Artificial intelligence tools could benefit chemists with disabilities. So why aren’t they?

Hoby Wedler remembers his years in graduate school as many computational chemists do: long hours writing code, running calculations, and poring over papers. Unlike many computational chemists, however, Wedler had a constant companion while he did those things—an assistant who could do things like describe figures in the article he was reading. Because Wedler is blind, he needed the support to complete his research.

Over the past decade or so—at the same time Wedler was working on his PhD—chemists around the world have been thinking about ways they could take repetitive lab work out of scientists’ hands. These new tools include robots that can move around a lab and operate equipment as a human would, as well as automated systems contained entirely within a fume hood to mix reagents.

While large-scale automation has been largely the domain of pharmaceutical companies, low-cost robotics and sensors, combined with more sophisticated artificial intelligence algorithms, have put affordable, customizable, and relatively easy-to-use automation in reach for academic chemists and others working outside the walls of major corporations. These researchers have been developing automation tools that scientists can program to carry out individual chores or even run experiments with little human intervention.

The time is now fast approaching, some say, when these robots could help chemists like Wedler and replace human assistants.

Alexander Godfrey is a chemistry automation consultant with the National Center for Advancing Translational Sciences. Until 2017, he ran an automated synthesis program at Eli Lilly and Company, where he collaborated with groups at other companies and institutions. These groups would run experiments in his lab without ever setting foot in it. They would give instructions for conditions and reagents to the automated synthesis setup, press “go,” and then collect the results. That got him thinking about chemists with disabilities. Veterans, for instance, could continue their careers with similar tools even after a serious injury, Godfrey says. Still, Godfrey says he doesn’t know of any chemists with disabilities who have tried to use the automation systems he helped develop at Lilly.

Chemists with disabilities are already working in academia and the public and private sectors, but they do so without automated tools. Some have physically modified their laboratories to accommodate a wheelchair. Others have enlisted the help of software that can read text aloud.

Greg Williams, a computational chemist who is blind, works for a company called Independence Science that is developing accessibility technology for STEM (science, technology, engineering, and math) education. The firm has added speaking software—a basic version that doesn’t rely on artificial intelligence—to a commercial data-collection device called LabQuest 2. The software speaks aloud results and data that would normally be displayed on a screen, providing students with visual impairments easier access to results. Williams says adaptive equipment has become more common in classrooms but is much rarer in academic and industry labs.

Chemists with disabilities are well aware that automation could give them new ways to work. In Wedler’s case, maybe a robotic assistant could have replaced the human one he worked with. But these chemists are skeptical that they’ll be using these tools anytime soon. Existing automation technology is too expensive for most researchers on an individual basis. And with a relatively small number of people with disabilities working in chemistry, some think there may not be enough demand to spur progress toward automation tools designed for them to use.

It makes for a frustrating problem. People with disabilities may choose not to pursue chemistry because the field isn’t accessible.

“It really is a catch-22,” says Mona Minkara, a computational chemist. Like Wedler and Williams, Minkara is blind. She recently finished a postdoc at the University of Minnesota and is now applying for faculty positions. “There aren’t technologies being made for blind scientists because there aren’t blind scientists. There aren’t blind scientists because there aren’t technologies to support them,” she says.

Automated tools might encourage more people with disabilities to go into chemistry. But the small population might explain why all the automation researchers C&EN spoke with said they are interested in automation’s potential to make chemistry more accessible, but few of them are working on it explicitly.

One researcher C&EN contacted for this story said automation to increase accessibility is very important but declined an interview because he didn’t want to set the expectation among members of the community that he’s working in that direction.

Some of the researchers that C&EN spoke with were inspired to think about how automation could make chemistry labs more accessible after experiences with scientists who found themselves temporarily unable to work as they normally did.

Andrew I. Cooper of the University of Liverpool has a story like that. His group has been developing an autonomous system to help discover new photocatalysts. The team’s robot rolls around a modified laboratory manipulating samples and equipment with a customized mechanical arm. Artificial intelligence algorithms help it decide how to move and, on the basis of the results of its experiments, what to do next.

Cooper recalls a time before his group was building these robots when one of his students broke a leg playing soccer. The student spent the next few months writing a review paper, which Cooper says is the kind of thing you do when you can’t work at the bench. “But had we had the system,” Cooper says, the student could have been doing his syntheses instead. Although Cooper says his group has discussed how automation could make labs more accessible, the team is not currently adapting its technology for chemists with disabilities.

Nicola L. B. Pohl had a similar experience to Cooper’s when one of her students was on crutches after an accident. Pohl is developing automated synthesis tools with her carbohydrate chemistry group at Indiana University Bloomington. Watching that student try to work, she saw how an automated system could make chemistry more accessible. The student had trouble using his hands and balancing in front of his fume hood at the same time. With the group’s automated equipment, all he had to do was mix up reagents and load them into the system, and then he could control his experiments from his desk. Pohl says she didn’t see a huge drop in her student’s productivity during his recovery. “That gets me excited about the possibility of being able to put this in labs everywhere—that you can start accommodating people with all their different abilities,” Pohl says.

Pohl has incorporated what she’s learned from that student and others into her automated systems. She started adding volatile organic compound sensors to her lab’s fume hoods after working with a student who didn’t have a sense of smell. Pohl is committed to making tools designed to help people with disabilities. But she says she hasn’t spoken with disabled chemists yet about how automation could make science more accessible for them.

Reaching out to chemists with disabilities is on Pohl’s to-do list, and she says she hopes to find a person with disabilities in her area whom she could work with. She explains that she wants to release a working version of her system before she starts augmenting it with new elements, which could include more accessibility features.

One group of researchers is actively thinking about designing for chemists with disabilities. Nitesh V. Chawla and Olaf G. Wiest of the University of Notre Dame; Abigail Doyle of Princeton University; Robert Paton of Colorado State University; Richmond Sarpong of the University of California, Berkeley; and Matthew S. Sigman of the University of Utah are working on a US National Science Foundation proposal to fund a center for developing computer-aided synthesis technology. This kind of software uses machine-learning algorithms to plan feasible synthetic routes to target molecules. The group has talked to members of the American Chemical Society Committee on Chemists with Disabilities about how the group’s artificial intelligence tools could help scientists with disabilities. ACS publishes C&EN.

Like other automation researchers, Wiest says his interest in accessibility stems from interactions with students with disabilities. Like Minkara, Wiest acknowledges that there is little apparent demand for accessibility tools because there are so few chemists with disabilities working in the lab. But, he says, “if we can help a handful of people, we’re going to double the number of opportunities” in chemistry for people with disabilities.

Wiest’s group’s decision to reach out to the ACS Committee on Chemists with Disabilities was vital in terms of connecting with the community. Chemists with disabilities say they’d prefer that automation tools are created with accessibility in mind from the outset, rather than tools being retrofitted later. “Retrofitting is always more complicated and more expensive,” Independence Science’s Williams says. He says writing software code to accommodate users with disabilities is much easier than redoing thousands of lines of code to add those features.

People developing new technology should also be thinking from the beginning about how users will interact with machines, says Brad Duerstock, an engineer at Purdue University. “Having multimodalities in user interfaces, to me, would be easier to do at the forefront and save a lot of headaches,” he says. Duerstock has tetraplegia, and his first attempt at accessibility engineering was making a microscope so he could finish his neuroscience PhD. Now he’s developing software that can translate images for blind scientists using sound or touch from a haptic device. One of his PhD students, Ting Zhang, is adding an artificial intelligence component that can anticipate users’ needs and make the software more efficient.

Minkara agrees that these tools will better serve chemists with disabilities if accessibility is baked in at the start, but she’s pessimistic that will happen. She thinks researchers will more likely develop and optimize their automation technologies for a broader audience first and that chemists will need to fit accessibility tools to the systems that are available.

On her wish list for artificial intelligence–powered accessibility? Maybe a way to interact with her computer by voice rather than typing. Wedler wants software that could describe graphs and figures. Also, if he’d had a robotic system that could run reactions so he could test his calculations experimentally—maybe something like what Godfrey or Pohl have worked on—he says he might have stayed in academic chemistry.

Williams says being able to run an experiment in the lab is a great goal, but most people with disabilities can’t even interact with computers and other equipment efficiently.

He believes that artificial intelligence and automation tools could make chemistry more accessible to people with disabilities. But he says that outcome is most likely if automation researchers engage people with disabilities now, while they’re still developing automation tools. Those interactions, Williams says, will increase the chances that accessibility tools are really useful for the people who will use them.