I write a blog about recommender systems with Pri Bengani and Jonathan Stray. We try to make it rigorous and accessible. Lots of people seem to like it.

How they work ↗ Preferences ↗ Engagement ↗ Amplification (1) ↗ Measurement ↗ Transparency ↗ Influence ↗ Filter Bubbles ↗ Amplification (2) ↗ Search ↗ Diversity ↗ Experiments ↗

Collaboration + Opportunities

Increasingly, I have more worthy research opportunities open to me than I have the bandwidth for, often with funding attached. If you are interested in collaborating or having offers of work passed in your direction, please get in touch!

Reference Works

Definitions Metaphors Places to publish

All

Bridging in Social Media Feeds Censors Controversial Topics ↗

Toward Democracy Levels for AI ↗

Views on AI aren't binary — they're plural ↗

The Moral Case for Using Language Model Agents for Recommendation ↗

Societal Sorting as a Systemic Risk of Recommenders ↗

Views on AI aren't binary — they're plural ↗

Places to publish

Underexplored Ways to Regulate Social Media ↗

Mapping the space of social media regulation ↗

Experiments are the Best Kind of Transparency ↗

Research Agenda for Sociotechnical Approaches to AI Safety ↗

Recommender Systems Very Different From Ours

How to redesign social media algorithms to bridge divides ↗

Bridging Systems ↗

What is Media Diversity and Do Recommender Systems Have It? ↗

What's the Difference Between Search and Recommendation? ↗

SHAPE: A Framework for Evaluating the Ethicality of Influence ↗

Error in the Euclidean Preference Model ↗

Submission to Australian Government ↗

Evaluating recommender systems in relation to the dissemination of illegal and harmful content in the UK ↗

Polarization as Probabilistic Dependence ↗

Error in the Euclidean Preference Model ↗

Making Amplification Measurable ↗

When You Hear “Filter Bubble”, “Echo Chamber”, or “Rabbit Hole” — Think “Feedback Loop” ↗

Submission to UNESCO ↗

Is Optimizing for Engagement Changing Us? ↗

Optimizing Language Models for Argumentative Reasoning ↗

Algorithmic Risks to Collective Decision-Making ↗

A Menu of Recommender Transparency Options ↗

How to Measure the Effects of Recommenders ↗

Submission to International Observatory on Information and Democracy ↗

What Will “Amplification” Mean in Court? ↗

The IC Rating Scale as a Measure of Analytic Rigor ↗

What's Right and What's Wrong with Optimizing for Engagement ↗

What Does it Mean to Give Someone What They Want? ↗

Using argument mapping to improve clarity and rigour in written intelligence products ↗

How Platform Recommenders Work ↗

The Impact of Evidence on Decision Making ↗

Towards Argument Embeddings ↗

Analytic Rigour in Intelligence ↗

Charting our way back to normal ↗

Enhancing human analysis of large corpora via argument mining and backlinks—a prototype ↗

Discourse-tionary

Classification of failure modes in public discourse

Disagreement is asymmetric

Tools to Ease the Choice and Design of Protein Crystallisation Experiments ↗

Metaphors for public discourse

Fractional Brownian motion

Using Anki for mathematics

What is a probability?

What's good is bad

License + Disclaimer

This site is licensed under the Creative Commons public domain (CC-0) license, but attribution is appreciated. Views expressed are my own, and probably don't coincide with those of any entity with which I am affiliated.