Online games, particularly in the competitive MOBA scene, are breeding ground for some really toxic interactions. I’ve been on the receiving end of any number, and — to be fair — have been guilty of the occasional tirade myself. It’s a problem that’s seen as intractable, a feature of the community where players need to “suck it up” and “deal with” if they’re going to play. Riot Games, however, have taken a different approach, and their studies suggests the issue is far from unmanageable.

In a post on Re/Code, lead game designer of social systems for League of Legends Jeffery Lin highlights the work and studies Riot Games have been conducting over a three year period. “As we spend more and more of our time online, we need to acknowledge that online harassment and toxicity is not an impossible problem, and that it is a problem worth spending time on.”

Some of the results are genuinely surprising. In analysing the interactions of their 67 million players, Lin states that if you plot players on a spectrum of positive or negative players in terms of their general sentiment, 87% of the  toxic behaviour — in the form of abuse, slurs, etc. — came from neutral or positive players in isolated incidents. That’s incredible, considering that any casual player of a MOBA will tell you, at least anecdotally, that abuse is a constant facet of the game. Despite online toxicity coming mainly from positive to neutral groups, they also identified that pairing negative players against each other results in a escalation of negative behavior.

Riot’s answer was initially to engage the community in the form of a public tribunal, where games reported as being toxic would have their match and chat logs made available for public comment. The community could discuss, rank and identify aspects they considered toxic or abusive. The majority of North American players railed against hate speech evident in the logs, which was an encouraging indicator that opening the administrative process to other players was a good step.

Riot then turned to machine learning – at its most basic definition, an algorithm that attempts to identify patterns and adjust its criteria for learning these on the fly – to help it identify phrases across multiple languages and deal with more subtle nuances of communication online, such as passive-aggressive behaviour and language that aided in “conflict resolution”.

Finally, to further refine the process, Riot made the data set available for external researchers, who helped identify the evolution of online language used in the game over time, collaborative behaviour and the link between online toxicity and age – surprise, there is none.

This has resulted in a governance system that provides real-time consequences and adjusts its learning behaviour based on feedback from the community, whether it’s “honouring” a player for good behaviour or reporting negative activity. Players have questioned the effectiveness of the system, but as Lin explained in a Reddit thread on the topic, there are various factors at play – for example, reports with a low weight due to coming from reporters who provided inaccurate reports in the past to a rolling ban queue resulting in bans only being enacted much later.

The effort of the past three years has returned results. According to Lin, ” As a result of these governance systems changing online cultural norms, incidences of homophobia, sexism and racism in League of Legends have fallen to a combined 2 percent of all games. Verbal abuse has dropped by more than 40 percent, and 91.6 percent of negative players change their act and never commit another offense after just one reported penalty.”

While some of the definitions are rather vague — for example, what exactly counts as “abusive” language or how toxicity is defined — it highlights excellent, proactive work being done on the part of Riot Games in an industry that currently suffers a serious issue with abuse across both games and people involved in games on social media platforms, and those saying that managing such communications is “impossible” should take note. But what about you? Recently played League of Legends? Noted an improvement? Edumacate us in the comments!

Source: Re/Code, Reddit

More stuff like this: