Slipping to the Extreme: A Mixed Method to Explain How Extreme Opinions Infiltrate Online Discussions

Abstract

Qualitative research provides methodological guidelines for observing and studying communities and cultures on online social media platforms. However, such methods demand considerable manual effort from researchers and can be overly focused and narrowed to certain online groups. This work proposes a complete solution to accelerate the qualitative analysis of problematic online speech, focusing on opinions emerging from online communities by leveraging machine learning algorithms. First, we employ qualitative methods of deep observation for understanding problematic online speech. This initial qualitative study constructs an ontology of problematic speech, which contains social media postings annotated with their underlying opinions. The qualitative study dynamically constructs the set of opinions, simultaneous with labeling the postings. Next, we use keywords to collect a large dataset from three online social media platforms (Facebook, Twitter, and Youtube). Finally, we introduce an iterative data exploration procedure to augment the dataset. It alternates between a data sampler --- which balances exploration and exploitation of unlabeled data --- the automatic labeling of the sampled data, the manual inspection by the qualitative mapping team, and, finally, the retraining of the automatic opinion classifiers. We present both qualitative and quantitative results. First, we show that our human-in-the-loop method successfully augments the initial qualitatively labeled and narrowly focused dataset and constructs a more encompassing dataset. Next, we present detailed case studies of the dynamics of problematic speech in a far-right Facebook group, exemplifying its mutation from conservative to extreme. Finally, we examine the dynamics of opinion emergence and co-occurrence, and we hint at some pathways through which extreme opinions creep into the mainstream online discourse.

Opinion Market Model: Stemming Far-Right Opinion Spread Using Positive Interventions

“…We construct the Bushfire Opinions dataset, containing 90 days of Twitter and Facebook discussions about bushfires and climate change between Nov. 1, 2019 to January 29, 2020. The Facebook postings are a subset of the So-cialSense dataset (Kong et al 2022), which was collected with the approval of the Human Research Ethics Committee of the University of Technology Sydney (approval number: ETH19-3877); we select posts and comments about bushfires and climate change (SocialSense also contains discussions around COVID-19). Using CrowdTangle 2 , we unobtrusively collected public far-right Australian Facebook discussions, identified via a digital ethnographic study (see (Kong et al 2022) and (Calderon, Ram, and Rizoiu 2024) for details).…”

“…The Facebook postings are a subset of the So-cialSense dataset (Kong et al 2022), which was collected with the approval of the Human Research Ethics Committee of the University of Technology Sydney (approval number: ETH19-3877); we select posts and comments about bushfires and climate change (SocialSense also contains discussions around COVID-19). Using CrowdTangle 2 , we unobtrusively collected public far-right Australian Facebook discussions, identified via a digital ethnographic study (see (Kong et al 2022) and (Calderon, Ram, and Rizoiu 2024) for details). We build the Twitter discussions using the Twitter Academic v2 API; we collect tweets emitted between November 1, 2019 to January 29, 2020 that mention bushfire keywords such as bushfire, arson, australiaburns (see the full list in (Calderon, Ram, and Rizoiu 2024)).…”

“…Our focus on the 2019-2020 Australian bushfires is motivated by the availability of human-annotated topics, opinions (Kong et al 2022) and stance classifiers (Ram et al 2022) trained on the same topic and timeframe. We use these classifiers to filter and label our dataset.…”

“…Moderate and Far-Right Opinion Labeling. To filter and label relevant Facebook and Twitter postings, we use the textual topic and opinion classifiers developed by Kong et al (2022), with a reported 93% accuracy in classifying Facebook and Twitter posts on bushfires and climate change. We select the following most prevalent six opinions, covering 95% of Twitter and 81% of Facebook postings: 0.…”

“…We test OMM on two real-world datasets 1 . The first contains Facebook and Twitter discussions expressing moderate and far-right opinions on bushfires and climate change (Kong et al 2022). The second captures the YouTube views and Twitter mentions for the most popular VEVO artists' songs in 2017 (Wu, Rizoiu, and Xie 2019).…”