The Honesty Project is supported by Wake Forest University, with the help of a very generous grant from The John Templeton Foundation. It is associated with, but funded independently from, the ERC-funded PRODEMINFO grant.
Numerous indicators suggest that democracy is under threat worldwide, with misinformation cited as a major concern. Misinformation is troubling not only because it lingers in memory even if people know it has been corrected, but also because under certain circumstances people come to view and value inaccuracy as a signal of “authenticity.” Within a populist logic, blatant lies violate the “establishment” norm of accuracy, thereby signaling the “authenticity” of a champion of “the people”. A lying politician may be considered “honest” because they are authentically “speaking their mind”.
Cooper, Cohen, Huppert, Levine, & Fleeson, 2021
Such belief-speaking is one component of a tripartite model of honesty (see figure) developed by the Honesty Project funders (Cooper, Cohen, Huppert, Levine, & Fleeson, 2021) and involves only the speaker’s beliefs and feelings, irrespective of factual accuracy. For a democracy, “belief-speaking” is problematic because it allows leaders to “honestly” speak beliefs without seeking common ground based on the actual state of the world.
We therefore urgently need a better understanding of what is considered politically “honest”, and under what circumstances. Within the tripartite model, this quest requires exploration of the balance between the three components. We pursue this quest computationally, using large- scale data analysis and text modeling of corpora (e.g., Google ngram, New York Times, Congressional Record).
Stage 1: Historical text-modeling. We examine the semantic evolution of “honesty” and related concepts (e.g., “truth”; “fact”; “mislead”) over time (1970s onward). We expect the components of the tripartite model to emerge as topics. This will illuminate 50 years of Americans’ thoughts about honesty.
Stage 2: Presidential dishonesty. We will relate the linguistic markers of honesty (from Stage 1) to events that arguably involved dishonesty by the president (viz. Watergate, Iran-contra, Lewinsky affair, Weapons of Mass Destruction, and the recent rise of “belief speaking”). For each event, we focus on relevant text (e.g., discussing Watergate) to explore the tripartite model.
Stage 3: Language and attitudes. We will link text modeling to behavioral data from existing surveys, e.g., American Freshman Survey (AFS). The AFS has sampled freshmen since 1966, querying political views and social / political engagement. We will relate survey data to the historical events and to linguistic markers (from Stages 1-2).
Stage 4: Honesty and society. We will link text modeling to societal indicators. Guided by precedent, we expect economic inequality, via its established links to polarization, to create obstacles for finding shared facts. Other variables address social capital (e.g., trust, civic engagement), whose decline has been posited as catalyst for misinformation.
Concern with “fake news” has become a central theme around the world. We believe that honesty must matter in a democracy or it will not remain a democracy. Our project will identify factors that determine how people view honesty, and what we might do to make accuracy matter again.