Donate to the Igs

Inauthentic Paper Detector?

InformaticsLogo.gifHow authentic is the Indiana University School of Informatics’ Inauthentic Paper Detector? Is it (as scoffers scoff) an inauthentic detector of paper or (as some detectives detect) an authentic detector of inauthentic paper or papers?

The inventors make a claim that some may find cryptic:

We are trying to detect new, machine written texts that are simply generated not to have any meaning, yet appear to have meaning on the surface.

(Thanks to Investigator Kristine Danoski for bringing this to our attention.)


UPDATE later in the day:

Investigator Tuomas R?s?nen writes:

I tried the inauthenthic paper detector by applying it to some example texts found from Project Gutenberg.

The results were staggering: It turns out that Hamilton Wright Mabie, and Thomas Babington Macauley were not human!

The beginning of the first chapter of Mabie’s Books and Culture, as well as the beginning of chapter Hallam’s History from the Critical & Historical Essays were both found fakes.

I wonder if anyone has even suspected before.


FURTHER UPDATE, still later in the day:

ArunGiridhar.jpgInvestigator Arun Giridhar writes:

As a follow-up to your blog post on the Inauthentic Paper Detector, I tried a few further tests on different machine-generated texts:

1. The SciGen-generated paper on Rooter was rated Inauthentic with a 21% chance of being authentic.

2. Alan Sokal’s paper on quantum gravity, itself the reason for a previous Ig Nobel Prize, scored Inauthentic, with a 21% chance on being authentic.

3. Several randomly generated essays from the Postmodernism Essay Generator scored Authentic, with the ratings being 94.7%, 84.8%, 86.6%, 95.4% and 95.8%.

I leave you to draw your own conclusions.

Improbable Research