What can social data tell us about
cyberbullying and hate speech online?

We’ve joined forces with Brandwatch to analyze 19 million Tweets over a four-year period to explore the current climate of cyberbullying and hate speech online.

As a digital anti-bullying charity, we wanted to better understand the dynamics and nature of bullying online so that we can continue to evolve and develop the amount of support available on our website.

This infographic is best viewed on a device with a larger screen, or you can download the full report.

Cyberbullying and Hate Speech

What can social data tell us about
cyberbullying and hate speech online?

1,460 Days19,000,000

We’ve joined forces with Brandwatch to analyze 19 million Tweets over a four-year period to explore the current climate of cyberbullying and hate speech online.

For the first time ever, we’ve used social data from Twitter to track and measure the rates of cyberbullying and hate speech used online. Our previous research shows that 6 in 10 young people will, at some point, experience cyberbullying.

The issue continues to grow and as a digital anti-bullying charity, we wanted to better understand the dynamics and nature of bullying online so that we can continue to evolve and develop the support services we offer.

Please be aware that some of the data presented in this infographic is in raw format and in places is uncensored. We recommended that anybody below the age of 16 has parental consent before exploring this infographic.

Framing the Global Issue

The Internet is a powerful tool for connecting communities and people together. However, unfortunately the Internet can sometimes be used for insidious purposes. It is now possible for somebody to attack another person online without ever meeting them, or even being in the same country.

This infographic explores online abuse from across the UK and USA, but we acknowledge that bullying is a global problem. This report provides just a fraction of an insight into a wider, growing issue.

Geographic data

The data points shown in the map represent 3% of the body of hate speech and bullying dialogue captured by Brandwatch in the USA and UK.

Cyberbullying: a very short introduction

By tracking the context and frequency of abusive terms used, we were able to paint an accurate picture of abuse – whilst also enabling us to account for culturally reclaimed words. This data makes it possible to track future abuse in real-time, which may enable us to better target our interventions.

1
19,170
2
17,871
3
17,297
4
16,819
5
16,510
6
15,981
7
11,510
8
10,366
9
9,651
10
8,411
11
8,142
12
7,013
Cyberbullying – key terms
1 Bitch 5 Idiot 9 Piece of shit
2 Dumb 6 Stupid 10 Punk
3 Fucking 7 LMAO 11 Hate
4 Fat/Ugly 8 Moron 12 Nigga

What is cyberbullying?

Cyberbullying is classified as generalized abuse, largely towards the appearance, interests, intelligence or previous posts of the recipient. Cyberbullying is not a separate issue from traditional bullying; rather an expansion of the platforms on which it exists.

How is it defined?

Bullying online and offline is a subjective phenomenon – what some people consider friendly banter others may construe as abusive. To define abuse while taking account of this subjectivity, we tracked both keywords and context, identifying abusive terms used with malicious intent.

When does cyberbullying happen?

Each week, Ditch The Label helps thousands of young people overcome bullying through the pioneering support services available freely on our website. By understanding how the prevalence of abuse varies over time, we can better tailor and coordinate campaigns and support services to match demand.

Social data for provisioning

The power of social data comes from the fact that it is observational and objective – enabling us to be scientific and targeted in our approach to providing cutting-edge advice and support to help counteract the damage of online abuse.

Which topics of conversation were most likely to attract abuse?

Discussion of certain topics was much more likely to attract abuse – most notably politics, sports, food and music. However, the data shows bullying occurs in response to a wide range of conversation subjects, showing that it is not limited to one or two niche interest groups. This data suggests a need for stronger tolerance and respect for the preferences of others.

Topics
1 Politics 7 Gaming
2 Sports 8 Health
3 Food 9 TV
4 Music 10 Hair
5 Bullying 11 Relationships
6 Response 12 Travel

While some divisive topics such as politics emerged, overall we found a broad range of topics preceded bullying. There is no simple cue found to trigger trolling on Twitter.

What happens when people respond to cyberbullying?

To conclude the section on cyberbullying, we explored the relative frequency of outcomes experienced by recipients who respond to their aggressors. The data shows that it really is true: responding to abuse can often only embolden the attacker.

Key

Negative
Neutral
No reply
Positive

Methodology

Positive vs negative outcomes were assessed through manual categorization of a random sample. Each conversation was allocated a sentiment based on how the discussion closed.

Example negative outcome:

Bullying author:

You fat f*ck dog sh*t idiot loser

Recipient:

You're a loser

Bullying author:

I've always been better than you, my life is better, you're a loser

Bullying author:

That's one ugly haircut

Recipient:

Dude, I have the same haircut as you...

Bullying author:

I don't get haircuts

Bullying author:

You're one dumb b*tch

Recipient:

Omg what did I even do to you?

Bullying author:

[did not respond]

Bullying author:

Ladies, you're both stupid af

Recipient:

Stick to Pokemon and YouTube, little guy

Bullying author:

Ok, I will

As one of the world’s largest anti-bullying charities, we’re here for you. Getting instant advice and support is just a click away.

Hate speech: a very short introduction

In the following sections, we will explore the phenomenon of racism, homophobia, and transphobia on Twitter, including data on how it has evolved over time and the authors who are generating it. This data will help us create and target new awareness campaigns, and will feed into the support that we provide.

1
128,066
2
61,821
3
53,590
4
49,545
5
47,999
6
39,627
7
33,467
8
31,491
9
29,688
10
29,375
11
26,690
12
10,288
13
8,397
14
4,529
15
1,975
16
1,753
17
1,651
18
1,240
19
1,032
20
164
21
94
22
5
Hate speech – key terms
1 Faggot 9 Black 17 Die
2 Bitch 10 LMAO 18 Lady boy
3 Nigga 11 Fuck 19 Shemale
4 Dirty 12 Tranny 20 Gender bender
5 Fucking 13 Dyke 21 He-she
6 Hoe 14 Lesbian 22 She-he
7 White 15 Gay
8 Pussy 16 Dumb

What is hate speech?

For the purpose of this research, hate speech is defined as abuse that directly targets a unique factor beyond the control of the recipient. We recognize that there is a broad spectrum of hate speech; in this research we focus on race, sexuality and gender identity.

How does hate speech differ from cyberbullying?

Hate speech is differentiated from cyberbullying in being defined as abuse directed specifically toward a unique, non-controllable attribute of a group of people.

The phenomenon and types of hate speech on Twitter

According to the data, racism is the most prevalent form of hate speech on Twitter – far surpassing homophobia and transphobia. Over time, discussions of all types of hate speech have increased, which indicates an encouraging increase in awareness of these issues.

Racism

Discussion volume 7,010,192
Abuse volume 7,681,280
Ratio 1.1:1

Homophobia

Discussion volume 3,142,496
Abuse volume 390,296
Ratio 0.12:1

Transphobia

Discussion volume 455,942
Abuse volume 19,348
Ratio 0.04:1

Note: This study measures explicit prejudiced language. Be aware that not all prejudice is explicit (for example, derogatory but non-gendered comments about trans people) and this may impact some category volumes more than others. Detecting implicit bias online is therefore an important avenue for further research.

How has volume changed over time?

Four years of data has enabled us to track the volume of hate speech on twitter alongside the rate of discussion of those same issues, over a long time period. Discussion has increased greatly over the time frame, while the growth in abuse varied between types.

Brexit
Alton Sterlin
#RNCinCLE
Sochi 2016 Games
Orlando Shootings
Joan Rivers calls Michelle Obama
a ‘Tranny’
Discrimination against Caitlyn Jenner

Key

Abuse
Discussion

We have plotted this data against real life events such as Brexit and the USA elections to consider what influence they may have had upon rates of racist, homophobic and transphobic tweets.

Who’s doing it? Understanding professions and interests

We explored the professions and interests of users most likely to perpetrate hate speech. By understanding authors, we are able to delve deeper into their psychology and motivations to tailor campaign materials and help prevent the issues.

Abuse ratio

Key

Abuse
Discussion

We recognize that there is a huge volume of data regarding authors of hate speech posted on the Twitter platform. This does not mean that everybody from a specific profession or with a certain interest is using Twitter to be abusive, but it does help us identify groups with higher levels of perpetration.

Select hate speech type

Select breakdown

Who’s doing it? Understanding location and political affiliation

In the Annual Bullying Survey, we consistently find that the rates and nature of bullying behaviors vary from region to region, due to factors such as population make-up, socioeconomics, access to quality education, and local culture. This data shows that these regional variations are equally apparent online.

Regional varations in dialog

In these graphics, the brightest red areas display a higher volume of insensitive language compared to constructive debate; darker areas have a high volume of debate and a comparatively lower volume of abusive language.

Please note that areas that are more sparsely populated are relatively low volume and that data of this kind should be interpreted with caution.

Select hate speech type

Select country

24 hours365 days

Nobody should be subjected to any type of bullying, harassment or abuse and we’re here for anybody who needs advice and support.

Click here to ask us a question and we’ll do our best to point you in the right direction.

As an independent charity we are reliant upon the generosity and support of people like you who believe in a world that is fair, equal and free from all types of bullying. It costs an average of £3 to help us positively change the life of one young person. Please help us take a stand by donating to keep our services open and available to even more young people in need. Donate now.

This research takes us a step closer to beating bullying once and for all.

We use data, science, and psychology to proactively prevent bullying from happening and to help those who are affected it. We provide a crucial and life-changing support line for thousands of young people every week through our pioneering digital support programs and campaigns.

This data will help us continue to innovate and improve the support that we provide across all of our markets; primarily the UK, USA and Mexico.

Find out more: brandwatch.com