Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Recommender Systems to Reduce Content Gaps on Peer Production Platforms

Mo Houtti, Isaac Johnson, Morten Warncke-Wang, Loren Terveen

TL;DR

The paper investigates whether recommender systems can reduce content gaps on peer production platforms like Wikipedia without sacrificing engagement. Through an offline Study 1 on SuggestBot and a three‑month live Experiment (Study 2), it shows that surfaceing underrepresented topics can increase edits in those areas while maintaining overall uptake, provided relevance is carefully managed. The findings highlight that discoverability processes shape editing, revealing a 'peer production filter bubble' and suggesting avenues to counteract it via diversity‑aware recommendations. Recommender systems are a valuable component of a broader strategy to improve content equity, but deeper systemic changes — including editor diversity — are necessary for lasting impact.

Abstract

Peer production platforms like Wikipedia commonly suffer from content gaps. Prior research suggests recommender systems can help solve this problem, by guiding editors towards underrepresented topics. However, it remains unclear whether this approach would result in less relevant recommendations, leading to reduced overall engagement with recommended items. To answer this question, we first conducted offline analyses (Study 1) on SuggestBot, a task-routing recommender system for Wikipedia, then did a three-month controlled experiment (Study 2). Our results show that presenting users with articles from underrepresented topics increased the proportion of work done on those articles without significantly reducing overall recommendation uptake. We discuss the implications of our results, including how ignoring the article discovery process can artificially narrow recommendations on peer production platforms.

Leveraging Recommender Systems to Reduce Content Gaps on Peer Production Platforms

TL;DR

The paper investigates whether recommender systems can reduce content gaps on peer production platforms like Wikipedia without sacrificing engagement. Through an offline Study 1 on SuggestBot and a three‑month live Experiment (Study 2), it shows that surfaceing underrepresented topics can increase edits in those areas while maintaining overall uptake, provided relevance is carefully managed. The findings highlight that discoverability processes shape editing, revealing a 'peer production filter bubble' and suggesting avenues to counteract it via diversity‑aware recommendations. Recommender systems are a valuable component of a broader strategy to improve content equity, but deeper systemic changes — including editor diversity — are necessary for lasting impact.

Abstract

Peer production platforms like Wikipedia commonly suffer from content gaps. Prior research suggests recommender systems can help solve this problem, by guiding editors towards underrepresented topics. However, it remains unclear whether this approach would result in less relevant recommendations, leading to reduced overall engagement with recommended items. To answer this question, we first conducted offline analyses (Study 1) on SuggestBot, a task-routing recommender system for Wikipedia, then did a three-month controlled experiment (Study 2). Our results show that presenting users with articles from underrepresented topics increased the proportion of work done on those articles without significantly reducing overall recommendation uptake. We discuss the implications of our results, including how ignoring the article discovery process can artificially narrow recommendations on peer production platforms.
Paper Structure (39 sections, 3 figures, 3 tables)

This paper contains 39 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Content Gaps on Wikipedia
  4. Recommender Systems on Wikipedia
  5. Fairness in Recommender Systems
  6. Background: SuggestBot
  7. Methods (Study 1)
  8. Dataset and Features
  9. Analysis
  10. Results (Study 1)
  11. Control Model
  12. Gender
  13. Geography
  14. Important Topics
  15. Observational Study to Controlled Experiment
  16. ...and 24 more sections

Figures (3)

  • Figure 1: An example set of recommendations, posted by SuggestBot to a user's Talk page. Roughly half of the recommendations are part of an experimental group, but the user is not given any indication as to which.
  • Figure 2: Diagram describing SuggestBot's process of filtering recommendations, and how we intervened in the filtering process to produce treatment groups. Black elements represent additions for the experiment. "Add Experiment Criterion" was given a random 55% chance of being yes.
  • Figure 3: Percentage of edited recommendations that belonged to underrepresented categories. Previous years are in red and study period is in blue. Graphs show consistent increases across all categories, likely due to our intervention.