Response to “uncertainty in emotion recognition”
| Pages | 295-298 |
| DOI | https://doi.org/10.1108/JICES-07-2019-0072 |
| Date | 12 August 2019 |
| Published date | 12 August 2019 |
| Author | Katleen Gabriels |
| Subject Matter | Information & knowledge management |
Response to “uncertainty in
emotion recognition”
Katleen Gabriels
Department of Philosophy, Faculty of Arts and Social Sciences,
Maastricht University, Maastricht, The Netherlands
Abstract
Purpose –This study responds to Agnieszka Landowska’s paper about the lack of accuracy in emotion
recognition.
Design/methodology/approach –The approachis purely theoretical. The paper also refers to empirical
studies.
Findings –The author first elaborates on Landowska’s“postulates”(normative guidelines) and then
shortly expands on how virtual chatbots such as “AI therapists”pose considerable challenges to emotion
recognitionalgorithms as well.
Originality/value –This viewpoint’s value is to elaborate and expand on an ongoing discussion on
emotion recognitiontechnologies.
Keywords Regulations, Chatbots, Emotion recognition, Complex emotions
Paper type Viewpoint
The paper “uncertainty in emotion recognition”questions the accuracy of emotion
recognition algorithms. It draws upon a meta-analysis of several studies to show that
uncertainty in emotion recognition is still high. During the past years, Agnieszka
Landowska and her team conducted a series of studies which reveals that one should
not put too much trust in these technologies yet. Landowska aptly remarks that this
unreliability is (ethically) problematic on a number of levels, not in the least regarding
commercial tools that are already on the market, even though their accuracy is low.
Users of these technologies, including people who are affected by their outcomes (e.g. in
job recruitment), have to be informed about this.
In this response, I first elaborateon Landowska’s“postulates”(normative guidelines)and
then I shortly expand on how virtual chatbots suchas “artificial intelligence(AI) therapists”
pose considerablechallenges to emotion recognition algorithms as well.
Postulates
The eight postulates with which Landowska concludes her article serve as concrete
recommendations to improve the current situation. To raise awareness, Landowska states
that producers and developers (software developing companies and individuals) of emotion
recognition algorithms have to provide the uncertainty representation of the results in a
transparent way. In addition, they should communicate clearly about the conditions that
affect the outcomes’validity, such as camera angle (i.e. how faces are filmed) and lighting.
These postulates can be classified within the scope of professional ethics and codes of
conduct.
Next to producers and developers, another involved group is researchers. According to
Landowska, they should provide detailed guidelines and recommendations regarding how
data sets were annotated,because otherwise the validity might be threatened.This postulate
Uncertainty in
emotion
295
Received6 July 2019
Revised6 July 2019
Accepted19 July 2019
Journalof Information,
Communicationand Ethics in
Society
Vol.17 No. 3, 2019
pp. 295-298
© Emerald Publishing Limited
1477-996X
DOI 10.1108/JICES-07-2019-0072
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/1477-996X.htm
To continue reading
Request your trial