Dutch DPA: ‘AI systems that can recognize emotions are dubious and risky’
Businesses are increasingly using artificial intelligence (AI) to recognize emotions in people. However, this is based on controversial assumptions about emotions and their measurability. This poses serious risks and ethical questions.
That’s the conclusion of the Dutch data protection authority (DPA) in the latest edition of its Report on AI & Algorithms Netherlands (RAN).
From social interaction to perception, how we learn and what decisions we make: emotions play a defining role in our daily lives. A growing number of AI systems claim to be able to recognize emotions based on biometrics.
According to the privacy regulator, more and more organizations are using AI-enabled emotion recognition because they think it can help to improve their products and services, marketing, and customer contact.
The market for such applications has steadily grown in recent years. However, the fundamental assumptions underlying these systems are shaky at best. Therefore, their effectiveness is questionable.
The Dutch DPA concludes that AI systems that are able to recognize human emotions must be handled with extreme caution. The claim of recognizing emotions is based on controversial assumptions. Biometric characteristics such as voice, facial expression, or heart rate are translated into emotions without nuance.
“Emotions strongly affect human autonomy. If you want to recognize emotions, this must be done very carefully and based on reliable technology. That’s often not the case now,” Aleid Wolfsen, Chairman of the Dutch DPA, warns.
The assumption that an emotion is experienced in the same way by everyone is incorrect. Let alone that those emotions can be measured using biometrics. There can be significant differences between cultures in how people experience, express, and name emotions. Differences can also exist between individuals,” Wolfsen continues.
The Chairman of the Dutch privacy regulator thinks it’s an ethical question whether we as a society think it’s desirable to use this technology. “This requires social and democratic consideration: whether you want to use these systems and if so, in what form and for what purpose.”
The RAN report concludes by stressing the importance of transparency about the operation and use of such systems so that organizations can make informed decisions.