Affective Computing: How Can Computers Have Feelings?

To say that machines can have feelings sounds effective, but it is actually a banalization of a more complex topic, which we classify in the field of affective computing and which will be discussed in this article.
The truth is that many AI experts are interested in how the human mind works, so that they can use these mechanisms to teach machines to better understand human actions by becoming more emotionally intelligent.

Reading Time: 3 minutes

affective computing

Illustration: Lenka Tomašević

What is affective computing?   

Affective computing is a branch of artificial intelligence that promotes emotional intelligence in algorithms, for recognizing, processing, interpreting and simulating human empathy. It is an interdisciplinary field that includes computer science, psychology, and cognitive science. 

What does affective computing, in principle, cover? 

Most often we talk about the following areas: 

  1. Facial detection; 
  2. Analysis of facial expressions, gestures and movements of people in pictures and video; 
  3. Detection of emotions based on audio recordings of speech; 
  4. Application of techniques for the analysis of natural speech for the detection of feelings; 
  5. Multimodal approaches for processing social signals in complex social interactions, based on simultaneous sound and video processing. 

Interest in affective computing is growing, and recently there have been important advances in research studies, but also in real-world applications. 

How does affective computing work?   

We will explain how affective computing works on an easy-to-understand example. 


The new Apple Watch Ultra uses sensors to detect body temperature, skin temperature, body posture, movement, and other physiological data that it then communicates with AI systems for health analysis.

Source: Neuroscience News 

What happens after the sensors send this information to the system? 

Machine learning techniques then produce certain “labels”; for example, if a person frowns, computer vision may begin to recognize it as “confusion,” “concentration,” or “mild anger.” 

Something similar happens when it comes to changes in speech and voice. Some emotions are easier to identify in this sense, for example, when the voice becomes elevated if the person is excited or angry, while, if tired, the speech is quiet, monotonous and slow. 

What does the accuracy and accuracy of this data depend on? 

As in other branches of artificial intelligence, these systems depend largely on the data they are fed. This is also one of the biggest challenges – to choose the right material for training the system. 

That’s why, for these purposes, the voice of those who best govern emotions and voice, namely actors, is most often used. Not only are they capable of producing variations of basic emotions in a controlled environment, but this means they can provide high-quality audio recordings, optimally articulated expressions, for later voice detection. 

Where can you learn affective computing?   

It is also an indispensable element in formal education, so, for example, students of the Faculty of Technical Sciences in Novi Sad meet with it at master studies, within the subject “Emotional Artificial Intelligence and Affective Computing”. 

“The aim of the course is to introduce students to the ways of using methods and techniques of artificial intelligence and machine learning to analyze people’s emotions and process social signals (observable behaviors of people in social interactions, which are related and reflect the internal emotional state of people). As part of the course, students will be introduced to the most influential approaches to the analysis of emotions based on classical methods of artificial intelligence, as well as deep learning,” reads the course description. 

Upon completion of this course, students are trained to use classical artificial intelligence and deep learning techniques to solve practical problems in the field of information technology. In addition, it will master the practical skills of developing software solutions, using OpenCV, scikit-learn, Caffe, TensorFlow and PyTorch environments to develop artificial intelligence systems. 

Where can affective computing be applied?   

Applications of affective computing are numerous and can have many benefits for everyday life, such as: 

  • Education – This is especially important for distance education. Based on this data, teachers could obtain information about students’ feelings and thus adapt their teaching methods and plans. 
  • Healthcare – Robots increasingly used in healthcare will be significantly improved thanks to affective computing. They can be of particular benefit in countries whose populations are aging and where there are fewer and fewer doctors to provide them with health care. In addition, these robots can also have applications in communication with people on the autism spectrum. 
  • Video games – Affective video games use user information in a completely new way while playing a particular video game. This information is called biofeedback and refers, for example, to the strength with which the player presses the joystick, which testifies to his excitement. 
  • Other – Some of the other applications relate, for example, to the identification of the user after they have tried a particular product. In this way, companies can decide whether to change something before launching a product on the market. 

A journalist by day and a podcaster by night. She's not writing to impress but to be understood.