We decided to run a co-design brainstorming session with DALL-E. Our intention to include DALL-E in the process was to further explore within a narrowly defined scope. Our 1 hour brainstorming session was broken down into 2 parts: priming and exploration of broad ideas, and refining existing ideas. We recruited 20-30yo HCI students, and ran 2 sessions.
After the brainstorming session, participants were asked to fill out a survey asking about the activities
As expected, individual brainstorming with post-its definitely generated more quantity ideas. However, there is a tendency for individual brainstorms to have similar ideas. In group discussions, people may be able to bounce ideas off of each other, and have participants think about various scenarios. These two, I feel are mainstays of a brainstorming session.
Our assumption with DALL-E was that the tool can help participants have more confidence when creating visual artifacts to share. Additionally, that DALL-E might generate images that may spark more creativity for the participant. We've seen this with a participant who accidentally generated a tooth with a face, and decided this can become a virtual assistant. Another benefit to this was that participants found the tool to be fun which led to better engagement.
Some hurdles was that at times participants could spend too much time trying to perfect the image and use multiple searches, which took more time. Another hurdle was learning what combinations of words to use with DALL-E to get the images that participants are looking for. These take time and a bit of a learning curve.
Overall, I think the use of DALL-E to generate unique ideas was successful, however when deciding to use DALL-E in a brainstorm session largely depends on what the goal of the session is. With a narrow scope, participants with tech fluency and comfortability with AI generators and goal to create more unique ideas, DALL-E can be a great tool. However, when looking to generate a large quantity of ideas - post-its and discussions are unbeatable.