Processes, People, and Public Accountability: How to Understand and Address Harmful Communication Online

Executive Summary

During Canada's 2019 federal election, researchers and reporters documented three forms of harmful online communication: abuse of individuals (such as the thousands of negative messages and office defacement targeting Minister Catherine McKenna), intolerance and hate toward marginalized groups in public online spaces, and coordinated support for hate in private online spaces. This report, by Chris Tenove and Heidi Tworek for the Canadian Commission on Democratic Expression, proposes a framework distinguishing key dimensions of harmful online communication, summarizes findings from a study of online abuse against political candidates in that election, analyzes international policy responses by governments and platforms, and concludes with principles for Canadian policy development.

Drawing on interviews with more than 30 candidates and campaign staff and a machine-learning analysis of a corpus of approximately one million tweets directed at candidates (trained on a manually coded sample of over 3,300 tweets), the authors estimate that about 40% of tweets at candidates were uncivil, about 16% abusive or intolerant, and about 1% explicitly hateful, threatening or potentially defamatory — implying roughly 10,000 such high-negativity messages among the million tweets studied. Negative-tweet volume varied enormously, from 125,031 negative tweets aimed at Justin Trudeau (the highest recipient) to 4,445 for the tenth-ranked candidate, Adam Vaughan, down to 242 for the 100th-ranked candidate. High-negativity messages affected candidates' sense of wellbeing and security and sometimes required engagement with police, while medium- and low-negativity messages, at high volume, were demoralizing for candidates, staff and volunteers; the relationship between gender or visible-minority status and negative messaging proved complex, with no simple correlation found.

The authors recommend that Canadian policy focus on systemic processes rather than individual pieces of content, attend to the people who perpetrate, suffer and address harm, and promote public accountability of regulators and platform companies. Concrete proposals include establishing a social media council to develop context-aware content-moderation principles, supporting civil society groups, psychological-assistance providers and content moderators who work to address harmful communication, and measuring policy efficacy through greater platform transparency and independent researcher access to data. The report concludes that while harmful communication online will never disappear entirely, policies addressing processes, people and public accountability can substantially reduce its impact both online and offline.