Done
Beyond the Prompt: Measuring the Effects of AI-Assisted Ideation on Human Creative Originality and Ownership
How does AI-assisted ideation change the originality of human creative work, and the sense of owning it?
Human–AI interaction · Computational creativity · Generative AI · Creative cognition and perceived ownership
Empirical study and evaluation framework completed. Originality, self-reported ownership, and perceived agency were measured across unaided and AI-assisted conditions.
Study overview
This study is finished. I compared work made by unaided ideation with work made when a generative model could offer ideas. The tasks shared their instructions, their time limits, and their output requirements. Originality of the artifact, the participant’s sense of ownership, and their sense of agency were measured apart from each other, and the amount of model use was logged rather than assumed. What follows is the design, the measures, and the structure of the findings.
Completed investigation. Done means the study was carried out. It does not name a venue, a DOI, or an affiliation.
Motivation
A generative model can offer ideas during creative work. I compared unaided ideation with AI-assisted ideation, and kept five things apart: the originality of the artifact, the originality of a person’s contribution, perceived ownership, perceived agency, and the amount and kind of assistance.
Questions
How does access to AI-assisted ideation affect the measured originality of human creative outputs compared with unaided ideation?
How does AI assistance relate to participants’ perceived ownership of, and agency over, their creative work?
How do the degree of AI involvement and the nature of the creative task relate to originality and perceived ownership?
Study design
Participants completed the same creative tasks under unaided ideation, with no model, or AI-assisted ideation, where a generative model could contribute ideas. Instructions, time limits, and output requirements were shared. The assisted condition logged prompt count, prompting depth, idea adoption, and edit distance. After the task, participants reported ownership, agency, and creative control. Originality was scored independently and double-blind. Semantic and structural similarity were recorded beside that score, not in place of it.
Measurement
Each construct had its own instrument. A similarity number is not an originality rating, and a self-report of ownership is not either of those. The sources below are the instruments written for this study. No external scale is cited here, because none is named in the study record.
- Human-rated originality
- The study rubric, applied independently and double-blind, on novelty, complexity, and structural coherence. Source: the evaluation rubric.
- Similarity distance
- Semantic and structural distance, kept as a supplementary indicator. It does not replace the human rating. Source: the same scoring pass as the rubric.
- Ownership, agency, and control
- Post-task self-report of perceived ownership, creative agency, and personal locus of control. Creative satisfaction was collected with that set and was not folded into the originality score. Source: the study questionnaires.
- Interaction log
- Prompt count, idea-adoption rate, edit distance, and prompting depth. These describe the assistance. They are not a quality score. Source: the session log.
Findings
The study was carried out. Artifacts from both conditions were scored on the same double-blind rubric, and ownership, agency, control, and satisfaction were reported against how heavily the model was used. Assistance was split by prompt count, prompting depth, idea adoption, and edit distance, not treated as on or off.
The scored results are not in the portfolio record. This page does not state a sample size, a mean, a direction, or an interval.
Ethics and limits
Participants gave consent. Identifying records were anonymized before analysis, and only the data the study needed was kept. The result belongs to these tasks and these participants. It does not travel, on its own, to every artistic field.
Repository and materials
The repository, evaluation rubric, questionnaires, and analysis scripts are not published. This page does not link a file that is not there.