Page 176 - Demo

Supplemental Material174Retweet (RT) is the most relevant engagement metric for studying circulating visualisations on Twitter, considering Twitter’s affordance (Amit-Danhi & Shifman, 2018): they are not a mere reflection of exposure or attention, but rather an indication of an active engagement that influences a visualisation’s online presence (Boscarino, 2022; Geboers & Van De Wiele, 2020). The selection of tweets with high RT metrics left us with a list of 214 tweets. We manually cleaned this list of irrelevant tweets (tweets that are not about nanotechnology in food, e.g., tweets about food and Nano cryptocurrency; tweets about food and things that have nano in their name; tweets that are about nanotechnology AND food but not IN food). We also manually removed one tweet from a suspended account, and we merged tweets that were repeated with bot-like behaviour (multiple times, usually with small intervals, by a single user, and with the same text and visuals).Our clean dataset consisted of 90 tweets that contained 104 unique visualisations. Some tweets included more than one visualisation; some of the visualisations were repeated in multiple tweets. The visualisations were downloaded; the tweets and their visualisations were put in an Excel file for further analysis of the tone (of the text) towards nanotechnology in food and the storylines (of the visualisations and their accompanying text) narrated.The tone was coded as typically done in the analysis of media coverage (e.g., Baumgartner et al., 2008; Kuttschreuter et al., 2011). A tweet with positive wording regarding nanotechnology in food, such as ‘prevent food spoilage’, was coded as positive; a tweet with negative wording, such as ‘infecting the public’, was coded as negative; a tweet with wording that does not express any valence, such as ‘discussing the impact nanotechnology will have on the food chain’, was coded as neutral. The results of the tone coding were compared with the results of automated sentiment analysis. When there was a difference between the manual and the automated results, the manual analysis was discussed by the authors until agreement was reached.The storylines were coded inductively, drawing on Hajer’s (1995, p. 56, 2006, p. 71) definition of storyline. This notion of storyline is shown to be effective in analysing tweets with opinions about a policy issue (Jeffares, 2014, p. 144). Following Kress’s (2001, 2010) approach, each visualisation and the text accompanying it (its tweet’s text) were analysed together to reveal repeated narratives that give meaning to the topic of nanotechnology in food (Table C1). Coding disagreements were discussed by the authors, and storylines were merged and divided, until Efrat.indd 174 19-09-2023 09:47