Detecting Stance in Media On Global Warming

Detecting Stance in Media on Global Warming Yiwei Luo1 Dallas Card2 Dan Jurafsky1,2 Stanford University 1Department of Linguistics 2Department of Computer Science {yiweil, dcard, jurafsky}@stanford.edu Abstract Citing opinions is a powerful yet understud- ied strategy in argumentation. For example, an environmental activist might say, “Lead- ing scientists agree that global warming is a serious concern,” framing a clause which af- firms their own stance (that global warming is serious) as an opinion endorsed ([scientists] agree) by a reputable source (leading). In con- trast, a global warming denier might frame the same clause as the opinion of an untrustworthy source with a predicate connoting doubt: “Mis- taken scientists claim [...].” Our work studies opinion-framing in the global warming (GW) debate,1 an increasingly partisan issue that has received little attention in NLP. We introduce Global Warming Stance Dataset (GWSD), a dataset of stance-labeled GW sentences, and train a BERT classifier to study novel aspects of argumentation in how different sides of a de- bate represent their own and each other’s opin- ions. From 56K news articles, we find that sim- ilar linguistic devices for self-affirming and opponent-doubting discourse are used across GW-accepting and skeptic media, though GW- skeptical media shows more opponent-doubt. We also find that authors often characterize sources as hypocritical, by ascribing opinions expressing the author’s own view to source en- tities known to publicly endorse the opposing view. We release our stance dataset, model, and lexicons of framing devices for future work on opinion-framing and the automatic de- tection of GW stance. 1 Introduction Ascribing opinions to other people is a power- ful yet understudied strategy in argumentation. 1Throughout, we use the term debate to refer to the ex- istence of contrasting opinions about GW expressed in the media; it is important to emphasize that there is virtually 100% consensus among scientists regarding the reality of an- thropogenic global warming (Powell, 2017). SOURCE PREDICATE OPINION Few researchers Climate experts Most Americans believe claim tend to agree humans have negligible impact on the climate man-made greenhouse gases are responsible for global warming the report exaggerates climate change risks Figure 1. Examples of SOURCE, PREDICATE, and OPINION components, and within components, exam- ples of affirming and doubting framing devices. For example, an environmental activist might say, “Leading scientists agree that global warming is serious,” whereas a global warming denier could say, “Mistaken scientists claim that global warm- ing is serious.” In both these examples, the em- bedded clause (that global warming is serious) is presented as an opinion belonging to a source entity (scientists). However, differences in the choice of predicate (agree vs. claim) and in how the source is described lead to very different interpretations. We henceforth refer to the use of such [ENTITY] [EXPRESS] [STATEMENT] sentences as opinion- framing, and to the respective components as the SOURCE, PREDICATE, and OPINION (see Fig. 1). Despite its pervasiveness in argumentative dis- course, opinion-framing is understudied as a per- suasive strategy. This paper studies opinion- framing in the media coverage of global …

Content truncated to 500 words