Deconstructing Ad Hominem Attacks to Counter Contrarian Campaigns on Climate Scientists

Sergei Samoilenko

Advisor: Timothy Gibson, Katherine E. Rowan

Committee Members: Christopher E. Clarke, John Cook

Online Location,
July 27, 2020, 01:00 PM to 03:00 PM


The process of communicating climate science has become increasingly politicized and embedded in a conflict between two polarized worldviews. There is consensus among scientists that climate change is occurring at an alarming rate, and it is caused by human action (Cook et al., 2016; Stocker, et al., 2013). However, many in the public are confused about many elements of climate change. A large contributing factor to public confusion is misinformation about climate change, misinformation produced by various interest groups. Climate misinformation targets scientific topics, climate policy, and the reputation of scientists.

Character assassination against climate scientists is the most common type of misinformation strategy used by contrarians in climate debates (Coan, Boussalis, & Cook, in review). Despite their pervasiveness, climate-related ad hominem attacks are understudied in misinformation research. This study fills this gap by analyzing character attacks found in a corpus of 287,000 documents published in English between 2008 and 2017 in 55 contrarian blogs and websites of 15 think tanks. These materials were gathered between January 2018 and May 2018 by a team of software programmers, Coan, Boussalis, and Cook (n.d.), who wrote custom software to harvest this content, which contains over 104 million words (tokens).

The study applies Walton’s argumentation approach (Walton, 1998) to categorize these attacks and apply critical questions in response. Walton’s typology emerged in a humanistic scholarly field, that of argumentation (e.g., van Eemeren & Grootendorst, 1984), and therefore, additional steps were needed to transform this type of scholarship into a social science instrument, a codebook for establishing a reliable system for classifying ad hominem arguments with acceptable inter-rater reliability. To increase the usability of Walton’s typology for a social scientific classification of these attacks, his 21 types of ad hominem attacks were refined to three major types of attacks directed towards climate scientists: allegations of incompetence, immorality, and bias. Additionally, two structural ad hominems: inconsistent behavior (circumstantial guilt), and guilt by association were identified during an iterative process that involved two coders (the author and a second coder) coding multiple batches of character attack paragraphs.

To test Walton’s argumentation approach to analyzing ad hominem attacks, a body of 315 paragraphs was drawn from the corpus of attacks Coan et al. ( n.d.) assembled.  These attacks were assessed by two or more coders. Using Krippendorf’s alpha, inter-coder reliability of at least .60 was achieved when identifying discourse in three of these categories: attacks on scientists’ morality, alleged bias, and inconsistent behavior. Bias attacks—accusing climate scientists of political partisanship or ideological agendas—were found to be by far the most common form of contrarian ad hominem. I also find that differing types of ad hominem attacks often cluster together (e.g., allegations of immorality and bias often accompany each other) in distinct combinations.

These findings offer guidance for inoculation interventions to neutralize future character assassination campaigns on climate scientists. Said differently, the study’s findings may inform efforts to assist people in recognizing ad hominem attacks on scientists and in responding to these attacks by asking critical questions about the evidence supporting them. Also discussed are potential ways to further strengthen the study’s coding scheme and reliability assessment.