Working paper (E-prints, Working papers and Research blog)
"You Are Orthos": An Experiment in AI-Assisted Human-Rights-Based Reasoning for Content Moderation Disputes under the Digital Services Act
GRADONI, Lorenzo; MAGNINI, Matteo
2026
 

Files


Full Text
_paper-2026-orthos.pdf
(1.27 MB)
Download

All documents in ORBilu are protected by a user license.

Send to



Details



Keywords :
Digital Services Act; Content Moderation; Human Rights; Artificial Intelligence; Large Language Models; Legal Reasoning
Abstract :
[en] This paper investigates whether contemporary Artificial Intelligence (AI) systems can assist the resolution of online content moderation disputes within the institutional framework created by Article 21 of the Digital Services Act (DSA), which enables the establishment of certified out-of-court dispute settlement bodies. It presents Orthos, an AI-assisted decision-support system designed to generate human rights-based legal reasoning for disputes between users and platforms. The paper reports the results of an experiment in which the system was applied to a dataset of moderation cases, with the aim of assessing the capacity of state-of-the-art Large Language Models (LLMs) to support human rights-based adjudication at scale. Its central question is whether contemporary LLMs can generate forms of legal reasoning sufficiently disciplined and coherent to assist human rights-based adjudication in high-volume moderation disputes.
Disciplines :
European & international law
Computer science
Author, co-author :
GRADONI, Lorenzo  ;  University of Luxembourg > Luxembourg Centre for European Law (LCEL) > LCEL Research
MAGNINI, Matteo ;  University of Luxembourg > Faculty of Science, Technology and Medicine (FSTM) > Department of Computer Science (DCS) ; Esch sur Alzette, Luxembourg
Language :
English
Title :
"You Are Orthos": An Experiment in AI-Assisted Human-Rights-Based Reasoning for Content Moderation Disputes under the Digital Services Act
Publication date :
29 March 2026
Publisher :
Zenodo, Geneva, Switzerland
Number of pages :
17
Available on ORBilu :
since 29 April 2026

Statistics


Number of views
34 (0 by Unilu)
Number of downloads
35 (1 by Unilu)

OpenCitations
 
0
OpenAlex citations
 
0

Bibliography


Similar publications



Contact ORBilu