Automatically created publications

Automatically created documents play an important role in digitalized compliance processes. Publications are also created algorithmically outside of compliance processes. A well-known example is MIT’s fake papers, which were submitted to academic journals and, surprisingly, accepted [1]. On October 8, 2019, an employee of the Springer publishing house presented the first chemistry book written by a computer at a conference in Osnabrück. The book is not fake – it is intended as a serious publication. It cost EUR 100.99 on Amazon on the day of the conference, but can be downloaded free of charge as a PDF from the Springer website. One of the attendees did exactly that and checked whether the newly appointed Nobel Prize winners in chemistry appear in the Springer book.

You can guess the result: only two of the three laureates, John Goodenough, Stanley Whittingham and Akira Yoshino, appear in the book. Akira Yoshino is missing; it’s not clear why. What can we learn from this? Automatically generated publications and reports are nothing more than nicely presented search results. Since search results are never perfect, neither are auto-generated publications. Nor can you expect anything new or creative from such publications. So are they useless? They can be helpful for sifting large text collections. Problems arise when, due to a lack of media competence, the results are assigned disproportionate significance. However, a creative person may well discover something new in the automatically compiled data.