TruaceTracing the truth around AITuesday, August 25, 2026
Media & Arts·P Space·Evidence-backed problem·Published 2026-08-25

Generative AI Art and Creative Subjectivity: A Mixed-Methods Study Based on Grounded Theory and CRITIC

Abstract: With the rapid application of Generative AI technologies in image generation, style transfer, and creative assistance, the process and logic of artistic creation are undergoing a profound transformation.However, research on how AI art platforms influence creators' subjectivity remains limited, and few studies have systematically examined users' real-world experiences.To bridge this gap, this study investigated three representative AI art platforms -Midjourney, Runway ML, and Stable Diffusion -using a mixed-metho…

TRV-2026-0882Peer-reviewedPermanent record — cite & verify
Generative AI Art and Creative Subjectivity: A Mixed-Methods Study Based on Grounded Theory and CRITIC

Maria and Steve work on the new logo of the Netscape Internet browser (fictional photo) by Joseph Ayerle (artist). Public domain

The quick read

A mixed-methods study examined Midjourney, Runway ML and Stable Diffusion to understand how generative AI reshapes artistic subjectivity. Twelve practitioners were interviewed in spring 2025 to build a grounded-theory framework, followed by a survey of 426 users in summer 2025 evaluated with CRITIC weighting.

The work matters because it quantifies what creators value most in human-machine collaboration and what they overlook. While autonomy and customization dominate user priorities, lower weighting for bias and transparency suggests gaps in critical awareness that could affect design of platforms intended to support diverse cultural expression and creative identity.

Main points
  • Mixed-methods study of Midjourney, Runway ML and Stable Diffusion with 12 practitioner interviews March-May 2025 and 426 questionnaire responses June-August 2025.
  • Grounded theory coding produced three core categories, 12 main categories and 29 initial categories, then CRITIC weighting ranked factors.
  • Control over expressive outcomes w1=0.106 and technical accessibility of model customization w9=0.103 were top-weighted factors.
  • Cultural bias in training data and algorithmic transparency received lower weights, indicating limited awareness of structural issues.
Problem

Users of AI art platforms showed limited awareness of structural issues, as cultural bias in training data and algorithmic transparency were rated lower in importance than autonomy and usability.

The rundown

Researchers interviewed 12 experienced AI art practitioners between March and May 2025 and built a framework of three core categories, 12 main categories and 29 initial categories through grounded theory coding.

A 7-point Likert questionnaire was then given to 426 participants from June to August 2025, with CRITIC analysis quantifying relative importance of main categories.

The study synthesizes findings into a mechanism of "Generative AI intervention b2 reconstruction of the creative process b2 changes in subjectivity" to explain how platforms reshape creative control, identity perception and aesthetic decision-making.

Reader signal

How should this claim be treated?

The debate