The England football team made its country proud when it reached the finals of the Euros last month. There has been overwhelming support for the team and its players, despite missing out on the trophy during penalty kick-outs. However, the joy of the tournament has no doubt been tainted by the abhorrent racist abuse some of the players received online during and after the final match.
Players on Twitter and Instagram saw awful racist messages on their screens the days following the final, which prompted a debate across the UK about how abusive messages should be moderated online. It's clear that the platforms in question have a responsibility to protect their users wherever possible, but as we all know, moderating social media messages at scale isn't a simple task.
There were also calls from certain quarters to introduce ID verification for social media, in the hope that this might reduce the likelihood of known users sending hateful content, but we at diginomica argued strongly against this as a realistic idea.
Twitter has today revealed how it handled the influx of online hate towards football players, which gives us a clearer idea of the direction of travel for social media moderation, as well as some insight into how the Euros was handled as a specific case.
We will get into the details shortly, but in its latest blogpost Twitter says that it is sharing the information in order to help move the debate forward with more reliable information. It notes:
We condemn racism in all its forms - our aim is to become the world's most diverse, inclusive, and accessible tech company, and lead the industry in stopping such abhorrent views being seen on our platform. We were appalled by those who targeted players from the England football team with racist abuse following the Euro 2020 Final.
Having completed an initial review of our data related to Euros 2020, we wanted to share an update publicly and transparently. Our aim is to contribute to the shared understanding of these behaviours online and provide an overview of some of the steps we put in place since our initial blogpost was published in February.
Using a proactive approach
Twitter says that in advance of the Euro 2020 tournament, it put in place plans to help quickly identify and remove racist, abusive tweets targeting the England team and the "wider Euros conversation".
What's clear is that Twitter is now heavily relying on its automated tooling to identify harmful tweets, rather than pushing full responsibility to manual approaches and moderation teams (although these are still in place). The hope is that this deals with racist abuse more quickly and effectively - and the numbers tell an interesting story. Twitter UK says:
Following the appalling abuse targeting members of the England team on the night of the Final, our automated tools, which had been in place throughout Euro 2020, kicked in immediately to identify and remove 1622 Tweets during the Final and in the 24 hours that followed.
While our automated tools are now able to detect a majority of the abusive Tweets we remove, we also continue to take action from reports. New vectors of abuse are ever-emerging, which means our system is having to adapt on an ongoing basis. Therefore, to supplement our efforts, trusted partners are able to report any further Tweets directly to our front-line enforcement teams. In total, over 90% of the Tweets we removed for abuse over this period were detected proactively.
Twitter continued to remove content that violated its terms in the days that followed and by 14th July, 1,961 tweets had been moved proactively using technology (following the final), with a total of 126 removed using the reports approach.
Putting ID verification argument to bed
As noted above, the racism online during the Euros prompted a debate about whether ID verification for social media should be made mandatory, in an attempt to curb harmful comments. This idea doesn't really stand up to scrutiny, in our opinion.
There were also claims that most of the abuse was coming from accounts outside of the UK (although very little firm data was offered to support this).
Twitter has now also revealed that almost all of the abusive comments were coming from accounts that were identifiable and from users in the UK - meaning that people spewing hate online don't appear to care if they are known or not. Twitter states:
Given the international nature of the Euro 2020 Final, it was no surprise to see that the Tweets we removed came from all over the world. However, while many have quite rightly highlighted the global nature of the conversation, it is also important to acknowledge that the UK was - by far - the largest country of origin for the abusive Tweets we removed on the night of the Final and in the days that followed.
We also wanted to better understand the users we had permanently suspended over the course of the tournament. While we have always welcomed the opportunity to hear ideas from partners on what will help, including from within the football community, our data suggests that ID verification would have been unlikely to prevent the abuse from happening - as the accounts we suspended themselves were not anonymous. Of the permanently suspended accounts from the Tournament, 99% of account owners were identifiable.
Twitter adds that the racist behaviour isn't reflective of the general sentiment of people using its platform, as the word "proud" was more often used on the day following the final than any other day this year (although doesn't mitigate the harm suffered by the players that faced racism).
Finally, Twitter outlines how it plans to improve safety for its users going forward. It says that there is "no place for racist abuse on Twitter" and it is "determined" to do what it can to stop the abhorrent views and behaviours seen on its platform. It also acknowledges that it "can do better".
It's clear that Twitter is pursuing more technology use to tackle this problem, with the hope that increased automation and the use of intelligent tools will tackle violations more quickly and effectively. This includes:
Shortly trialling the development of a new product feature that temporarily autoblocks accounts using harmful language, such that users will be stopped from being able to interact with your account
Continuing to roll-out reply prompts, which encourage people to revise their replies to tweets when it looks like the language they use could be harmful. Twitter has found that when prompted, 34% of people revised their initial reply or decided not to send their reply at all. And after being prompted once, people composed on average 11% fewer potentially harmful replies in the future.
However, Twitter finishes its post by raising a very valid point that it can continue to invest in supporting a more sophisticated response to online abuse, the problem itself is societal and the responsibility of all of us to tackle. It notes:
However, we also have to be honest that the progress we will be able to make alone would be magnified by greater interventions across the board. As long as racism exists offline, we will continue to see people try and bring these views online - it is a scourge technology cannot solve alone. Everyone has a role to play - including the government and the football authorities - and we will continue to call for a collective approach to combat this deep societal issue.