My master’s thesis delves into the crucial role emotions play in communication design and investigates the extent to which artificial intelligence can assist in managing emotions.


Designers have the power to influence attitudes and behavior through their actions and must therefore make well-informed, reflective decisions. However, emotions are complex, subjective, and often unconscious phenomena, making it challenging to manage them effectively in design. Can artificial intelligence make emotions more accessible and facilitate critical reflection on design?


To tackle this question, I developed an evaluation and analysis tool called æffekt, which aims to help designers make responsible and empathetic decisions regarding emotional impact. By leveraging the capabilities of artificial intelligence, æffekt enables designers to analyze emotional design effects and evaluate their designs based on the insights gathered.


Literature Review

The project began with a literature review of three central topics: emotions, emotions in design, and emotional artificial intelligence (affective computing). To gain a comprehensive understanding of these topics, I drew from perspectives in psychology, philosophy, computer science, sociology, and cultural studies.

Online Survey

To better comprehend the subjective perception of media consumers, I conducted an online survey. The survey aimed to identify any possible correlations between emotionality and other parameters, such as comprehensibility or credibility. Participants ranked design examples based on emotional and functional parameters.

Expert interview

I also sought the input of experts on AI and emotional design. In three interviews, we explored the role emotions play in design practice, how they can be deliberately addressed, and the potential of artificial intelligence.

First concept sketches of the final application

Proof of concept

For the proof of concept, I created a prototype in Unity that analyzes five examples. The analysis is powered by an image recognition model that was trained using IBM Watson Visual Recognition. The results of the four analyzed parameters (pleasant-unpleasant/calm-arousing) are presented in an intuitively readable color code.


To obtain an initial assessment of the tool’s usability, impact, and added value in design practice, I conducted four tests with designers.


Oct ’20 – Feb ’21


Master’s thesis