AI and facial emotionality: non-inclusive and attentive to diversity
Emotion recognition from images of a face is anartificial intelligence-based technology that can detect an individual's emotional state through use of advanced AI algorithms. This type of technology is going through a significant growth period, with the market worth $12 billion in 2018 and optimistically estimated to grow to $90 billion by 2024.
A rush to use unsupported by scientific research
There are already a number of companies in the human resources sector that use emotion recognition to offer services aimed at screening candidates, sounding out a range of qualities and aptitudes that are difficult to detect with other types of traditional tests. These include, for example, qualities such as"grit." Another detail that is studied by these algorithms, however, is the frequency with which candidates smile. Despite the wide application of this technology, however, research shows that emotion recognition is based on a shaky foundation. Consider, for example, the theory developed by U.S. psychologist Paul Eckman in the 1990s, now widely discredited, which assumes that the expression of human emotions is universal, remaining constant from one culture to another.
>> ALSO READ: AI IN HR PROCESSES: WHAT IS THE RIGHT RATIO?
Companies' efforts toward diversity and inclusion at risk
The use of AI technologies for facial emotionality analysis applied to the evaluation of candidates or employees clashes in no small measure with very important values such as diversity and inclusion. Values that at this moment in history are instead a priority for all companies and that have been engaging all HR departments for some years now in a major collective effort geared toward creating more equitable corporate standards.
Looking first at the issue of diversity, several studies show that the results of such algorithms are strongly influenced by ethnic, gender and, above all, cultural differences. If we then go on to analyze the issue ofinclusion it must necessarily be pointed out that the results are also influenced by the quality of the images provided to the algorithm to make its evaluations. This assumes, especially in the selection phase, that candidates have a device with a good camera at their disposal, thus going to discriminate against all those candidates who have, instead, a medium-low quality device.
>> ALSO READ: THE CV CHALLENGE: ASSESSING CANDIDATES' SKILLS AT THE PRE-SELECTION STAGE
AI and facial emotionality: privacy issues
To all this must be added the issue of privacy, another essential point to make sure that the implementation of a technological solution is actually usable on a large scale. The question arises as to howcompliant these algorithmsactually are with theregulations of the GDPR, since the privacy guarantor in some opinions expressed on the subject has likened theanalysis of facial microexpressions to so-called biometric data. This type of data is regulated particularly strictly by the GDPR, precisely for the purpose of protecting the privacy and dignity of the individual.
For these reasons, it is believed that facial emotionality analysis technologies are still very critical and not ready to be used in the HR world. In contrast, AIs that rely on language analysis, whether in context analysis or text sentiment analysis, prove to be free of such criticalities. This could therefore prove to be aninteresting alternative for all those companies intent on bringing innovation to their candidate and employee evaluation process and who rightly do not want to give up the opportunities that technology can offer them.
