A year-long research project exploring how text-to-image generators fuel or fight design fixation.
Methods
Research-
through design
Participatory
design
Experimental
studies
Team
Angela Truong
Jenny Tan
Sam Gillespie
Francisco Ibarrola
My Role
Main researcher
Workshop Facilitator
Data Collector and Analyst
Timeline
12 months
Research Scope
This research explores how text-to-image (T2I) Generative AI (GenAI) tools impact design fixation, focusing on their potential to break fixation and support early ideation. Using two studies—exploratory co-design workshops and an experimental study—it examines how designers interact with GenAI systems and how visually diverse outputs can spark fresh ideas.
The Research Questions
How does Generative AI influence design fixation in creative processes, and what reduce or increase this effect?
How can AI systems help designers explore different perspectives and consider more ways to reframe their ideas?
Our Methods
Co-design workshops provided qualitative insights into how fixation occurs in a designer’s creative processes and the attitudes and strategies that influence it.
The experimental study introduced DiversifAI, a T2I system developed based on insights from the co-design workshops. This tool was used to test how visual diversity helps reduce fixation by promoting idea reframing and divergent thinking.
Both studies used an index called the MICSI (Mixed-Initiative Creativity Support Index) survey to quantify creativity support, human-AI collaboration, and fixation levels of participants.
What We've Found…
There is a relationship between alignment, commitment, and diversity.
Lower fixation effects occurred when users were open to new ideas, the AI provided diverse outputs, and there was minimal misalignment between user expectations and AI results.
GenAI needs to explain its processes to users.
Miscommunication between participants and AI was a common issue, as users felt the AI often ignored visual details or misinterpreted prompts, leading to frustration. Many preferred traditional methods due to the AI's inability to fully understand them, stemming from the limited communication channel of prompts.
Fixation can be disrupted by unexpected outcomes.
Participants were often “pleasantly surprised” by unexpected compositions, styles, or quality, which disrupted their cognitive patterns and inspired creativity. While not all deviations were welcomed, many found value in serendipitous results, such as discovering unknown art styles that reshaped their vision.
Implications
Designers should be trained to work with AI as an iterative collaborator, experimenting with prompts and remaining open to unexpected results. Encouraging adaptability and AI literacy in education can help reduce fixation and improve creative workflows.
Fixation is a dynamic process influenced by alignment, commitment, and diversity, showing that AI can both hinder and enhance creativity. Unexpected AI outputs can break cognitive patterns, but managing their balance is key to effective creative support.
AI tools need better communication methods beyond text prompts, such as sketch inputs or interactive refinements, to align with user expectations. Enhancing transparency and user control over diversity can reduce frustration and make AI collaboration more effective.
