Is AI-assisted Creativity an “Original Sin”?: Lay Judgments of Qualities Justifying Copyright Protection for Artworks Derived from AI- vs. Human-generated Sources

This repository contains materials for the Cognitive Science Proceedings 2025 paper “Is AI-assisted Creativity an “Original Sin”?: Lay Judgments of Qualities Justifying Copyright Protection for Artworks Derived from AI- vs. Human-generated Sources” by David G. Kamper, Alice Lin, and Keith J. Holyoak.

Introduction

The rise of generative artificial intelligence (AI) has upended traditional legal and cognitive frameworks governing creativity and authorship. A central issue in this disruption is whether AI-assisted visual artworks—products of human-AI collaboration—qualify for copyright protection under current law. Existing copyright doctrine, rooted in human authorship and the originality of fixed expressions, faces significant challenges when applied to works co-created by humans and AI systems. A pivotal example is the 2022 case of Jason Allen’s “Théâtre D’opéra Spatial,” an AI-generated artwork denied copyright registration by the U.S. Copyright Office (USCO). Despite Allen’s extensive effort—refining 624 text prompts, iterating within Midjourney, and making post-hoc edits in Adobe Photoshop—the USCO ruled that the work’s “traditional elements of authorship” originated from the AI system rather than human creativity (USCO, 2023). This decision reflects a legal framework that rigidly distinguishes human and machine contributions, relegating AI-assisted works to the status of “mechanical reproductions” unworthy of copyright protection.

From a cognitive science perspective, this strict dichotomy raises fundamental questions. Empirical studies suggest that human evaluators struggle to reliably distinguish AI-generated from human-created art (Sun et al., 2022; Nightingale & Farid, 2022), and that professional artists using AI tools often produce works perceived as more creative than standalone AI outputs (Orwig et al., 2024). These findings challenge the assumption that AI operates merely as a “tool,” instead positioning human-AI collaboration as a hybrid creative process that transcends traditional authorship paradigms.

This research examines two interrelated questions:

Legal-Cognitive Tension: How do lay evaluations of AI-assisted artworks align with copyright law’s criteria for protection?
Creative Agency: To what extent does human effort in guiding AI systems influence perceptions of authorship and creativity?

To address these questions, we conducted two preregistered experiments. Study 1 employed a mixed factorial design to investigate how modification level (none, slight, dramatic) and creator type (AI vs. human) shape lay evaluations of three legal criteria: transformativeness, essence change, and creativity. Study 2 extended this inquiry by manipulating creative effort (less than 1 hour, 10 hours, 100 hours) in AI prompting versus human creation, probing whether labor invested in refining AI outputs impacts judgments of copyrightability. Our findings suggest that lay evaluators prioritize perceptual transformation over creator attribution, challenging the legal assumption that AI-generated content inherently lacks human authorship.

Preregistrations

Preregistrations for all experiments are available on the Open Science Framework:

Repository Structure

.
├── README.md
├── Code
│   └──  R
├── Qualtrics
├── Data
└── Figures

Code

R

AIArtStudy1_Final.Rmd and AIArtStudy2_Final.Rmd are the primary analyses files. They contain the analyses and visualizations for all experiments in the paper.

Qualtrics

Contains survey flow and questions.

Data

Contains raw data files for each of the experiments.

Figures

These contain the data visualizations generated in R and included in the paper.

Paper Results

All results is analyzed and in visualized code/R/AIArtStudy1_Final.Rmd and code/R/AIArtStudy2_Final.Rmd All paper results are included in the document. Knitted document can be viewed Code/R/AIArtStudy1_Final.html and Code/R/AIArtStudy2_Final.html

Software versions

Analysis was performed in R version 4.2.2.

R package versions are indicated in the knitted analysis file at Code/R/AIArtStudy1_Final.html and Code/R/AIArtStudy2_Final.html.

CRediT author statement

What is a CRediT author statement?

Term	David G. Kamper	Alice Lin	Keith J. Holyoak
Conceptualization	X		X
Methodology	X	X	X
Software	X	X
Validation	X	X	X
Formal analysis	X
Investigation	X
Resources	X
Data Curation	X	X
Writing - Original Draft	X
Writing - Review & Editing	X	X	X
Visualization	X
Supervision			X
Project administration	X
Funding acquisition	X		X