The PROMPT codebook: the contribution of narratives, rhetorics and information manipulation techniques to the fight against disinformation

By Anda Rozukalne, Vineta Kleinberga, and Sara Mercereau

Media professionals consider the fight against disinformation as one of the most important challenges to their work. Citizens expect media professionals not only to be a source of trusted and verified information, but also to protect the public from the influence of disinformation. Even more, transformation in disinformation strategies and techniques have changed the way journalists process information, pushing them to adapt and adopt new working methods.

What can a new project add to the field when there are already many studies, fact-checking initiatives and organisations (though few disinformation detection apps)?

PROMPT partners recognise that disinformation is an ever-changing phenomenon, with new topics, players and techniques/methods being increasingly instrumentalised. Studying AI and disinformation means dealing with a moving target that keeps picking up pace with technological innovation.

The most important yet difficult task in disinformation studies is to decipher the impact of disinformation. Central to PROMPT is the hypothesis that propagators of disinformation increasingly leverage culture, community, and language specific patterns to achieve greater impact. We need to understand how narratives are formed and reshaped to resonate in specific communities. PROMPT employs AI-driven methods and Large Language Models (LLM) to help monitor disinformation narratives, how they propagate and transform in the process of its movement and transformation across social platforms and local contexts.

The process is not without challenges. The main task of a group of researchers of the Faculty of Social Sciences at Rīga Stradiņš University (Latvia), in cooperation with opsci.ai, the coordinator of the PROMPT project, was to create a codebook – to detect and deconstruct the main disinformation narratives and their associated arguments.

What is new in the PROMPT codebook?

Using previous disinformation research, best practices from practitioners, and analysis of disinformation “samples”, we created a codebook of 30 code groups and more than 240 codes. Building on an interdisciplinary approach (media, communication, semantic studies), we developed a disinformation coding system that combines various linguistic elements, rhetorical devices, information manipulation techniques, as well as codes that help to recognise the purpose of disinformation, the intentions of its propagators and its possible effects.

In the codebook, the form of disinformation (how it’s presented) is as important as its substance (what it talks about).

The main part of the codebook explores various linguistic and rhetorical categories. Steered by the intuition that distributors of disinformation increasingly use culture, community and language specific patterns to achieve greater effect, we focused on the rhetoric of disinformation using a multimodal (text, photos, videos) and multilingual (8 languages) filter. We deciphered these practices by classifying persuasion mechanisms and information manipulation techniques, lexical devices and a diverse list of rhetorical figures. To analyse deeper layers of disinformation, we used axiologically-semantic categories to determine and define what values these social media posts promote.

The codebook also helps identify whether the information aims to sow discord, spread fear and influence public opinion by using belligerent words that convey aggression or hostility (e.g., “enemy”, “traitor’, “invader”, “conflict” or “threat”); or if it promotes discrimination against certain groups by employing stereotypical words that reinforce clichés – e.g “catho-facho”, “welfare queen” or “lazy immigrant”. A sense of urgency, fear or panic can be facilitated by alarmist words, often used to exaggerate a situation or provoke a strong emotional response, e.g., “catastrophic”, “urgent”, “epidemic”, “tragic” or “crisis”. However, the identification of specific lexicon does not necessarily confirm the presence of disinformation; rather, it helps see if these are highly correlated with disinformation content.

After the first phase of testing, realising the capabilities of the in-depth analysis tool, we also supplemented the codebook with the analysis of how the language in social media posts connects with different political ideologies. This approach significantly enhances the tool’s ability to integrate a broader context of political culture in the understanding of disinformation.

Powered by AI (LLMs), all categories will support the PROMPT disinformation analysis arsenal. It will be capable of detecting subtle nuances in the language of public information – nuances that might go unnoticed to the naked eye, but which may fundamentally alter the message, shifting its meaning and playing a crucial role in shaping perception and reception of disinformation across the digital space.

How did we develop the codebook and what did we learn in the process?

All PROMPT partners participated in the creation and validation of the matrix. In 3 technical workshops, project partners proposed and analysed examples of disinformation and helped revise the matrix.

One key lesson: the codebook should reflect both the standards upheld by journalists (categories on facticity, disinformation typology, verifiability), as well as categories less explored in this field (axiological attributes and rhetorical devices). This pluralist approach is central to PROMPT’s contribution to the research field. We tested the matrix across a sample of country-specific disinformation cases, including debunked claims extracted from fact-checkers’ work.

The matrix embodies trade-offs between theory and practice, combines a literature-based approach and an empirical investigation of disinformation samples across six countries (Latvia, Lithuania, Estonia, Romania, Italy and France). It will account for the peculiarities of disinformation dissemination across these countries and how this may be impacted by unique linguistic traits, the structure of the communication environment, as well as locally specific socio-political and cultural discourses.

As great a challenge as the development of PROMPT’s matrix may be, it should help disclose and enhance our understanding of the – rhetorical but not only – mechanisms of disinformation. It should bring additional tools to uncover those seeking to manipulate public opinion, polarise the digital space, and ultimately, undermine democratic institutions.

This article was originally published on de.ejo-online.eu Read the original article here: https://de.ejo-online.eu/digitales/das-prompt-codebuch-narrative-rhetorik-und-techniken-der-informationsmanipulation-zum-kampf-gegen-desinformation