Filling a vacancy takes a lot of (costly) time. Automated pre-processing of applications using artificial intelligence technology can help to save time, e.g., by analyzing applications using machine learning algorithms. We investigated whether such systems are potentially biased in terms of gender, origin and nobility. For this purpose we created a corpus of common German reference letter sentences on which we performed sentiment analysis using the cloud services by Amazon, Google, IBM and Microsoft. We established that all tested services rate the sentiment of the same template sentences very inconsistently and biased at least with regard to gender.
%0 Conference Paper
%1 8940848
%A Folkerts, Finn
%A Schreck, Vanessa
%A Riazy, Shirin
%A Simbeck, Katharina
%B 2019 IEEE International Conference on Humanized Computing and Communication (HCC)
%C New York
%D 2019
%I IEEE
%K Arbeitszeugnis Digitalisierung Diskriminierung Künstliche_Intelligenz hbs-2017-369-2
%P 1-6
%R 10.1109/HCC46620.2019.00009
%T Analyzing Sentiments of German Job References
%U https://ieeexplore.ieee.org/document/8940848/
%X Filling a vacancy takes a lot of (costly) time. Automated pre-processing of applications using artificial intelligence technology can help to save time, e.g., by analyzing applications using machine learning algorithms. We investigated whether such systems are potentially biased in terms of gender, origin and nobility. For this purpose we created a corpus of common German reference letter sentences on which we performed sentiment analysis using the cloud services by Amazon, Google, IBM and Microsoft. We established that all tested services rate the sentiment of the same template sentences very inconsistently and biased at least with regard to gender.
@inproceedings{8940848,
abstract = {Filling a vacancy takes a lot of (costly) time. Automated pre-processing of applications using artificial intelligence technology can help to save time, e.g., by analyzing applications using machine learning algorithms. We investigated whether such systems are potentially biased in terms of gender, origin and nobility. For this purpose we created a corpus of common German reference letter sentences on which we performed sentiment analysis using the cloud services by Amazon, Google, IBM and Microsoft. We established that all tested services rate the sentiment of the same template sentences very inconsistently and biased at least with regard to gender.},
added-at = {2020-03-10T10:02:30.000+0100},
address = {New York},
author = {{Folkerts}, Finn and {Schreck}, Vanessa and {Riazy}, Shirin and {Simbeck}, Katharina},
biburl = {https://www.bibsonomy.org/bibtex/2c7af5e25fc6318262506a2e639daa863/meneteqel},
booktitle = {2019 IEEE International Conference on Humanized Computing and Communication (HCC)},
doi = {10.1109/HCC46620.2019.00009},
eventdate = {2019-09-25},
eventtitle = {2019-09-27},
interhash = {564d8e40f21cfdda60997f7ea93ac3ac},
intrahash = {c7af5e25fc6318262506a2e639daa863},
keywords = {Arbeitszeugnis Digitalisierung Diskriminierung Künstliche_Intelligenz hbs-2017-369-2},
language = {eng},
month = {Sep.},
pages = {1-6},
publisher = {IEEE},
timestamp = {2020-03-11T10:41:17.000+0100},
title = {Analyzing Sentiments of German Job References},
url = {https://ieeexplore.ieee.org/document/8940848/},
venue = {Laguna Hills, Calif.},
year = 2019
}