Study identifies weaknesses in how AI systems are evaluated
4 November 2025
Largest systematic review of AI benchmarks highlights need for clearer definitions and stronger scientific standards.
Latest news
OII researchers to explore LLM interpretability at EMNLP 2025
31 October 2025
Researchers, including those from the Oxford Internet Institute’s Reasoning with Machine’s Lab (OxRML), will attend the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP) in Suzhou, China from 4-9 November.
New leadership at the Oxford Internet Institute
27 October 2025
The OII is pleased to announce Dr Scott A. Hale appointed as its next Director, and Dr Kathryn Eccles as Deputy Director.
Next events
14
Nov
Enshittification and Extraction: The Internet Sucks Now
17:30 -19:00, Rewley House
With Cory Doctorow, and Prof Timothy WuJoin Tim Wu and Cory Doctorow, for a lively, infuriating and invigorating exploration of the forces that drove the internet to its current miserable state, and a prescription for taking it back and making a new, good internet that is fit for purpose
Four graduate degrees
Academic spotlight
Find out more about research by OII academics
Professor Andrew Przybylski
Professor of Human Behaviour and Technology
Professor Przybylski is a psychologist who studies how online social media and video games platforms influence users’ mental health. His research focuses on how data from these platforms can inform regulation, policymaking, and expert advice.
Professor Mark Graham
Professor of Internet Geography
Mark Graham is an economic geographer. His research focuses on digital labour, the gig economy, and digital inequalities. He is the author, most recently, of Feeding the Machine: The Hidden Human Labour Powering AI
Dr Ekaterina Hertog
Associate Professor in AI and Society
Ekaterina Hertog is an Associate Professor of AI and Society at Oxford Internet Institute and Institute for Ethics in AI. She studies how the rising digitalisation is reshaping private lives across the world.
Latest projects
Conversations between governments and citizens in a digital society
This project seeks to understand how citizens listen and speak to public institutions, and how alternative AI-based models and framing might encourage democratic communication.
The Promise of AI in Childcare
With mobile devices ubiquitous among young people, it is not surprising that parents are increasingly turning to technology for help in childcare. This project examines monitoring technologies and how parents use them as part of child supervision.
Understanding the Causes of the Reverse Personal Computing Revolution
The shift to cloud represents a reversal of the 1980s personal computing revolution: computation is moving from personal devices back into large, centralized facilities. This project seeks to understand the economic drivers of this reversal.
Latest report
By Andres Raieste, Andri Rebane, Madis Tapupere, Keegan McBride
In a world increasingly defined by crises—from pandemics and climate disasters to escalating wars—governments that fail to build digital resilience risk crumbling under pressure.
Press coverage
AI’s capabilities may be exaggerated by flawed tests, according to new study
NBC News, 06 November 2025
Researchers behind a new study say that the methods used to evaluate AI systems’ capabilities routinely oversell AI performance and lack scientific rigour.
How Sky News investigated X’s algorithm for political bias
Sky News, 06 November 2025
In recent months, X has become a focus of political conversation in the UK, with social media experts and politicians saying Elon Musk's platform is sowing political division across the country. With commentary from Dr Mohsen Mosleh.
Viral Truths & Virtual Walls: What 10 Million Posts Reveal About Us
Somewhere on Earth: The Global Tech Podcast, 05 November 2025
Listen to the OII's Dr Mohsen Mosleh talking about his latest research analysing news-sharing trends across seven major social media platforms.
Latest podcast
The science behind successful start-ups
Fabian Braesemann and Paul McCarthy discuss the impact of founder personalities on startup success, how a predictive model outperforms industry standards to predict company success, and what factors can help us predict which startups will succeed.
