CHAMPAIGN, Ill. — Architecture professors at the University of Illinois Urbana-Champaign are working to create new tools to make the study of architecture accessible to blind and low-vision students.
Architecture professors John Clark and Hugh Swiatek teach design studio classes. Their inspiration for transforming the educational experience for blind and low-vision students comes from their work with Daniel Bein, a sophomore architecture student from Fisher, Ill., who is legally blind.
The professors said there are only a few blind or low-vision students in any architecture schools, so their work is novel territory.
“Part of the reason why we don’t get more blind and low-vision students is because the curriculum and software are inaccessible,” Swiatek said. “There are some career paths where you can’t be blind. I don’t think that’s true of architecture.
“Architecture is an oculo-centric. We try to talk about other sensory conditions, but we privilege the visual in architecture education. In making architecture education accessible to someone who is blind or low vision, we’re thinking about how to rebalance that,” he said.
A primary goal for Swiatek and Clark is to develop a nonvisual feedback loop by creating tools and workflows that allow blind and low-vision students like Bein to have a studio experience comparable to their sighted peers, including the ability to work independently, receive feedback and incorporate it into their designs.
Bein said he decided to study architecture because he wanted to pursue a creative field, and because he knew from taking advanced geometry in high school that he was good at math and at understanding spaces.
“Architecture itself is a fundamentally multisensory art form. It’s just the way we communicate architecture before a building goes up that’s visual,” Bein said.
He can learn the layouts of buildings very quickly and keeps a mental map of each place he visits. He experiences buildings through all his senses ― trailing his hands along a wall, sweeping the floor with his cane to understand its texture and tapping the floor to hear the acoustics of a space.
“I can relate much more to the built environment than the drawn environment. I can’t hear the acoustics of a drawing,” Bein said.
For design work, Bein has been using Microsoft PowerPoint and OpenSCAD, a software program for creating solid 3D models. He positions and dictates the dimensions of simple figures within the programs by entering coordinates and geometric operations.
“It’s all the things you learn in geometry class applied to model building,” he said, but “it’s a very slow process. It’s hard to go back and change something.” PowerPoint isn’t precise enough for architectural drawings, and OpenSCAD is slow and doesn’t integrate with industry standards.
Clark, Swiatek and Bein meet weekly to discuss the ideas they are developing and testing in class and to discuss what is working and where obstacles remain. Their work has included a strong artificial intelligence component, including using AI to build extensions that incorporate text-based commands for graphical software using a universal code such as Python. The goal is to allow Bein to use graphical software in a text-based way, Swiatek said.
People who are blind or have low vision often use screen readers, which read aloud the information below a computer’s cursor. However, screen readers don’t work with the architecture software used for graphical images, Clark and Swiatek said. They are using AI’s large language models to describe images on a screen, as well as specifics that design students need, such as the proportions of a drawing.
One of the biggest hurdles is getting a large language model to describe the scale of a structure. They are quite good at describing how something looks in a flat plane ― Clark calls such a model an “image description machine” ― but they don’t work well in understanding and describing its 3D geometry, though the team is optimistic about the future capabilities of these tools.
“The machine is very good at qualitative descriptions ― a column is made out of this material, it’s this color, this is the design of it,” Bein said. “It can only take an educated guess at aspect ratio, and I don’t know if it’s right. If I wanted to know the aspect ratio of a column or wanted to recreate it later, I couldn’t do that from a picture plugged into an LLM machine.”
Clark and Swiatek also are looking at integrating technology, such as cameras, to capture drawings or models and convert them into data or a tactile representation.
“The heart of our research is to build upon what Daniel has already started on his own ― how to create orthographic and 3D representations,” Clark said.
The curriculum teaches students to use the design process to determine how to meet the challenges of a project, rather than starting to draw with a finished product in mind. It’s a process that requires multiple iterations of a design, incorporating feedback along the way. One of the goals of the project is to speed up the time it takes for Bein to create a design, review it, receive and understand feedback, and revise the design. Even with tools that allow Bein to draw more quickly and efficiently, he still needs to know if what he has drawn accurately represents what he wants and he needs to understand and incorporate feedback ― tasks that are typically done visually.
Clark, Swiatek and Bein are experimenting with tactile tools to develop a rapid tactile feedback loop while designing. These tools include converting 2D and 3D images into tactile drawings using “swell paper” and 3D printers; traditional models made from Styrofoam, cardboard and wire; and a 3D printing pen. Swell paper is a type of paper that, when used to print a drawing, produces raised lines to provide a tactile representation of the design. While working on his final design project, Bein drew on tracing paper with a 3D pen, which lays down a bead of filament ― the same kind used in 3D printing ― to create a raised line.
Such tools need to capture meaningful information, Swiatek said, not just the outlines of a space. The team means to develop a series of tactile graphical standards using different lineweights and textures to illustrate different aspects of architectural designs.
In December, Bein presented his final design project for the fall semester with his partner on the assignment, Ethan Anderson, a senior in sustainable design. The project reimagined the former site of a lead recycling plant in East Chicago, Indiana, and examined using structures to capture sand and soil and restore the dunes that are the site’s original ecology.
Bein and Anderson used a variety of tools to develop and refine their project, including an adjustable 3D-printed model that helped them to decide the placement of canals and buildings on the site. They applied a mix of tactile techniques from the research including Python-scripted computer models output to 3D prints, swell paper, traditional wire and chipboard models, and drawings made with a 3D pen. Bein said the process of drawing and revising the design was considerably faster than other methods he’s been using.
In addition to creating tools to help with the design process, Clark and Swiatek said students like Bein need real-time explanations of what their classmates are doing and of the images their teachers are using in the classroom. The goal is to create tools that will get blind and low-vision students through the entire curriculum and into the working world.
“We teach design studios, and the tools we are developing are for design studios, but they can be used in structures, history or other classes that are visual-forward,” Clark said. “It’s creating a feedback loop for the individual but also allowing them to work in groups, or for the relationship between student and teacher and eventually between designer and client.”
This research was supported by funding from the College of Fine and Applied Arts and the Office of the Chief Information Officer. Clark and Swiatek have consulted with researchers on campus with expertise in related areas, such as computer science, informatics and accessibility both in analog and digital worlds. They are looking for others who might be interested in supporting the technical aspect of their work.
Editor’s notes: To contact John Clark, email jsclark2@illinois.edu. To contact Hugh Swiatek, email swiatek3@illinois.edu.
