Page 1 of 85
On Risk Assessment and Mitigation for
Algorithmic Systems
About the Integrity Institute
The Integrity Institute is a non-profit professional community and think tank working
to advance the theory and practice of protecting the social internet, powered
by our community of integrity professionals.
With years of experience mitigating harms to people and communities within more than 55
online platform companies, we bring seasoned, insider knowledge to leaders theorizing,
building and governing online platforms and help them put integrity front and center.
Here’s how we do it:
● We build and empower a community of integrity professionals in
● tech, giving them the tools and research they need to make online platforms safer and
healthier for people and societies.
● We advise online platforms, policymakers and academics to put integrity at the heart of
company governance, compliance and tech regulation.
● We educate the public about what an integrity-first future looks like for the social
internet.
Acknowledgements
This report was authored by Jeff Allen and Abagail Lawson, with input from members of the
Integrity Institute community. It does not reflect the views of all Integrity Institute members.
The authors are extremely grateful for the contributions from experts listed below. Their
inclusion here doesn’t mean they endorse or support any of the recommendations or
statements in this report.
Diane Chang, Sagnik Ghosh, Andrew Gruen, Spencer Gurley, David Harris, Anna Lenhart, Jenn
Louie, Swapneel Mehta, Arushi Saxena, Theodora Skeadas, Amanda Wall, Grady Ward
Page 2 of 85
Page 3 of 85
Table of Contents
Introduction and Background
Chapter 1: Design and Operation of Algorithmic Systems
Introduction
Common Components of Algorithmic Systems
Summary of different algorithmic systems used by platforms
Content Lifecycle
Unifying Framework for Content Ranking, Recommendation, and Search
Engagement Based Ranking
Quality Based Ranking
Content Moderation Systems
Advertising Systems
Design Choices
Algorithmic Systems and Risk
Engagement Based Ranking Significantly Increases the Risk of Algorithmic Systems
Page 3 of 85