Designing an Intervention Tool for End-User Algorithm Audits in Personalized Recommendation Systems

TL;DR

This paper presents MapMyFeed, a new tool that enables everyday users to audit personalized recommendation algorithms for bias and fairness, addressing literacy and noise challenges to promote transparency.

Contribution

It introduces a novel intervention tool with guiding prompts, audit tracking, and visualization features to facilitate user-driven algorithm audits in recommendation systems.

Findings

Supports user literacy and awareness of algorithms

Enhances transparency and fairness in recommendations

Addresses noise and unstructured audit paths

Abstract

As algorithms increasingly shape user experiences on personalized recommendation platforms, there is a growing need for tools that empower end users to audit these algorithms for potential bias and harms. This paper introduces a novel intervention tool, MapMyFeed, designed to support everyday user audits. The tool addresses key challenges associated with user-driven algorithm audits, such as low algorithm literacy, unstructured audit paths, and the presence of noise. MapMyFeed assists users by offering guiding prompts, tracking audit paths via a browser extension, and visualizing audit results through a live dashboard. The tool will not only foster users' algorithmic literacy and awareness but also enhance more transparent and fair recommendation systems.