Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Auditing Elon Musk's Impact on Hate Speech and Bots

Daniel Hickey, Matheus Schmitz, Daniel Fessler, Paul Smaldino, Goran Muric, Keith Burghardt

TL;DR

The paper investigates whether Elon Musk's 2022 Twitter acquisition elevated hate speech and bot activity. It combines a Reddit-derived hate lexicon with Perspective API to detect hateful tweets and uses Botometer to assess bot prevalence, conducting timelines of hateful users and year-long hate tweet volumes. Results indicate substantial increases in both hate speech and bot scores post-acquisition, with astroturf bots declining in some control groups, though the study refrains from claiming causal effects of specific policies. These findings highlight potential risks to platform safety and underscore the need for causal analyses and robust moderation strategies in the evolving Twitter environment.

Abstract

On October 27th, 2022, Elon Musk purchased Twitter, becoming its new CEO and firing many top executives in the process. Musk listed fewer restrictions on content moderation and removal of spam bots among his goals for the platform. Given findings of prior research on moderation and hate speech in online communities, the promise of less strict content moderation poses the concern that hate will rise on Twitter. We examine the levels of hate speech and prevalence of bots before and after Musk's acquisition of the platform. We find that hate speech rose dramatically upon Musk purchasing Twitter and the prevalence of most types of bots increased, while the prevalence of astroturf bots decreased.

Auditing Elon Musk's Impact on Hate Speech and Bots

TL;DR

The paper investigates whether Elon Musk's 2022 Twitter acquisition elevated hate speech and bot activity. It combines a Reddit-derived hate lexicon with Perspective API to detect hateful tweets and uses Botometer to assess bot prevalence, conducting timelines of hateful users and year-long hate tweet volumes. Results indicate substantial increases in both hate speech and bot scores post-acquisition, with astroturf bots declining in some control groups, though the study refrains from claiming causal effects of specific policies. These findings highlight potential risks to platform safety and underscore the need for causal analyses and robust moderation strategies in the evolving Twitter environment.

Abstract

On October 27th, 2022, Elon Musk purchased Twitter, becoming its new CEO and firing many top executives in the process. Musk listed fewer restrictions on content moderation and removal of spam bots among his goals for the platform. Given findings of prior research on moderation and hate speech in online communities, the promise of less strict content moderation poses the concern that hate will rise on Twitter. We examine the levels of hate speech and prevalence of bots before and after Musk's acquisition of the platform. We find that hate speech rose dramatically upon Musk purchasing Twitter and the prevalence of most types of bots increased, while the prevalence of astroturf bots decreased.
Paper Structure (10 sections, 3 figures, 1 table)

This paper contains 10 sections, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Hate Speech and Content Moderation
  4. Bots on Twitter
  5. Methods
  6. Hateful tweet extraction
  7. Hateful user timeline collection
  8. Bot detection
  9. Results
  10. Discussion

Figures (3)

  • Figure 1: Hateful users increased their hate speech after Musk bought Twitter. Proportion of hate words in hateful users' tweets over time before and after the purchase. The inset plot represents the average daily hate speech before and after the takeover. Black lines in inset plot represent standard errors; p-value calculated from a Mann-Whitney U test.
  • Figure 2: Overall hate speech on Twitter increased after Elon Musk bought the platform. Number of hateful tweets sampled during each week in 2022, compared to a baseline of tweets collected during the same time period. The spike in March, 2022 coincides with the Canada convoy protests.
  • Figure 3: Mean Botometer scores before and after Musk's purchase. Black lines represent standard errors. P-values are calculated from the two-sided Mann-Whitney U test. *:$p < 0.05$, **:$p < 0.01$, ***:$p < 0.001$, ****:$p < 0.0001$