Exploring the human bias against songs partially generated by AI

Abstract

Artificial intelligence has been expanding its presence across numerous industries, revolutionizing operations and transforming the user experience. The entrance of AI into artistic fields has sparked scepticism and fear, as creativity is traditionally seen as uniquely human. In music, this has led to a human bias against AI that can be observed when listeners rate a song higher when they believe, or are told, it was created by human artists rather than by AI. This dissertation focuses specifically on AI covers which are songs where AI replaces the vocals of the original piece by the voice and style of another artist. It aims to understand what are the factors contributing to the bias against AI covers and how strongly they influence listeners perceptions. Furthermore, the study examines how the perception of these factors differs between generations. A review of the existing literature identified factors that possibly contribute to this bias, such as believing AI covers lack emotional depth, believing they have an unnatural sound, perceiving AI as unable to be creative, fearing AI will surpass human intelligence and considering AI covers are unethical. To test these hypothesis, in-depth interviews and an online survey were performed, revealing that the perception that AI covers have an unnatural sound is the strongest factor contributing to the bias. Additionally, beliefs regarding emotional depth, creativity and ethics also play significant roles. Finally, the analysis showed that younger generations perceive these factors more negatively than the older generations, revealing stronger biases against AI covers.

Date of Award	7 Feb 2025
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Miguel Rita (Supervisor)