Page 177 - Demo

Supplemental Material175Sagreement was reached. Tweets in which no storyline was revealed were labelled as not applicable (NA).C-b. Steps for answering RSQ2 (which of the visualisations revealed through RSQ1 circulate frequently on the open Web, in what topical context, with what textual tone, and as part of what nanotechnology-in-food storyline (if at all)?)1. Narrowing the dataset: cross-platform circulatingUsing a Google reverse image search (Rogers, 2019, p. 19, 2021, p. 13) and the 104 visualisations extracted from the tweet list, we revealed the frequency of the visualisations’ circulation on the open Web. Google reverse image search is a black-boxed algorithm (Rogers, 2019, p. 19) that returns results of a source visualisation as well as modified versions of it (e.g., a version with different colours, a version that is a fragment of the source visualisation, Nieuwenhuysen, 2019). To include only those visualisations that circulated frequently on the open Web, we selected the visualisations whose Google reverse image search returned at least 200 results.52 The narrowed dataset consisted of 14 visualisations. 2. Analysing URLs In all the English URLs on which the 14 visualisations appeared,53 the visualisation was compared with the source visualisation. When there was a visual match (n=667 URLs), the text was coded for its main topical context. For this analysis, the complete text of the URL was analysed, unless the URL contained several distinct sections (e.g., several blog posts). In that case, the section containing the visualisation was analysed. Applying the same strategy as applied to clean the tweet dataset (Step 1 above), we indicated when an URL text discusses nanotechnology in food. 3. Narrowing the visualisations dataset; Adding URLs52 This number includes URLs that were repeated multiple times, URLs that were unavailable (‘broken links’), URLs that although appearing in the result list did not include the visualization, and URLs that included a visualization that was a modified version of the source visualization. These ‘false results’ were filtered in the next step.53 We analysed only the URL results in English on each of the first 20 pages. Results in English were generally followed by results in another language. In 13 of the 14 visualizations, all results in English seemed to appear on pages 1 to 20.Efrat.indd 175 19-09-2023 09:47