A group of researchers, including some from the University of Maryland, are developing a new video-calling platform that aims to assist in accommodations and minimize miscommunications between autistic and non-autistic people in majority-neurotypical workplaces.

The Fostering Inclusivity through Technology platform is a video interface that seeks to remedy workplace miscommunications between autistic and non-autistic people by analyzing facial features, speech and language produced over the course of a conversation. This analysis helps identify markers that a misunderstanding may be happening between autistic and non-autistic people or when a miscommunication might occur.

As winners of the university’s Grand Challenges Team Projects Grant, the FIT project will receive up to $1.5 million worth of funding over the course of three years.

“What’s so important about this technology and our approach is that we’re not creating technology to try and fix anyone,” assistant research professor and assistant director of this university’s language science center Shevaun Lewis said. “What we’re trying to fix is conversations.”

[UMD and IonQ celebrate National Quantum Laboratory opening]

Autistic and non-autistic people think differently from one another, which can lead to miscommunications, according to FIT platform researchers. Every autistic person experiences autism differently, but some commonality among workplace experiences can be found in having a different understanding of social rules from non-autistic people.

Some of the differences autistic people face include how they comprehend and clarify task-related directions given by non-autistic bosses or colleagues and feel the pressure to mask — meaning to camouflage their authentic selves by mirroring non-autistic people — in the workplace, university researcher Quentin Leifer said.

Leifer also said autistic people would often benefit from access to tools and environmental changes in the workplace because they sometimes experience sensory sensitivities to workplace noises or lighting.

Lewis, who is director of this university’s autism research consortium, emphasized that miscommunications between autistic and non-autistic people are two-sided. Non-autistic people have as much difficulty understanding autistic people as autistic people have understanding non-autistic people, according to FIT platform researchers.

The FIT platform will look similar to other video-calling applications — like Zoom, Microsoft Teams and Skype — but will have additional features that are specific to the project, according to Andrew Begel, an associate professor of computer science at Carnegie Mellon University who works on the project.

Specific functions of the FIT platform are still being considered and developed based on research participant feedback, according to Leifer. Potential functions might be focused on speech, facial expression, body language and additional non-verbal behavior analysis.

[UMD professors launch collective to assist communities, activists with urban equity issues]

Currently, a list of around 20 potential features for the platform are being examined. Researchers are still receiving feedback from both the autistic community and people connected to them in the workforce to ensure their opinions on how the platform can be beneficial for both autistic and non-autistic people are heard.

Functions being considered by research participants at the moment include one that allows participants to submit text that can be translated into more direct or polite mannerisms and, based upon the large interest shown by the research participants, an agenda generator that takes topic submissions before meetings in order to create a randomized agenda to ensures each participant’s topics are broached during conversation.

Artificial intelligence will also be used to help shape these platform functions by studying conversations of autistic and non-autistic people and analyzing what signs of behaviors to look out for from both participants, Begel said. The FIT platform will have to develop its own non-verbal communication AI because pre-existing models are primarily trained on non-autistic people.

“Autistic body language and autistic facial expressions are very different, and those AI models don’t work,” Begel said. “We’re doing this from scratch.”

A platform like FIT wouldn’t have been possible 15 or 20 years ago, according to Lewis. But with the change in attitude toward autism among researchers and society, she is hopeful the FIT platform will help facilitate workplace conversations between autistic and non-autistic people.