Researchers have created a new AI-assisted digital art tool designed to help art therapy patients better express themselves while maintaining the efficacy of the process.
The tool, dubbed DeepThInk, was designed by researchers at the Cheriton School of Computer Science and the Southern University of Science and Technology in collaboration with art therapists. DeepThInk grew out of the challenges the therapists faced when the COVID-19 pandemic forced them to conduct their work virtually.
“Not everyone has many art supplies at home,” said Jian Zhao, a Professor at the Cheriton School of Computer Science who directs the Waterloo Visualization research group. “Many of the existing digital art tools either don’t offer enough options or are too complex and intimidating for people who don’t know how to use them.”
DeepThInk was developed to solve this problem. The program incorporates both traditional drawing and painting tools as well as an AI brush that transforms backgrounds or landscapes suggested by the user into complex AI-generated images.
“For example, you start with the AI brush, and draw a big colour segment representing the ocean, then reiteratively generate ocean landscapes until you find one that matches your vision for the piece,” said Xuejun Du, a computer science master’s student at the Cheriton School of Computer Science. “After that, you add your own stuff using the user brushes. For example, a boat on the ocean.”
The team collaborated with five registered therapists from the Canadian Art Therapy Association in a reiterative process that lasted 10 months. Their work with the therapists taught them to focus not on using AI to generate complex and flashy images but rather to imagine how humans and AI could work to co-create art.
- Read the full article on Waterloo News
To learn more about the research on which this article is based, please see Xuejun Du, Pengcheng An, Justin Leung, April Li, Linda E. Chapman, Jian Zhao. DeepThInk: Designing and probing human-AI co-creation in digital art therapy, International Journal of Human-Computer Studies, vol. 181, 2024.
