Ten activities were scheduled for the first half of the project. Here we highlight the most promising outcomes.
Analysis 1.1. A computational history of "truth"
This analysis is nearly complete, with additional explorations ongoing as this has turned out to be even more interesting and valuable than anticipated. To date, this analysis has yielded one publication in PNAS Nexus (Lasser et al., 2022) and one in Nature Human Behaviour (Lasser et al., 2023).
Analysis 1.2. Truth discourse and 20th century fascism
A tentative preliminary analysis has shown that the lead-up to fascism in 1930s Germany was associated with a dramatic shift towards intuition-based language, exactly as expected.
Analysis 1.3. Truth, discourse, & democracy: towards an early-warning system
A preliminary analysis showed a mild positive correlation (r = 0.4) between some of the dimensions extracted from large-scale existing survey data and the average level of fact-speaking (as opposed to belief-speaking) in the congressional records.
Study 1.4. Public discourse vs. personal attitudes
We successfully executed data collection in our five target countries (Germany, Italy, Spain, UK, and Hungary). We have analysed the data sets and extracted the best set of items to compare people’s subjective ontology of truth, captured by what we call the Evidence-Intuition Scale (EIS), across countries.
Study 1.5. Public discourse vs. personal tweets
This study is an extension of Study 1.4 where participants’ historical Twitter discourse is linked to their survey responses to examine whether individual ontologies of honesty and/or truth are detectable in spontaneous social media activity. Thus far, we have only found a very modest association between people’s speech on Twitter and their responses to the EIS.
Study 1.6 Truth in people's own words
For this analysis, we have leveraged collaborations with the Horizon 2020-funded JITSUVAX project, which granted us access to text contributions from participants regarding their opinions on vaccination. We applied our dictionaries to the text and administered the EIS to participants, finding a small-to-modest correlation between their speech in response to a prompt and their epistemic orientation.
Analysis 2.1. A European misinformation corpus
This analysis is dedicated to curating a European misinformation corpus through a search of fact-checking archives in the five target countries to select items – from both trustworthy and untrustworthy sources – for potential use as experimental stimuli. The corpus has been presented at the European Chapter of the Association for Computational Linguistics (EACL) in 2023, and it is a valuable resource for both quantitative and qualitative misinformation studies. In parallel, we have commenced a collaboration with the German-Austrian Digital Media Observatory (GADMO), which is working on a similar project in German-speaking countries.
Analysis 2.2 A synthetic misinformation corpus
This analysis endeavours to create a “synthetic corpus of misinformation” for use in research in different languages. The material for this corpus will consist of synthetic stimuli (i.e. stimuli created from scratch according to various criteria but without a knowable truth value) from a large variety of studies conducted by the PRODEMINFO team and all collaborators. Items will be labeled and classified according to the type of misleading rhetoric they contain. This work is almost complete and has benefited from collaboration with Mubashir Sultan at the Max-Planck Institute for Human Development in Berlin.
