How the public reacts to CRISPR? Complete timeline since 2015
The study focused on the mean sentiment of tweets. A sentiment approximates emotional weight of text – typically from -1 to 1, corresponding to a polarity from negative to positive opinions. Sentiment can be quite accurately assessed by modern machine learning tools. BERT, a model used in the study, correctly distinguishes negative and positive movie reviews in 95% of cases.
In a period from 2013 to 2019, the first two years brought too few tweets to provide reliable mean sentiment. Early quotes about CRISPR were limited to scientific circles, often associated with conferences.
From June 2015 to May 2019, the mean sentiment data were summarized to a graph:
The public remains mostly positive about CRISPR – but there are isolated peaks of negative reactions to certain events. The events which negatively stimulated the public include Sanger Institute’s study about off-target activity of CRISPR (July 2018), He Jiankui’s experiment (November 2018), and Wired’s story about biohacker “malware” (February 2019).
Precise classification of tweets showed that negative reactions were almost exclusively limited to embryos and humans – since 2015. Modifications of bacteria, animals, or plants were seen as mostly positive.
#GMO and #CRISPRbabies were the only popular tags associated with negative mean sentiment. In contrast, tags such as #ResearchHighlight, #Cancer, or #GeneTherapy consisted of mainly positive tweets. In addition to hashtags, a keyword “mutation” was negatively correlated, whereas “disease” and “treatment” were among the main positive topics.
Our results suggest that, overall, the CRISPR technology was discussed in a positive light, which aligns well with a previous study which considered the coverage of CRISPR in the press. However, more recently the sentiment reveals a series of strong negative dips, pointing to a more critical view.
As the authors remark, it’s important to remember about basic limitations of the study – Twitter users are not a representative group of the whole population and the platform is prone to automatic, bot-driven activities. At the same time, this is certainly insightful approximation of the public’s view about CRISPR. The preprint suggests, that the data might be used in the future to provide additional voice in discussions around policy:
Specific topics, such as the discussion surrounding a potential moratorium of CRISPR, may be analysed in more detail and provide actionable outcomes. Since the presented analysis can automatically process a large amount of data in almost real-time, it extends the traditional toolset of empirical methods for discourse analysis. It may therefore help analysing public opinion and support policy and decision making.
Preprint: Martin Müller, Manuel Schneider, Marcel Salathé, Effy Vayena (2019). Combining Crowdsourcing and Deep Learning to Assess Public Opinion on CRISPR-Cas9. Biorxiv. Doi:10.1101/802454.