Pandemic recalibrates screen habits, online abuse

Spike in social media use during lockdown sparks content moderation rethink

Attacks on the physical appearance of online users is driving incivility in digital platforms for Canadians, according to a February Microsoft study | Tero Vesalainen/Getty

Best known for automating content moderation on streaming media and social media platforms, Two Hat Security Ltd. has experienced massive increases in the chat volumes it handles for other companies.

Carlos Figueiredo, the Kelowna-based tech company’s director of community trust and safety, said that prior to the pandemic lockdown Two Hat was processing about one billion interactions each day, including the exchange of chat messages and photos.

By April, that figure spiked to an average of 2.95 billion interactions a day as people turned to their screens to bide their time during the lockdown.

“The reality is that a lot of the companies up to this point before the lockdown had policies and procedures for content moderation [and had to] basically throw it all out the window and say, ‘We have to reinvent the way we do content moderation,’” Figueiredo said.

One client, for instance, reported a 3,000% jump in chat volumes from February to March.

The result is that companies are being pushed to address spikes in abusive behaviour online or else risk losing users.

A 2016 Data & Society Research Institute study, Online Harassment, Digital Abuse and Cyberstalking in America, found 21% of respondents stopped using social media after facing online harassment.

Two Hat, meanwhile, has been working primarily with gaming companies to deploy AI-powered tools to weed out inappropriate language or abusive content, such as pornographic images, on their respective social networks.

“Traditionally this has been done by humans really doing the brunt of the work and doing it manually,” said Figueiredo, whose company counts Vancouver-based developer Kabam Games Inc. among its big clients.

Last month, Two Hat’s efforts caught the attention of the World Economic Forum, which named it to its influential list of 100 top technology pioneers of 2020, often seen as a barometer for hot tech companies that have the power to address global issues.

In November 2018, the company acquired Texas-based ImageVision, which specializes in detecting pornographic images.

The integration of the two companies’ technologies has boosted its detection accuracy to 99% from 97%, Two Hat CEO Chris Priebe told Business in Vancouver last year.

But for Canada’s part, online abuse is not declining, according to a Microsoft Corp. (Nasdaq:MSFT) study conducted just before the pandemic.

The tech giant’s annual Digital Civility Index measures online risks across four categories: reputational, behavioural, sexual and personal/intrusive. 

Its 2020 global rankings, released in February, place Canada at No. 6 with a score of 60% (the lower the score, the better).

While the score is unchanged from a year earlier, Canada moved up two spots in the rankings.

The global index now stands at 70% – a four-year low.

Physical appearance (24%), race (22%), sexual orientation (21%), politics (21%) and religion (20%) were the topics driving most of Canada’s online incivility.

But not only has Two Hat had to help clients navigate through spikes in online interactions and reinventions of content moderation policies, the pandemic has amplified concerns about hate speech.

Figueiredo noted that some clients experienced an uptake in users blaming certain nationalities for the spread of COVID-19.

And in the Netherlands, experts detected an increase in users wishing others would catch the coronavirus.

“COVID-19, the term itself it’s not a concern, but when it’s used to harass others and wish disease upon others it becomes a harassment weapon,” Figueiredo said. “That’s what our clients are seeing across languages as well.”