What Comes After the Algorithm? An Investigation of Journalists’ Post-editing of Automated News Text

Abstract

ABSTRACT Journalists often perceive that automated journalism produces news texts that lack narrative and editorial quality. Therefore, they sometimes manually edit automated output before publication, creating so-called “post-edited” variants. Using an exploratory sequential mixed-methods design, this study aims to advance research on automated journalism by investigating the steps journalists say they take when post-editing and testing whether—and if so, how—these steps are reflected systematically on a larger scale in post-edited news texts. First, journalists’ statements about the post-editing process were gathered in semi-structured interviews. Then, qualitative content analysis of automated news stories and their post-edited offspring examined whether post-edited stories contain evidence of forms of editing not mentioned by the journalists. Second, the qualitative findings were investigated quantitatively using comparative content analysis (N = 282). The qualitative findings suggest that a range of editorial steps may be taken during post-editing. The quantitative analysis shows that some of these steps are reflected systematically in post-editing, resulting in significant differences between automated news stories and their post-edited offspring. However, findings also show differences between journalists’ reports and post-editing practice on a larger scale, including evidence that forms of editing were taking place that were not mentioned in the interviews.