Segun Aroyehun

@uni-graz.at featured our brand-new article on collective moderation of online spaces published in @pnasnexus.org 🥳 Wanna read what citizen-based strategies can be a remedy against online hate? Click here: ow.ly/csZK50XGteK @janalasser.bsky.social @aroyehuns.bsky.social @dgarcia.eu

Linguistic and Embedding-Based Profiling of Texts Generated by Humans and Large Language Models Sergio E. Zanotto, Segun Aroyehun. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Cool findings! It aligns very with results from our recent work on a similar topic: aclanthology.org/2025.emnlp-m...

How does human language differ from texts generated by large language models? #UniKonstanz @excinequality.bsky.social study explored this question – and shows that while AI writes with increasing fluency, it could at the same time endanger the diversity of human expression: https://t1p.de/xo0x0

Computational analysis of US congressional speeches reveals a shift from evidence to intuition - Nature Human Behaviour This research explores the linguistic traces of evidence-based reasoning and intuitive decision-making in congressional speeches from 1879 to 2022. The analysis suggests that evidence-based language h...

In our most recent study, which just appeared in Nature Human Behaviour with Segun Aroyehun as lead author (doi.org/10.1038/s415...) , we drew a wider historical arc and examined the rhetoric in all speeches given on the floor of Congress between 1879 and 2022.
16/n

How politicians communicate shapes online discourse in ways we might overlook.

Our new paper shows that their choice between a fact-based (evidence-driven) and a belief-based (sincerity-driven) honesty creates a "contagion" effect, influencing how users engage and respond. ⬇️(1/8)