A public research hub connecting projects, papers, datasets, and artifacts around:
- agentic AI
- synthetic media and disinformation
- human perception and cognitive vulnerability
- provenance and verification
- information integrity on the Web
This repository is the top-level entry point for a broader research program. It does not duplicate the individual project repositories. Instead, it provides:
- an overview of how the projects fit together
- a curated index of research artifacts
- mobile-friendly landing pages in
docs/ - a stable QR-friendly destination for conference materials
An empirical platform for measuring how humans judge AI-generated news on dual axes: source attribution and authenticity.
- Repo: https://github.com/aloth/JudgeGPT
- Demo: https://judgegpt.streamlit.app/
- Preprints: https://arxiv.org/abs/2601.21963, https://arxiv.org/abs/2601.22871, https://arxiv.org/abs/2404.03021
- Zenodo: https://doi.org/10.5281/zenodo.18703384
A controlled stimulus generation engine for synthetic news and misinformation research.
- Repo: https://github.com/aloth/RogueGPT
- Preprints: https://arxiv.org/abs/2601.21963, https://arxiv.org/abs/2601.22871, https://arxiv.org/abs/2404.03021
- Zenodo: https://doi.org/10.5281/zenodo.18703137
A privacy-first mobile framework for on-device image provenance verification using C2PA and related signals.
- Repo: https://github.com/aloth/origin-lens
- App: https://apps.apple.com/us/app/origin-lens/id6756628121?platform=iphone
- Paper (WebSci '26): https://doi.org/10.1145/3795513.3806658
An expert survey on GenAI disinformation threats, reproducible provenance, and mitigation strategy preferences.
- Repo: https://github.com/aloth/verification-crisis
- Survey: https://forms.gle/BCwYFtfqxmZewkL97
- Preprint: https://arxiv.org/abs/2602.02100
- Zenodo: https://doi.org/10.5281/zenodo.18703600
A credibility-oriented dataset and infrastructure layer for source-level information integrity research.
- Repo: https://github.com/aloth/cred-1
- Dataset: https://doi.org/10.5281/zenodo.18769459
- Preprint / paper page: https://doi.org/10.2139/ssrn.6448466
This repo is intended to host:
- landing pages for talks, posters, and symposium presentations
- cross-links to published papers and preprints
- links to datasets and demos
- project overviews and navigation pages
- future research synthesis pages
- a stable QR-friendly destination for already published and online research outputs
GitHub Pages content is stored in docs/.
Current pages:
/- main hub/websci2026/- WebSci materials/judgegpt/- JudgeGPT landing page/origin-lens/- Origin Lens landing page/cred-1/- CRED-1 dataset landing page, with the Trackless Links production integration and the WebSci 2026 free-codes campaign/trackless-links/- Trackless Links landing page: first production deployment of CRED-1, on-device credibility warnings, app and code links/phd-symposium/- PhD Symposium page
- RogueGPT: https://doi.org/10.5281/zenodo.18703137
- JudgeGPT: https://doi.org/10.5281/zenodo.18703384
- Expert Survey / verification-crisis: https://doi.org/10.5281/zenodo.18703600
- CRED-1: https://doi.org/10.5281/zenodo.18769459
- Blessing or Curse? A Survey on the Impact of Generative AI on Fake News: https://arxiv.org/abs/2404.03021
- Industrialized Deception: The Collateral Effects of LLM-Generated Misinformation on Digital Ecosystems: https://arxiv.org/abs/2601.21963
- Eroding the Truth-Default: A Causal Analysis of Human Susceptibility to Foundation Model Hallucinations and Disinformation in the Wild: https://arxiv.org/abs/2601.22871
- The Verification Crisis: Expert Perceptions of GenAI Disinformation and the Case for Reproducible Provenance: https://arxiv.org/abs/2602.02100
- Origin Lens: Reclaiming Trust on the AI-Mediated Web Through On-Device Image Provenance Verification: https://doi.org/10.1145/3795513.3806658
- CRED-1: An Open Multi-Signal Domain Credibility Dataset for Automated Pre-Bunking of Online Misinformation: https://doi.org/10.2139/ssrn.6448466
This repository is designed to make the broader research agenda legible:
When synthetic media becomes indistinguishable from authentic content, trust must increasingly be supported by infrastructure rather than intuition alone.
Alexander Loth
Frankfurt University of Applied Sciences
GitHub: https://github.com/aloth
Website: https://alexloth.com/