Article published by lmu.de
Readers find news articles automatically generated by AI more difficult to understand compared to texts written by journalists, mainly due to word choice and handling of numerical data.
Traditional articles written by journalists are more understandable than those produced automatically, according to a study. LMU University study, published in the journal Journalism: Theory, Practice, and CriticismThe research team from Department of Media and Communication (IfKW) conducted a survey of over 3,000 online news readers in the United Kingdom. Each participant rated one of 24 survey texts, half of which were produced by automation and the other half were written by journalists. “Overall, readers found the 12 automated articles significantly less understandable,” concluded the lead author, Sina Thäsler-KordonouriThis happened despite the fact that the articles created by artificial intelligence were edited by journalists before publication.
Problems with numbers and word choice
According to the survey, one of the reasons for readers’ dissatisfaction was the choice of words used in AI-generated texts. Readers noted that automated articles contained many inappropriate, difficult, or unusual words or phrases. Furthermore, they were significantly less satisfied with the way automated articles handled numbers and data.
Readers' perceived shortcomings in the handling of numbers and word choice in automated articles are among the reasons why these articles were more difficult to understand, according to the researchers. However, readers were equally satisfied with the character of the writing and the fluidity of the narrative structure in automated articles and those written by journalists.
Need more human editing
Professors Neil Thurman, who led the project, suggests that “when creating and/or editing automated articles, journalists and technologists should aim to reduce the amount of numbers, better explain words that readers may not understand, and add language that helps readers visualize the content of the story.”
This study is the first to investigate the relative clarity between manual and automated articles and explain the reasons for the differences. “Our results show the importance not only of human involvement in the production of data-driven news, but also of improving this process,” emphasizes Sina Thäsler-Kordonouri.