The Marga Institute is currently monitoring mainstream media and social media to identify instances of harmful/hate speech. Monitoring of harmful/hate speech on social media was assigned to Popkorn (PVT) Limited. Following the planning meeting and discussions on methodology and key focus areas, Popkorn conducted preliminary monitoring between 1 September and 19 October 2020. The monitoring was conducted in two stages: first, between 1 September and 30 September, and second, between 1 October and 19 October.
In the first stage of monitoring, a broad range of keywords were included in the tracking tool, to identify the scope of the tool’s reach. It was found to be able to access posts in public pages and profiles across Twitter and Facebook, but could not access private posts in private profiles or groups on Facebook.
Key Findings
September Monitoring
The tool picked up relevant terms regardless of the context in which it was used. Thus, following the initial capture of keywords, a member of Popkorn had to individually comb through each use of the keyword to identify if it was harmful or not (i.e. positive/negative/neutral) and to what ‘sentiment’ it belonged. The stratification of sentiments is broken up into ‘racial’, ‘controversial’, ‘names’ and ‘other’. Meanwhile, the channels on which these ‘sentiments’ were featured were also recorded: (i) Twitter, (ii) Web, (iii) Facebook, (iv) Instagram, and (v) YouTube.
According to the findings of the search tool, ‘controversial’ and ‘racial’ keywords featured more prominently in the share of voice than ‘names’ and ‘other’ during the monitoring period. ‘Controversial’ keywords were featured most on Twitter and the Web, with significantly less mention on Facebook, and no mention on Instagram and YouTube. Meanwhile, key ‘names’ appeared on Twitter, the Web, Facebook and Instagram (in that order) but occupied less share of voice than ‘racial’ and ‘controversial’.
By contrast, in terms of share of impressions, ‘names’ comprised 68%, with ‘controversial’ keywords taking only 15% of the share of impressions, and ‘other’ taking 16%. ‘Racial’ keywords appear not to have made any impression.
The names monitored in the month of September 2020 were:
Rishad Bathiudeen / Rauff Hakeem / Athureliya Rathena Thero / Zahran Hashim / Thawheed
Jamath / Ali Sabry / Dilith Jayaweera / Amith Weerasinghe / Dan Priyasad / Chatura Alwis / Iraj/ Gnanasara Thero / Ampitiya Sumanarathana Thero.
These names left an impression primarily on Facebook, with far less impression made on Twitter and the Web, suggesting that engagement with these actors is highest on Facebook.
The type of ‘influencers’ varied from individual personalities – such as IRAJ, and politicians – to radio and TV channels, newspapers and gossip sites. Hiru and its associated channels (i.e. Hiru News, Hiru FM, hirufm.lk, hirunewslk) emerged leading influencers in both the ‘controversial’ and ‘other’ categories. The prominence of Hiru in these early findings is helpful in terms of identifying relevant TV channels or programmes that should be monitored alongside social media monitoring.
Overall, in September 2020, there appears to have been less than 250 negative mentions in total. This low number is likely to be due to two factors. First, methodologically, the keywords included in the search terms were not specific enough. It is likely that a number of mentions were missed due to this limitation (that has since been rectified). Second, there appears to have been a decline in overall harmful speech targeting ethnic and religious minorities. This trend can be contrasted with the pre-election period when anti-minority harmful speech was significant.
October Monitoring
For the monitoring undertaken in October, Popkorn and Marga decided to refine the tool and the keywords to adapt to the changing socio-political context in Sri Lanka. Thus, in addition to tracking mentions of longstanding ‘key issues’, such as the Easter Attacks, boycotts of minority businesses, and the cattle slaughter ban, it also monitored harmful speech relating to the COVID-19 cluster emerging in the Brandix manufacturing factory in Minuwangoda. This ability to track new and emerging issues is likely to be of use in the future, as it enables the project team to adapt and respond to a dynamic socio-political environment.
In October 2020, ‘people’ dominated both the share of voice (40%) and the share of impressions (72%). There were 299 ‘negative’ mentions of ‘people’ between 1 and 19 October. The majority of these negative statements mentioned Rishad Bathiudeen, in connection with his arrest (i.e. the initial failure to locate and arrest him).
Assuming that the tool has managed to effectively capture all mentions of the selected keywords, it appears that the trend for October reflects the trend evident in the September monitoring findings – that overall anti-minority harmful speech is not occurring at a significant level. It raises questions over the use of anti-minority harmful speech in the pre-election/pre-August period as part of electioneering by political parties and their affiliates. However, as there was no baseline study conducted by Popkorn/Marga before September, it is difficult to make claims without empirical evidence. Yet alternative monitors of anti-minority harmful speech appear to have made similar observations that anti-minority harmful speech was rife in the lead up to the elections and even before elections. Monitoring the trajectory of these trends over the next 6 months to a year will be important.