🍻New paper published on @BDCC_MDPI, collab. w. @mvitevit and @Fede_Botta:
👇
Cognitive Networks Extract Insights on COVID-19 Vaccines from English and Italian Popular Tweets: Anticipation, Logistics, Conspiracy and Loss of Trust
👇
mdpi.com/2504-2289/6/2/…
👇
Cognitive Networks Extract Insights on COVID-19 Vaccines from English and Italian Popular Tweets: Anticipation, Logistics, Conspiracy and Loss of Trust
👇
mdpi.com/2504-2289/6/2/…
@BDCC_MDPI @mvitevit @Fede_Botta We investigate almost 5k Popular tweets about #COVID19 and #vaccines from early 2021.
Remember when vaccines were announced?
How did massively read content frame vaccines? With which stances and emotions?
We use #cognitive networks and #machinevision to address these points!
Remember when vaccines were announced?
How did massively read content frame vaccines? With which stances and emotions?
We use #cognitive networks and #machinevision to address these points!
We found #emotional #polarisation in the framing of "vaccines" in English tweets.
Emotions like trust and anticipation co-existed with negative emotions of sadness and anger.
A circumplex model corroborated this dichotomy. These emotions rose from discussions about cases/deaths:
Emotions like trust and anticipation co-existed with negative emotions of sadness and anger.
A circumplex model corroborated this dichotomy. These emotions rose from discussions about cases/deaths:
We also identified #keywords in social discourse through closeness centrality, which in cognitive networks capture semantic relatedness and prominence.
Social discourse focused on #vaccines as linked to the pandemic, workers and #logistic administration.
Social discourse focused on #vaccines as linked to the pandemic, workers and #logistic administration.
We then implemented a #multimodal pipeline combining cognitive #textual forma mentis networks (peerj.com/articles/cs-29…) with #OCR, #face and #entity recognition.
In this way, we were able to extract more #text from tweets and assess the #pictures portrayed in them.
In this way, we were able to extract more #text from tweets and assess the #pictures portrayed in them.
#Semantic frames for #vaccine were populated by different emotions when grouping tweets by their #pictures.
Tweets with people showing faces with no facemasks elicited more #trust than those with people wearing masks.
Disgust was strong only in tweets with pics showing no person.
Tweets with people showing faces with no facemasks elicited more #trust than those with people wearing masks.
Disgust was strong only in tweets with pics showing no person.
👉What happened with #AstraZeneca and social discourse in March 2021?
In #English popular tweets, read million times, #vaccine retained trust but #AstraZeneca lost it. 🚨
This loss of trust did not happen in the #Italian twittersphere in that same period and for pop tweets.
In #English popular tweets, read million times, #vaccine retained trust but #AstraZeneca lost it. 🚨
This loss of trust did not happen in the #Italian twittersphere in that same period and for pop tweets.
Our approach combines #text analysis and #cognitive networks with #machine vision and #AI.
We apply it for to reconstruct how #vaccines were framed in early 2021.
We found patterns of emotional polarisation, topics of #logistics, #conspiratorial associations and losses of trust.
We apply it for to reconstruct how #vaccines were framed in early 2021.
We found patterns of emotional polarisation, topics of #logistics, #conspiratorial associations and losses of trust.
By understanding how widely popular content, flowing online, frames specific entities, like vaccines, we can better identify how massive audiences are exposed to information and to #emotional contagion.
This is a key step also for fighting misinformation, cf. @IRIS_Academic .
This is a key step also for fighting misinformation, cf. @IRIS_Academic .
Our approach also highlights #country specific patterns, much like recent works by @gaveltri and colleagues about vaccine hesitancy in different EU countries.
There is no one-size-fits-all and this underlines the need to develop robust #AI analysing multiple social environments.
There is no one-size-fits-all and this underlines the need to develop robust #AI analysing multiple social environments.
I was honoured to work with such an amazing multidisciplinary psych/datasci team!
Thanks to the @UniofExeter for supporting us in publishing this work. Tagging also @UniExeterIDSAI and @exetercompsci. And thanks to @BDCC_MDPI for a terrific editorial management!
Thanks to the @UniofExeter for supporting us in publishing this work. Tagging also @UniExeterIDSAI and @exetercompsci. And thanks to @BDCC_MDPI for a terrific editorial management!
@threadreaderapp unroll
• • •
Missing some Tweet in this thread? You can try to
force a refresh
