Artificial Intelligence in Speech-Language Pathology – Expert Interview

“AI has the potential to transform the future of speech-language pathology in numerous aspects,” Van Echo says. “Speech-language pathologists (SLPs) may use information from AI to assist in clinical decision-making and guide treatment recommendations. AI can be also used across the SLP scope of practice, including speech, language, cognitive, and social communication.”

For all its anthropomorphic imitations, AI is still a piece of technology, and the use of tech in speech-language pathology is not new. The pandemic demonstrated that online speech-language therapy could be as effective as its in-person version, and apps on smartphones and tablets have become an important resource for SLPs in therapy sessions. But with any technology, it’s important to find out what it does well and what it doesn’t. 

AI is fast and organized, adept at improving documentation accuracy and generating personalized exercises based on an SLP’s recommendations. It’s also an effective data analyst, collecting and sorting data to help identify communication changes and provide biofeedback during speech production.

But AI has limitations, too. It can be expensive and inaccessible, placing additional technological burdens on those needing an SLP’s services. It requires additional training on the part of an SLP and comes with ethical implications around misuse, privacy, and algorithmic bias. AI is blind to elements that are not easily categorized. It’s impersonal and incapable of forming relationships. Even an AI tool like ChatGPT3 would tell you: without an expert to assist, train, and recontextualize, AI applications in speech-language pathology have little value.

“An ASHA-certified and licensed SLP creates relationships with their clients and students and can better determine their needs and create treatment goals using collected data than machine learning can,” Van Echo says. “Speech-language pathology treatment is client-focused. Meaningful client goals guide treatment. AI and machine learning data may be more accurate when predicting treatment targets, but humans use relationships and discussion of individual needs to determine treatment options and targets.”

Today, hundreds of apps and programs use algorithms to support people with speech and language disorders. The best ones work closely with SLPs every step of the way, from development to deployment to implementation. 

SLP consultations and contributions can help AI applications identify developmentally appropriate targets, incorporate accurate treatment recommendations for lessons, and provide oversight on data collection and storage. SLP consultants can offer insight into cultural considerations to reduce interpreting differences as errors; their ethics ensure the consumer’s welfare and protect the profession’s reputation and integrity.

SLPs will continue to shape how technology is used in speech-language pathology. Some have created their own apps and services, like SmartyEars, that aim to help other SLPs make the most of their sessions. And the ASHA community is a hotbed of research and conversation around the potential and implementation of new technological applications. 

“I am particularly excited about the work done by the ASHA Special Interest Groups to inform each other of best practices and research that is relevant to communication disorders, digital technologies, and AI in both publications and in the ASHA Member Community,” Van Echo says.

Recent publications in Perspectives of the ASHA Special Interest Groups include one study on using speech-to-text as a form of biofeedback support for intervention, and another on how speech analytics can predict neurological disease. The ASHA Leader continues to publish the work of SLPs and audiologists related to AI and machine learning, including AI’s use in helping hearing aids filter speech in noise and even in developing algorithms that translate thoughts into speech. 

“The future of providing direct client care or service to students using AI is unpredictable, but very promising,” Van Echo says. “As with any technology, the benefits could be transformative, but risks for abuse will require monitoring. As professionals, SLPs will need to have some knowledge of digital literacy to be critical of the claims and recommendations of machine learning.”

AI has exciting potential for augmented and alternative communication (AAC), improving users’ experiences with speech-generating devices. As machine learning models train themselves further, the algorithms may become more adept at predicting words, personalizing vocabularies, and adding important characteristics like inflection, volume, and humor to digital voices. But predictive features can quickly become too much, and Van Echo notes that these applications would need to be monitored regularly to verify that the thoughts and needs of the AAC user are attributed to their own ideas and not overly predictive of their language skills.

“It will take people with experience in language and communication, linguistics, ethics, philosophy, and computer science all working collaboratively to create a technology that will support all people,” Van Echo says. “AI use with speech-language pathology treatment should continue to be a support, not a replacement, for expert diagnosis and treatment planning by a highly qualified SLP that will improve the lives of people experiencing communication disorders and difficulties.”