lee, Jakub Simko, Maria Bielikova
Veröffentlicht in: Findings of the Association for Computational Linguistics: EMNLP 2024, 2024
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2024.FINDINGS-EMNLP.369

GateNLP at SemEval-2025 Task 10: Hierarchical Three-Step Prompting for Multilingual Narrative Classification

Autoren: Iknoor Singh, Carolina Scarton, Kalina Bontcheva
Veröffentlicht in: 2025
Herausgeber: Association for Computational Linguistics

SheffieldVeraAI at SemEval-2023 Task 3: Mono and Multilingual Approaches for News Genre, Topic and Persuasion Technique Classification

Autoren: Wu, Ben; Razuvayevskaya, Olesya; Heppell, Freddy; Leite, João A.; Scarton, Carolina; Bontcheva, Kalina; Song, Xingyi
Veröffentlicht in: 2023
Herausgeber: Association for Computational Linguistics
DOI: 10.5281/ZENODO.8159066

UKElectionNarratives: A Dataset of Misleading Narratives Surrounding Recent UK General Elections

Autoren: Jisun An, Yu-Ru Lin, Yelena Mejova, Eni Mustafaraj, Juhi Kulshrestha, Ingmar Weber
Veröffentlicht in: 2025
Herausgeber: AAAI Press
DOI: 10.5281/ZENODO.15228283

KInITVeraAI at SemEval-2023 Task 3: Simple yet Powerful Multilingual Fine-Tuning for Persuasion Techniques Detection

Autoren: Hromadka, Timo; Smolen, Timotej; Remis, Tomas; Pecher, Branislav; Srba, Ivan
Veröffentlicht in: Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023), 2023
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2023.SEMEVAL-1.86

Label Set Optimization via Activation Distribution Kurtosis for Zero-Shot Classification with Generative Models

Autoren: Yue Li, Zhixue Zhao, Carolina Scarton
Veröffentlicht in: Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, 2025
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2025.EMNLP-MAIN.1617

Analysing State-Backed Propaganda Websites: a New Dataset and Linguistic Study

Autoren: Heppell, Freddy; Bontcheva, Kalina; Scarton, Carolina
Veröffentlicht in: 2023
Herausgeber: Association for Computational Linguistics
DOI: 10.18653/V1/2023.EMNLP-MAIN.349

Other (6)

Authorship Obfuscation in Multilingual Machine-Generated Text Detection

Autoren: Macko, Dominik; Moro, Robert; Uchendu, Adaku; Srba, Ivan; Lucas, Jason Samuel; Yamashita, Michiharu; Tripto, Nafis Irtiza; Lee, Dongwon; Simko, Jakub; Bielikova, Maria
Veröffentlicht in: 2024
Herausgeber: TBC
DOI: 10.48550/ARXIV.2401.07867

Comparison between parameter-efficient techniques and full fine-tuning: A case study on multilingual news article classification

Autoren: Razuvayevskaya, Olesya; Wu, Ben; Leite, João A.; Heppell, Freddy; Srba, Ivan; Scarton, Carolina; Bontcheva, Kalina; Song, Xingyi
Veröffentlicht in: PLOS ONE, Ausgabe 19(5), 2024
Herausgeber: Zenodo
DOI: 10.1371/JOURNAL.PONE.0301738

A Ship of Theseus: Curious Cases of Paraphrasing in LLM-Generated Texts

Autoren: Tripto, Nafis Irtiza; Venkatraman, Saranya; Macko, Dominik; Moro, Robert; Srba, Ivan; Uchendu, Adaku; Le, Thai; Lee, Dongwon
Veröffentlicht in: 2023
Herausgeber: TBC
DOI: 10.48550/ARXIV.2311.08374

Disinformation Capabilities of Large Language Models

Autoren: Vykopal, Ivan; Pikuliak, Matúš; Srba, Ivan; Moro, Robert; Macko, Dominik; Bielikova, Maria
Veröffentlicht in: 2023
Herausgeber: TBC
DOI: 10.48550/ARXIV.2311.08838

Does Explanation Matter? An Exploratory Study on the Effects of Covid 19 Misinformation Warning Flags on Social Media

Autoren: Barman, Dipto; Conlan, Owen
Veröffentlicht in: 2023
Herausgeber: TBC
DOI: 10.48550/ARXIV.2309.16305

IMGTB: A Framework for Machine-Generated Text Detection Benchmarking

Autoren: Spiegel, Michal; Macko, Dominik
Veröffentlicht in: 2023
Herausgeber: TBC
DOI: 10.48550/ARXIV.2311.12574

Peer reviewed articles (2)

The disinformation lifecycle: an integrated understanding of its creation, spread and effects

Autoren: Kimberley Kruijver, Neill Bo Finlayson, Beatrice Cadet, Sico van der Meer
Veröffentlicht in: Discover Global Society, Ausgabe 3, 2025, ISSN 2731-9687
Herausgeber: Springer Science and Business Media LLC
DOI: 10.1007/S44282-025-00194-5

Are Publicly Available (Personal) Data 'Up for Grabs' ? A Discussion of Three Privacy Arguments

Autoren: Elisa Orrù
Veröffentlicht in: SSRN Electronic Journal, 2024, ISSN 1556-5068
Herausgeber: Elsevier BV
DOI: 10.2139/SSRN.5042634

Book chapters (1)

A Lightweight Approach for User and Keyword Classification in Controversial Topics

Autoren: Ahmad Zareie, Kalina Bontcheva, Carolina Scarton
Veröffentlicht in: Lecture Notes in Computer Science, Social Networks Analysis and Mining, 2025
Herausgeber: Springer Nature Switzerland
DOI: 10.1007/978-3-031-78538-2_21