Is X Less Safe Than Other Social Media Platforms?

Elon Musk appeared at the Cannes Lions International Festival of Creativity last week and was interviewed by WPP CEO Mark Read. A typical post on LinkedIn read like what you’ve probably seen many variants of: “X is a minor advertising platform that enables (supports?) hate, misogyny, antisemitism and bigotry. It's a haven for fake news, lies and deception. Buyer beware.”

I’ve long wondered the degree to which this is true, as I’m no longer active on any social media except LinkedIn. Given that my work is at the intersection of marketing and management consulting, I dug into the topic in an analytical and impartial manner.

My ingoing hypothesis was that there is indeed hate speech on X. And everywhere else. It’s the Internet, after all. What I didn’t know is how much of it is on X relative to other platforms. Is it fair for the media and others to demonize Elon more than, say, Zuck? Ad spend on X is down by approx. 50% since Elon bought it, whereas it's grown on Meta.

Summary: I couldn’t find any numeric data to support the claim that there’s more hate speech on X than other platforms. All of them have objectionable content, with one making notable strides toward improvement (see table below).

Methodology: In looking for studies or meta-studies on the topic, I...

• Searched in my DuckDuckGo browser “Quantitative analysis of hate speech on X (formerly twitter) vs. Facebook vs. Instagram vs. Google vs. Tiktok.”

  • Read many pages and skimmed hundreds more across the first dozen articles and white papers that appeared (cited below and in the footnotes). While none showed an apples-to-apples comparison with proper quantification, I was able to stitch together a few sources to form the table below.

• Asked ChatGPT to “Quantify the amount of hate speech… (on each platform).”

Findings:

Color shading of boxes added by author. Red = bad; Yellow = better; Green = good. Blanks indicate no data available.

Here are links to the sources cited in the table:

• One of the most illuminating and potentially credible studies was by The Anti-Defamation League and Tech Transparency Project.

• A Wall Street Journal article released last week (June 20, 2024) found that Instagram Recommends Sexual Videos to Accounts for 13-Year-Olds, Tests Show. SnapChat and TikTok fared better in this regard, and yet TikTok is in the regulatory crosshairs for banning.

• Some numbers – and balanced, countervailing views – were provided in this Variety article:

“Among the Center for Countering Digital Hate’s research into Twitter, it has found that the volume of tweets containing slurs have risen by up to 202% since Musk’s takeover of Twitter.”

but…

“Twitter/X has claimed that hate speech occurs at extremely low rates on the platform and estimates that (these) impressions have declined 30% since Musk took over Twitter in October 2022… More than 99% of content users and advertisers see on Twitter is healthy. According to the company, Sprinklr’s independent model showed that daily English-language hate speech impressions are even lower than Twitter’s own model estimates. Sprinklr estimated the average daily number of hate speech impressions vs. overall impressions to be 0.003% compared with Twitter’s estimate of 0.012% for the period of Jan. 1-May 31, 2023.“

Hate speech can’t be both up 202% and down 30% unless different methodologies or timeframes are used, and there’s a subjective element to identifying and classifying hate speech or unsafe content generally. But even if the 202% figure is correct, then together with Sprinklr’s findings the amount of hate speech on X is minuscule in the absolute (up to 3-in-100,000 from 1-in-100,000) and less than Facebook's 5-in-10,000 (per ChatGPT). I couldn’t find comparable data on Meta’s platforms from Sprinklr or on NBC News.

CyberGhost took a crack at quantifying social media toxicity, defined as "spreading fake news, enabling bad behaviors and scams, promoting unhealthy comparisons, cyberbullying, trolling." While informative in that it's based on users' perceptions, take it with a grain of salt because Facebook has the most users which likely pull these numbers higher.

Implications

• Media Agencies and Consultants - Cite facts and specific examples to support your recommendations about where your clients should allocate their media dollars. If you think X is unsafe, back it up with data that’s been equally applied to other media platforms. Social proof and peer pressure are powerful forces, but a hundred people saying the same thing doesn’t make it true. Example from the LinkedIn post mentioned above: “We've managed over $1 billion in media reviews recently and not a single client wanted X.” There might be valid business and marketing reasons behind this, but the logic could also be circular and self-reinforcing.

• Marketing Executives - Do your homework. You’re entitled to your biases – and each of us has them – but remember that when making media buying decisions, it may be your budget, but it’s not your money, and you have a fiduciary responsibility to your company’s shareholders / owners to do so in the most efficient way possible.

• CEOs - Ask your marketing leader whether their budget is allocated in a way that maximizes eyeballs within your target audiences. Your job and theirs entails maximizing enterprise value at every turn.

• Business Owners - Make whatever decisions you like. It’s your prerogative to trade off business results with your personal feelings about a media company and/or its owner.

I welcome any additional data sources, inputs, and perspectives that shed light on a topic on which I may have only scratched the surface.

Additional Sources Reviewed:

1. USA Today: from Nazi propaganda to Holocaust denial, social media is pushing hate on users: study

2. Thirty years of research into hate speech: topics of interest and their evolution

3. Analyzing hate speech dynamics on (Portuguese) Twitter/X: Insights from conversational data and the impact of user interaction patterns

4. Racism, Hate Speech, and Social Media: A Systematic Review and Critique

5. Exploring Automatic Hate Speech Detection on Social Media: A Focus on Content-Based Analysis

6. Hate Speeches on Twitter and Facebook in South Asia: A Case Study Of Malala Yousufzai

7. Bridging the gap in online hate speech detection: a comparative analysis of BERT and traditional models for homophobic content identification on X/Twitter

8. A survey on hate speech detection and sentiment analysis using machine learning and deep learning models

9. Offensive, aggressive, and hate speech analysis: From data-centric to human-centered approach

10. Meta's Hate Speech Problem