Research map / Large language models / AI and society
Operationalizing Essential Characteristics of Creativity in a Computational System for Music Composition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents Pop*, an autonomous lyrical pop music composition system built around seven operationalized characteristics of computational creativity, and evaluates it with a 125-person survey rating each characteristic and overall creativity.
Problem
Building and evaluating computational systems whose primary objective is creativity, with explicit operationalization of the essential characteristics of creativity in a highly structured domain (Western lyrical lead sheet generation).
Approach
Operationalizes seven characteristics of computational creativity (generation, knowledge representation, intentionality, aesthetic, domain knowledge, autonomy, self-evaluation) in a single music composition system that composes from Twitter-derived intentions using constrained non-homogeneous Markov models, and evaluates each characteristic against overall creativity ratings.
Key results
- Overall creativity (Q12) received the highest average scores across all groups, exceeding ratings for each individual characteristic.
- Novelty (Q6) averaged 3.16-3.77 across songs and was the highest-rated artifact-based characteristic; lyrics (Q2) was lowest at about 2.09-2.63 across songs.
- Participants familiar with the designers rated the system notably higher than those who were not (e.g., overall rating Q1 2.93 vs 2.00).
Surprising findings
- Overall creativity ratings exceeded every individual characteristic rating, which the authors interpret as creativity being an emergent property.
- Respondents who knew the designers rated the system consistently higher, indicating a familiarity bias.
Limitations
- Participants were not blind to the system; familiarity with designers inflated ratings.
- Lyric quality and self-evaluation and aesthetic ratings were weak.
- Single system and single domain (pop/rock/show tunes) limit generalization.
Experimental setup
Datasets: Lyric knowledge base (lyrics.com, 369,606 songs) Wikifonia pop lead sheets (6,673) Twitter posts
Metrics: Likert and star ratings (1-5) per characteristic overall creativity rating familiarity bias breakdown musician vs non-musician breakdown
Builds on: non-homogeneous Markov models hierarchical Bayesian program learning SPECS framework Empath
Closest papers by meaning
- Responding to the Call: Exploring Automatic Music Composition Using a Knowledge-Enhanced Model (AAAI 2024 · similarity 0.86)
- JEN-1 Composer: A Unified Framework for High-Fidelity Multi-Track Music Generation (AAAI 2025 · similarity 0.81)
- Position: Societal Impacts Research Requires Benchmarks for Creative Composition Tasks (ICML 2025 · similarity 0.81)
- Efficient Fine-Grained Guidance for Diffusion Model Based Symbolic Music Generation (ICML 2025 · similarity 0.81)
- MIDI-GPT: A Controllable Generative Model for Computer-Assisted Multitrack Music Composition (AAAI 2025 · similarity 0.81)
- MeLFusion: Synthesizing Music from Image and Language Cues using Diffusion Models (CVPR 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.