Study finds celebrity tweets likely shaped US negative public opinion of COVID-19 pandemic

In a recent study published in BMJ Health and Care Informatics, researchers evaluated the influence of celebrities on public attitudes toward the coronavirus disease 2019 (COVID-19) pandemic.

Background

The COVID-19 pandemic has significantly changed human lives globally. Millions of individuals have been forced out of public places, with pandemic-related conversations occurring online on different social network platforms. People in the public eye (PIPE), such as politicians, news anchors, entertainers, and athletes, have leveraged such platforms to discuss diverse topics and share their experiences and opinions on COVID-19, vaccines, and public health measures with their followers.

About the study

In the present study, researchers compared the COVID-19 vaccine-related tweets cooccurring with mentions of PIPE representatives between January 1, 2020, and March 1, 2022. PIPE representatives included Novak Djokovic (ND), Nicki Minaj (NM), Tucker Carlson (TC), Joe Rogan (JR), Donald Trump (DT), Aaron Rodgers (AR), Rand Paul (RP), Eric Clapton (EC), Ron DeSantis (RD), Candace Owens (CO), Phil Valentine (PV), and Ted Cruz (TeC).

They were selected due to statements made in the public that were against vaccination, misinformation, or propagated misinformation. PIPE representatives were grouped as news personnel, politicians, or athletes/entertainers. The team harvested about 13 million tweets; tweets from bots, repetitive news media/users, and duplicates were removed. The tweets were queried for COVID-19-19 vaccine(s) and PIPE terms. The final dataset comprised 45,255 tweets from 34,407 unique authors.

The sentiment for each tweet was calculated using the DistilRoBERTa model, trained by labeling 4,000 tweets as negative or positive, which were used with the back translation method to generate an augmented dataset of 50,000 tweets. After fine-tuning, the Hugging face pipeline was used to process data for sentiment analysis. The probabilistic confidence score (0 – 1) and polarity (positive/negative sentiment) were reported.

Findings

Among athletes/entertainers, there were 29,210 likes on tweets mentioning ND, with an average of 46.63 and the maximum and minimum positive sentiment of 0.44 in February 2021 and 0.1 in July 2021, respectively. The mean confidence score for negative and positive tweets mentioning ND was 0.96 and 0.88, respectively. For AR, the maximum positive sentiment was 0.3 in December 2021, while the minimum was 0.17 in February 2022.

The mean confidence scores were 0.95 for negative and 0.87 for positive tweets mentioning AR. The maximum positive sentiment for EC-mentioned tweets reached 0.19, with a minimum of 0.07 in June 2021. The mean confidence score for positive tweets mentioning EC was 0.84, and 0.97 for negative tweets. The maximum positive sentiment for tweets mentioning NM was 0.66 in May 2021, and the minimum was 0.1 in February 2022.  

For NM, the mean confidence score was 0.93 for negative tweets and 0.87 for positive tweets. Among politicians, 11,195 tweets in the dataset mentioned DT, with a maximum positive sentiment of 0.83 in February 2020, which declined sharply in the next month to 0.23 and remained consistently low ever since. The dataset comprised 1,520 tweets mentioning TeC, which reached a maximum positive sentiment of 0.22 in May 2020 and a minimum of 0.12 in January 2021.

RD mentions occurred in 3,885 tweets, reaching a maximum positive sentiment of 0.966 in June 2021 and a minimum of 0.25 in June 2020. There were 2892 RP-mentioned tweets, with a maximum positive sentiment of 0.6 in March 2020 and a minimum of 0.15 in December 2021. Among news personnel, 4,843 tweets mentioned TC, with a maximum positive sentiment of 0.355 in April 2021 and a minimum of 0.07 in August 2020.

There were 1,264 PV-mentioned tweets; the maximum positive sentiment was 0.4, and the minimum was 0.1. CO reached a maximum positive sentiment of 0.43 in January 2022 and a minimum of 0.03 in October 2021. A majority of the maximum/minimum sentiment occurred in different months. However, three months contained two occurrences of maximum/minimum sentiment.

Two entertainers (EC and NM) had their maximum positive sentiment in June 2021, while two politicians expressed their most-negative sentiment during December 2021. Several major events occurred in December 2021, including Omicron detection in the United States (US). NM and AR, two entertainers, had their lowest sentiment during February 2022.

Conclusions

In summary, the findings revealed consistent patterns of emotional content cooccurring with messages/tweets shared by PIPE representatives influenced public opinion. Overall, a broadly polarized negative tone was observed despite the faint differences in sentiment across PIPE subgroups. The public view was shaped by political ideologies, risk perceptions, and behaviors shared by PIPE representatives.

The authors believed that messages/tweets from news personnel and politicians strongly correlate with public health events. Further analysis of public response to the emotions shared by PIPE representatives could provide valuable insights into the role of social network-shared sentiment in disease prevention, COVID-19 control/containment, and response to future disease outbreaks.

Written by

Tarun Sai Lomte

Tarun is a writer based in Hyderabad, India. He has a Master’s degree in Biotechnology from the University of Hyderabad and is enthusiastic about scientific research. He enjoys reading research papers and literature reviews and is passionate about writing.