Milad Alshomary

ABB C525

College Court

Hamilton, Canada, L8S 4L8

I am a postdoctoral research fellow at the Computing and Software Department at McMaster University, with a background in Natural Language Processing (NLP) focusing on human argumentation and the explainability of machine learning models. My research explores the intersection of argumentation and explainability, driven by the increasing need to understand the behavior of AI systems. I earned my PhD in Computer Science from Paderborn University (July 2018 - December 2023). My doctoral research resulted in the dissertation titled “Audience-Aware Argument Generation,” aimed to advance the effectiveness of argument generation by emphasizing the importance of relevance, consideration of the opponent’s argument, and addressing the audience’s interests. As a postdoc at Columbia University (January 2024 - June 2026), I developed methods for authorship attribution, with a particular emphasis on making these models explainable. This work involves interpreting latent spaces to understand what aspects of style authorship attribution models capture. Most recently, at McMaster University, as Postdoctoral Research Fellow, I am investigating the mechanisms by which prompting AI to emulate a specific role can affect their subjective judgment.

Demos

Talks

Interpretability for Authorial Style Analysis. Institute for Natural Language Processing, University of Stuttgart, Germany. May 2026.
Explainability and Argumentation in the Age of AI. Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), Abu Dhabi, UAE. May 2025.
Explainability of Authorship Analysis Models. Institute of Artificial Intelligence, Leibniz University Hannover, Germany. August 2025.
Towards a Unified Model of Scholarly Argumentation. Dagstuhl Seminar 22432, Schloss Dagstuhl, Germany. October 2022.

News

Jul 1, 2026	I started a new position as a Postdoctoral Research Fellow at the Computing & Software department at McMaster Univerity, Hamilton, Canada. This position is supported by an NSERC Canada Impact+ Research Training Award (Postdoctoral). I will be working on NLP Perspectivism.
May 13, 2026	I gave a guest talk at the Institute for Natural Language Processing at Stuttgart University covering some of our work on interpretability for authorial style. The talk covered two directions. First, how we interpret the latent space of pre-trained language models to surface which stylistic features drive attribution decisions. Second, proposing two architectures that build interpretable models by design: One grounding embeddings at the word sense level, the other at the transformer layer level, making model decisions transparent by construction. Together, these approaches offer complementary paths toward trustworthy, human-readable authorship attribution.
Mar 24, 2026	I presented our paper iBERT: Interpretable Embeddings via Sense Decomposition that was accepted at the EACL 2026 conference.
Aug 26, 2025	I gave a guest talk at the Institute of Artificial Intelligence at Leibniz University Hannover (link). The talk covered my recent work on the explainability of authorship analysis models and how to make them robust for out-of-domain scenarios.
Aug 25, 2025	Excited to share that our paper, “Generalizable Analysis of Authorial Style by Leveraging All Transformer Layers,” has been accepted at EMNLP2025 as a main conference paper (link). We introduce a novel approach to address authorship analysis, showcasing state-of-the-art outcomes in out-of-domain scenarios while progressing towards more interpretable models. Looking forward to sharing more insights at the conference!
May 26, 2025	I gave a guest talk at MBZUAI on the role of explainability and argumentation in the age of AI (link)
Jan 20, 2025	Our proposal on “Art Images and AI: Latent Space Interpretability, Art History, and the Law” get funded by Columbia’s Data Science Institute Seed Funding program. (link) We will explore approaches for explainability of vision LLMs when they perform formal analysis of Art pieces.
Jan 20, 2025	My PhD. disseration recieved the the Dissertation Award 2024 of Paderborn University
Jun 1, 2024	I am co-chairing the publication committee for the (EMNLP-24 conference)
Jan 1, 2024	I started a new position at the Data Science Institute at Columbia University as Postdoctoral Research Scientist.
Dec 28, 2023	I successfully defended my Ph.D. thesis at Paderborn University (tweet). Thesis can be found here (link)
Jan 1, 2023	I am co-organizing the 10th workshop on Argument Mining (link)
Oct 23, 2022	I participated in the “Towards a Unified Model of Scholarly Argumentation” Seminar at Dagtsuhl (link)
Oct 12, 2022	Recently, I attended the COLING-22 conference to present our paper “A Dialogue Corpus for Learning to Construct Explanations” (link)
Aug 1, 2022	We are organizing a SharedTask on Identifying human values in argumentative texts at SemEval 2023 (Link)

Selected Publications

Latent Space Interpretation for Stylistic Analysis and Explainable Authorship Attribution

Milad Alshomary, Narutatsu Ri, Marianna Apidianaki, and 3 more authors

In Proceedings of the 31st International Conference on Computational Linguistics 2025
Proceedings of the 10th Workshop on Argument Mining

Milad Alshomary, Chung-Chi Chen, Smaranda Muresan, and 2 more authors

In Proceedings of the 10th Workshop on Argument Mining 2023
Semeval-2023 task 4: Valueeval: Identification of human values behind arguments

Johannes Kiesel, Milad Alshomary, Nailia Mirzakhmedova, and 4 more authors

In Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023) 2023
The Moral Debater: A Study on the Computational Generation of Morally Framed Arguments

Alshomary, Milad and El Baff, Roxanne and Gurcke, Timon and Wachsmuth, Henning

In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics 2022
Toward Audience-aware Argument Generation

Alshomary, Milad and Wachsmuth, Henning

Patterns 2021
Belief-based Generation of Argumentative Claims

Alshomary, Milad and Chen, Wei-Fan and Gurcke, Timon and Wachsmuth, Henning

In Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume 2021

Abs

When engaging in argumentative discourse, skilled human debaters tailor claims to the beliefs of the audience, to construct effective arguments. Recently, the field of computational argumentation witnessed extensive effort to address the automatic generation of arguments. However, existing approaches do not perform any audience-specific adaptation. In this work, we aim to bridge this gap by studying the task of belief-based claim generation: Given a controversial topic and a set of beliefs, generate an argumentative claim tailored to the beliefs. To tackle this task, we model the people’s prior beliefs through their stances on controversial topics and extend state-of-the-art text generation models to generate claims conditioned on the beliefs. Our automatic evaluation confirms the ability of our approach to adapt claims to a set of given beliefs. In a manual study, we additionally evaluate the generated claims in terms of informativeness and their likelihood to be uttered by someone with a respective belief. Our results reveal the limitations of modeling users’ beliefs based on their stances, but demonstrate the potential of encoding beliefs into argumentative texts, laying the ground for future exploration of audience reach.