音樂偏好的心理學：人工智慧作品與人類作品之比較

Psychology of Music Preference: Comparison Compositions Between Human and Artificial Intelligence

汪子翔(Zih-Hsiang Wang)

指導教授 ： 黃從仁

共同指導教授 ： 黃貞穎(Chen-Ying Huang)

國立臺灣大學/理學院/心理學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202002375

摘要

創造力與製作藝術被認為是人類所獨有的能力。雖然人工智慧（AI）技術曾被應用在創作藝術上，相關文獻卻指出相較於人們對於AI作品的喜好程度較低。然而，許多這類比較AI音樂與人類音樂的研究並沒有嚴謹控制（如樂曲調性與演奏者等）會影響音樂賞析的要素，於是無法確認是AI音樂本身尚未達到人類音樂的水準，抑或是人們對於AI音樂帶有負面的主觀偏見。因此，本研究控制這些混淆便項設計三個關於古典音樂子實驗，並透過音樂情緒及音樂認知問卷的填答，去回答兩個研究問題：人們對於AI音樂是否帶有偏見以及何謂音樂中的人性。在子實驗一觀察到AI音樂與人類音樂在盲測的主觀評定上沒有任何顯著差異。在子實驗二中觀察到人們的第一種偏見──把好聽的分類成人類音樂、難聽的分類成AI音樂的傾向。在子實驗三中則觀察到第二種偏見──若在聽音樂之前便知道作者是AI所做會給予較低的評價。此外，本研究發現受試者對於音樂是否為人類所做的判斷，跟人們是否喜歡該首音樂類似；說AI音樂缺乏人性相當於是第一種偏見。

關鍵字

人工智慧 ； 古典音樂 ； 音樂情緒 ； 音樂認知 ； 電腦藝術偏見

並列摘要

Creativity and art are considered to be unique to human beings. Although artificial intelligence （AI） technology has been used in creative arts, relevant literature has pointed out that people less prefer AI artworks than human ones. However, many of these studies do not strictly control the factors that affect music appreciation, such as melody and performers. Therefore, it could not conclude whether AI music itself has not reached the level of human music yet. Either it could not conclude whether people have a negative subjective bias against AI music. After controlling these confounding variables, this study designed three sub-experiments on classical music. We aim to answer two research questions through music emotion and music cognition scale （MEMC）: whether there is prejudice against AI music and what humanness is in tune. In Study 1, a condition of blind testing, no significant difference is observed between AI music and human music in the subjective assessment. The first kind of prejudice was found in Study 2 — the tendency to classify the euphonious excerpts as human music and the unpleasant ones as AI music. In Study 3, the second kind of prejudice was observed — knowing the author beforehand will cause a lower rating on AI excerpts. Besides, this study found that the judgment made by the subjects on whether the music is human is similar to whether people like music. The sense of lacking humanness in AI music is likely to be equivalent to the first prejudice.

並列關鍵字

artificial intelligence ； classical music ； musical emotion ； musical cognition ； computer-art bias

參考文獻

Agres, K., Forth, J., Wiggins, G. A. (2016). Evaluation of musical creativity and musical metacreation systems. Computers in Entertainment, 14(3), 3.

Aljanaki, A., Wiering, F., Veltkamp, R. C. (2016). Studying emotion induced by music through a crowdsourcing game. Information Processing Management, 52, 115-128.

Bartel, L. R. (1988). A study of the cognitive-affective response to music (Unpublished doctoral dissertation). University of Illinois at Urbana-Champaign, Champaign, IL.

Bartel, L. R. (1992). The development of the cognitive-affective response test—Music. Psychomusicology: A Journal of Research in Music Cognition, 11, 15.

Boden, M. A. (2010). The Turing test and artistic creativity. Kybernetes, 39(3), 409-413.