A Computational Approach to Assessing Rhetorical Effectiveness: Agentic Framing of Climate Change in the Congressional Record, 1994–2016

ABSTRACT

The goal of this paper is to consider rhetorical effects as the propagation of rhetorical expressions across large sets of texts, measured by the extent to which rhetorical expressions, structures, or practices become replicated in texts and sites of rhetorical in(ter)vention. The paper draws on lines of scholarship in the digital humanities and computational rhetoric – primarily, sequential structuring of semantic contexts, semantic parsing of unstructured text, and diachronic tracking of textual expressions – to extend their conceptual and methodological insights into a computational framework for assessing rhetorical effectiveness. It offers a test case for this concept through an analysis of how Congress has framed human agency toward addressing climate change.

KEYWORDS:

• Climate change
• digital humanities
• computational rhetoric
• rhetorical effects

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1. For similar systems perspectives on scientific rhetorical practices, see Derkatch (2018) and Keränen (2010).

2. This systems-view of communication is why I take prevalence rather than salience as a measure of rhetorical effectiveness, and measures of rhetorical effectiveness as a resource for communicative practice. To follow Luhmann (1992), prevalence of rhetorical expressions creates “redundancy,” a “memory that can be called on by many persons” (p. 254). These redundancies calcify into (communicative) institutions that can “guarantee acceptability” (i.e., reduce the risk of rejection) of utterances made by individual communicators; they identify positions “of connection for the next communication that can now either build on an already attained consensus or seek dissent” (pp. 255–256). Therefore, although a novel, inventive, or particular act of communication can, of course, be effective in intervening in a life-world if its content and structure is particularly salient to the moment, from a broader use-perspective, those acts of communication cannot be abstracted into a resource for others to use in other moments, nor can they be leveraged as points of critique for general practices of communication.

3. For a more complete discussion of effects in rhetoric, see Kiewe and Houck’s (2015) introduction to The Effects of Rhetoric and the Rhetoric of Effects.

4. See also Campbell and Jamieson (1978) on the “recurrence” of rhetorical forms, as an index of “human action,” the “influence of conventions on the responses of rhetors,” of “rhetorical options or commonplaces,” as “insight into the human condition,” and as making accessible (rhetorical) “affinities and traditions across time (pp 26, 27); or Greene (1998) in a critical/materialist turn, in which “rhetoric distributes different elements on a terrain of a governing apparatus,” yielding “a form of cartography” (pp. 38, 39).

5. The relative linguistic/syntactic simplicity of lexical framing serves to more clearly illustrate the computational dimension of the test case, because such frames have limited semantic and little syntactic variation (see subsequent section for how such variation is methodologically captured). My hope is that subsequent research can build on it to encode more complex rhetorical structures for tracking/mapping, whether they be frames with more complex structure (like thematic frames, or Burkean acceptance/rejection frames), tropes, figures, narrative structures, genres, etc.

6. The term global warming is included to capture documents, particularly those in the early years of the dataset, that still used this term over climate change.

7. See also Kessler and Graham (2018) for why diachronic approaches to analyzing rhetorical practices are necessary.

8. Links to all code written for this study are listed in the appendix.

9. Rationales for sequential over bag-of-words-based approaches to inductively discover semantically interesting features in large textual corpora include Koteyko et al. (2013) : “the lexical environment of a search term [… helps provide] a ‘feel’ of the corpus and a ‘map’ of key junctions” (p. 77); Nerlich et al. (2012): “[…] the simple and obvious fact that human language is an inescapably sequential phenomenon” (p. 48). See also Grundmann and Krishnamurthy’s (2010) argument on “the phraseological tendencies of all languages, whereby the choice of one word tends to favor the co-selection of other words, its collocates, within close proximity in a text” (p. 117).

10. See also Figure 2 and its description for the difference between a fully passive frame and the passive-agentic frame.

11. In a writeup of House hearings, Peter Folger (2006), then-director of research at the American Geophysical Union, noted that although Republicans saw the hearings as being about education and deepening understanding, Democrats on the committee “expressed their desire to move beyond the science and take action to address climate change.”

12. See also Hart’s (2001) contention that he has found his approach to computerized content analysis – a different method to the one exemplified in this study, but related conceptually – to be “creative” rather than mechanical, and requiring “subtlety” (p. 43).

Additional information

Notes on contributors

Zoltan P. Majdik

Zoltan P. Majdik is an associate professor at North Dakota State University in Fargo, ND. He has published about issues like the ones addressed in this article in the Quarterly Journal of Speech, Rhetoric Society Quarterly, Rhetoric & Public Affairs, Social Epistemology, and Argumentation.