Background/Question/Methods
How scientists write about genetically modified (GM) crops in peer-reviewed journals could influence science journalists, and thereby public opinion. Public opinion polls in China, USA, and the EU indicate low support for GM crops; between 12-38% have a positive stance. The objective of this study was to quantify and analyze trends in the stance of scientists on the topic of GM crops. We downloaded over 1700 peer-reviewed journal articles discussing GM crops and the environment. To analyze stance, 500 of these were classified into positive, neutral, and negative positions on GM crops. To address our own bias, at least four individuals classified articles, and only those with a three-quarters majority were included in our analyses. Linguistic features of these articles were used to train a random forest classifier to predict the stance of unread articles. We also tested the hypothesis that articles with similar stance cited similar sources. To our knowledge, this study represents the first systematic quantification of stance toward GM crops in peer-reviewed literature.
Results/Conclusions
Our results indicate that over half of the articles conveyed a neutral stance towards GM crops, yet 40% conveyed a non-neutral stance. Of the 230 articles with an agreed upon stance that discussed GM crops, 139 (60.4%) were neutral, 60 (26.1%) were positive, and 31 (13.5%) were negative. China had the highest ratio of positive to negative articles (8:1), followed by the USA (12:5). The EU had a more negative position on GM crops (5:7), possibly a reflection of the political stance toward GM crops. Articles with a positive stance toward GM crops mentioned Bt 10 times more than glyphosate, as opposed to a ratio of 2:1 for negative and neutral articles. Negative stance articles had a higher rate of publication in the early 2000’s and primarily focused on the environmental impacts of gene flow. Articles with positive or negative stance have distinct linguistic characteristics. Using our random forest model, we were able to accurately predict whether articles were neutral, positive, or negative more than 64% of the time, as opposed to random chance which would be 44% accurate. The opinions formed in scientific journal articles can influence media outlets and be amplified into public opinion. How we, as scientists, communicate in our publications may have unintended consequences in the debate over the health and environmental safety of GM crops.