Activists are expressing concerns about the Federal Communications Commission’s (FCC) push to adopt a new phone system for people who are deaf or hard of hearing, saying the services may not meet their needs and are potentially biased.
Under the Americans with Disabilities Act (ADA), qualifying deaf or hard of hearing people have access to the Internet Protocol Captioned Telephone Service (IP CTS), which provides transcription for phone calls, similar to television closed captions, through a combination of technology and human interpreters.
However, the FCC has pushed to authorize allowing Automatic Speech Recognition (ASR) technology, which translates speech into text by computers, as a replacement for the IP CTS service. The FCC proposed a rule on the issue last year, which was adopted in February.
When proposing the change, the FCC cited cost savings and argued the move would ensure the program “remains sustainable for those individuals who need it by reducing waste and thereby bringing under control the exponential growth of the program.”
The agency suggested such growth was driven in part by “individuals who could derive equal or greater benefit from less costly alternatives, such as high-amplification phones.” The FCC also suggested that providers have marketed the service to anyone with any level of hearing loss.
The agency, which emphasizes that it requires minimum quality standards for the services, is now seeking comment on three applications from ASR-only providers by Wednesday. Meanwhile, advocates who expressed reservations during the rulemaking process are similarly voicing concerns ahead of the applications deadline.
Advocates say the technology removes the human element from the service and is not yet ready to replace the existing service wholesale, according to Emily Ladau, a consultant for Clear2Connect, a coalition that works to preserve captioning technology for disabled people.
“Imagine relying on Siri for your most important telephone calls or even a 911 call. Without additional testing and protections, ASR-only service risks unleashing services that are not ready for prime time onto a population of vulnerable users,” Ladau told The Hill.
Many of these users, Ladau added, may not be aware of the existence of higher-quality alternatives they could switch over to.
“So once ASR-services are introduced, users may suffer in silence rather than switch providers: Think about how often your parents or grandparents wind up stuck in bad contracts. The same thing could happen here—an outcome that would be exactly what the ADA was intended to prevent,” she said.
In a statement to The Hill on Tuesday, a spokesperson for the FCC said the criticisms of the change “miss a number of key factors.”
“Automatic speech recognition has long been a part of IP-captioned phone service conversations. All but one of the IP-CTS providers have been using automatic speech recognition (ASR) for years—with a person also sitting in the middle of the call to ‘revoice’ the conversation,” the spokesperson told The Hill. “Last year we clarified that there is no legal requirement that a person be in the middle of the conversation, and we adopted a framework to ensure that introducing a pure ASR system wouldn’t adversely affect service quality.”
“We are currently reviewing applications from companies asking to provide these services. Regardless of the underlying technology being used for captioned phones, the FCC requires minimum quality standards. Any approved provider will be required to meet mandatory minimum standards, including verbatim transcription. And any approved provider must have already demonstrated an ability to meet this and other minimum standards,” the FCC added.
Ladau maintained to The Hill that before the FCC considers a transition, it must conduct a broader review of ASR-only providers, accounting for features such as accuracy, speed, synchronicity, privacy practices and suitability for emergencies.
Advocates are also concerned about potential algorithmic bias in ASR services, a frequent issue raised in regard to artificial intelligence or automated systems.
“The ADA requires that captioning providers must be able to handle ‘all types of calls,’ but it is well documented that, unless specifically trained to do so, ASR technologies often fail for specific categories of speakers, including women, children, people who are older, and minorities,” Ladau said.