COVID-19 vaccines have been largely debated by the press. To understand how mainstream and alternative media debated vaccines, we introduce a paradigm reconstructing time-evolving narrative frames via cognitive networks and natural language processing. We study Italian news articles massively re-shared on Facebook/Twitter (up to 5 million times), covering 5745 vaccine-related news from 17 news outlets over 8 months. We find consistently high trust/anticipation and low disgust in the way mainstream sources framed “vaccine/vaccino”. These emotions were crucially missing in alternative outlets. News titles from alternative sources framed “AstraZeneca” with sadness, absent in mainstream titles. Initially, mainstream news linked mostly “Pfizer” with side effects (e.g. “allergy”, “reaction”, “fever”). With the temporary suspension of “AstraZeneca”, negative associations shifted: Mainstream titles prominently linked “AstraZeneca” with side effects, while “Pfizer” underwent a positive valence shift, linked to its higher efficacy. Simultaneously, thrombosis and fearful conceptual associations entered the frame of vaccines, while death changed context, i.e. rather than hopefully preventing deaths, vaccines could be reported as potential causes of death, increasing fear. Our findings expose crucial aspects of the emotional narratives around COVID-19 vaccines adopted by the press, highlighting the need to understand how alternative and mainstream media report vaccination news.

The increasing availability of textual corpora and data fetched from social networks is fuelling a huge production of works based on the model proposed by psychologist Robert Plutchik, often referred simply as the “Plutchik Wheel”. Related researches range from annotation tasks description to emotions detection tools. Visualisation of such emotions is traditionally carried out using the most popular layouts, as bar plots or tables, which are however sub-optimal. The classic representation of the Plutchik’s wheel follows the principles of proximity and opposition between pairs of emotions: spatial proximity in this model is also a semantic proximity, as adjacent emotions elicit a complex emotion (a primary dyad) when triggered together; spatial opposition is a semantic opposition as well, as positive emotions are opposite to negative emotions. The most common layouts fail to preserve both features, not to mention the need of visually allowing comparisons between different corpora in a blink of an eye, that is hard with basic design solutions. We introduce PyPlutchik the Pyplutchik package is available as a Github repository (http://github.com/alfonsosemeraro/pyplutchik) or through the installation commands pip or conda. For any enquiry about usage or installation feel free to contact the corresponding author, a Python module specifically designed for the visualisation of Plutchik’s emotions in texts or in corpora. PyPlutchik draws the Plutchik’s flower with each emotion petal sized after how much that emotion is detected or annotated in the corpus, also representing three degrees of intensity for each of them. Notably, PyPlutchik allows users to display also primary, secondary, tertiary and opposite dyads in a compact, intuitive way. We substantiate our claim that PyPlutchik outperforms other classic visualisations when displaying Plutchik emotions and we showcase a few examples that display our module’s most compelling features.

This work introduces EmoAtlas as a computational framework extracting emotions and syntactic/semantic word associations automatically from texts. EmoAtlas implements a cutting-edge synergy of interpretable artificial intelligence (AI) and psychologically-validated lexicons. Our framework supports out-of-the-box emotion detection of 8 categorical emotions and syntactic network building in 18 languages. In the task of detecting emotions in human-labelled social media posts and news media articles, either in Italian or in English, EmoAtlas achieves performances analogous, or even superior (e.g., 70.2% precision for detecting joy in tweets), to state-of-art natural language processing techniques like BERT, RoBERTa, distillBERT and ALBERT (e.g., 67.9% precision for detecting joy in tweets). This performance comes with EmoAtlas being 12 times faster than BERT. In a psychometric task like reproducing human creativity ratings for 1,071 short stories, EmoAtlas and BERT obtain equivalent predictive power (ρ ≈0.495, p<10-4). Combining BERT's semantic features with EmoAtlas' emotional/syntactic networks of words, the cross-validated resulting AIs get substantially better at estimating creativity rates of stories (ρ ≈0.628, p<10-4). This indicates an interplay between the creativity of narratives and the semantic, emotional, and syntactic divergence of their words, pointing out how EmoAtlas and BERT-like models could be used in synergy in psychometrics. By outputting interpretable emotional profiles and syntactic networks, EmoAtlas can quantify how emotions are channelled through specific associations in texts, e.g., how did customers frame their ideas and emotions towards "beds" in hotel reviews? We release EmoAtlas as a standalone computational tool and discuss its impact in modelling texts as data, promisingly filling current research gaps in AI and cognitive science.

No abstract

We quantify social media user engagement with low-credibility online news media sources using a simple and intuitive methodology, that we showcase with an empirical case study of the Twitter debate on immigration in Italy. By assigning the Twitter users an Untrustworthiness (U) score based on how frequently they engage with unreliable media outlets and cross-checking it with a qualitative political annotation of the communities, we show that such information consumption is not equally distributed across the Twitter users. Indeed, we identify clusters characterised by a very high presence of accounts that frequently share content from less reliable news sources. The users with high U are more keen to interact with bot-like accounts that tend to inject more unreliable content into the network and to retweet that content. Thus, our methodology applied to this real-world network provides evidence, in an easy and straightforward way, that there is strong interplay between accounts that display higher bot-like activity and users more focused on news from unreliable sources and that this influences the diffusion of this information across the network.