| Creators: |
Riekers, Nils and Risius, Marten and Cheng, Tong |
| Title: |
Developing Multi-Agent Systems to Address AI Explainability Issues: The Case of Online Hate Speech Detection |
| Item Type: |
Conference or Workshop Item |
| Event Title: |
20th International Conference on Wirtschaftsinformatik (WI25) |
| Event Location: |
Münster, Deutschland |
| Event Dates: |
13. – 17. September 2025 |
| Projects: |
IDI |
| Date: |
16 September 2025 |
| Divisions: |
Informationsmanagement |
| Abstract (ENG): |
Online hate speech is a severe issue for individuals and society that can be addressed automatically by AI-based information systems. Through structured, court-inspired debates between specialized agents, our developed multi-agent system enables transparent decision-making and improves the explainability of multimodal hate speech detection. The human-in-the-loop design allows users to grasp the decision-making process and supports the iterative refinement of system configurations. Initial evaluations with the Hateful Memes dataset show promising results in terms of performance (F1 score of 0.70) and transparency. Future work will mainly focus on optimizing technical performance through fine-tuning and on systematically assessing and improving the quality of explainability based on stakeholder feedback. |
| Forthcoming: |
No |
| Language: |
English |
| Uncontrolled Keywords: |
Hate Speech, Explainable AI, Multi-Agent Systems, Multimodality |
| Citation: |
Riekers, Nils and Risius, Marten and Cheng, Tong
(2025)
Developing Multi-Agent Systems to Address AI Explainability Issues: The Case of Online Hate Speech Detection.
In: 20th International Conference on Wirtschaftsinformatik (WI25), 13. – 17. September 2025, Münster, Deutschland.
|
 |
View Item in edit mode (academic staff only) |
