The Role of Artificial Intelligence in Speech Therapy: Part 2

 

Author: Kendra Wormald, Reg. CASLPO Speech Language Pathologist
Date: Wednesday, April 24, 2024

Table of Contents

 

Part 1 of this series reviewed what artificial intelligence (AI) is and a broad overview of its role in speech and language therapy. This entry will focus on how specific populations can optimize on AI and how AI can be incorporated throughout the client life-cycle.

To note, here at WELL SAID,  we do not offer a specific artificial intelligence program such as Jessica. However we do intentionally aim to optimize on the artificial intelligence systems that you may already be interacting with on a daily basis such as Google assist, Siri, chatGPT, Zoom AI Summary and various forms of speech to text and speech synthesis. In our consent forms there is an explicit request for client's consent to use artificial intelligence systems within planning, assessment and intervention of speech therapy. You always have the right to withdraw consent at any time and it is within your right to be fully educated and informed of the pros and cons of these systems. All AI used will be screened and ensured that it is compliant with the Personal Health Information Protection Act (PHIPA).

 

Types of AI

  1. Narrow AI: Designed to complete very specific delegated actions; unable to learn independently 

  2. Artificial General Intelligence: Designed to learn, similar thinking and performances levels to humans 

  3. Artificial Superintelligence: Able to surpass human knowledge and capabilities

  4. Reactive Machine AI:: Capable of responding to external stimuli in real time

  5. Limited knowledge AI: Capable of storing knowledge and use it to learn and train

  6. Theory of Mind AI: Capable of sensing and responding to human emotions; performs tasks of limited memory machines 

  7. Self-Aware AI: Capable of recognizing others’ emotions, has a sense of self and human level intelligence

 

AI potential throughout the client lifecycle

Assessment 

  1. Consolidation of data: AI systems can condense data into a palatable format and summarize findings as it compares to age-matched peers

  2. Explaining Results: When conducting an assessment, there may be statistical information, diagnoses or a significant amount of information which may be challenging to both explain and interpret. AI can generate phrases and words to support communication translation of assessment results. 


Intervention 

1. Therapy Stimuli and Activities: Materials used in therapy such as reading material, picture scenes, opinion questions and basic speech tasks can become redundant. Although repetition of skills is imperative for growth, the stimuli used to support this repetition needs to be engaging, novel, appropriate for the client, and be able to be manipulated to be more or less challenging. Speech-Language Pathologists at times spend countless hours researching various stimuli for those reasons. Generative AI has the potential to develop stimuli that checks all of these boxes while saving the speech therapist time and effort that they could put elsewhere.

Examples: 

  1. Generate a 1-act play for 2 people about XYZ topic with one person having the majority of  speaking lines

  2. Generate list of multisyllabic words with 2-consonant blends in initial positions and /d/ in final position

  3. Generate a list of communication situations that may be be challenging for an adult with _____ diagnosis or disorder

  4.  If a client is targeting the /t/ sound and has a passion for bird watching. The story could be written using this target sound in various parts of the word while embedded in content that is motivating and engaging.

2. AI can also be seen in  augmentative and alternative communication (AAC) systems. AI can support AAC devices to adapt to unique user preferences, communication style and make it easier to access language. Language is more than what we say but also HOW we say it. A lot of AAC users miss out on the ability to express their tone through their device. With AI advancements, users can express their emotions more effectively by pressing a button to describe the tone in which they want their speech to be read . 


3. Evidence Based Practice: Research concerning speech, language and swallowing disorders, approaches, and assessment and intervention protocols is constantly evolving. The scope of a speech therapist is so vast that at times it may be challenging to keep up with research, especially in areas they don't often assess or treat. AI can be used to make suggestions for relevant research and evidence based practice techniques that are specific to targets

Home Exercises

We all know that it can be challenging to do home exercises. We either 'run out of time' or forget to do the work. AI can support consistency of practice to support progress between speech therapy sessions:

  1. AI can be used to not only generate material that is specific to skills and interests but also tailor education and practice to specific learning styles, reading levels and practice modality. 

  2. Reminders, Encouragement and Reinforcement: Let's be honest, even though research recommends spending a minimum of 15 minutes a day on intentional practice, we don't often have a spare 15 minutes each day. This is why I encourage clients to bring practice into routines they are already doing and conversations they are already having. AI can be programmed so that it sends reminders associated with your plans. For example, if you have "coffee chat with Tim" in your calendar, it can provide you with an associated reminder to focus on a specific communication skill during this conversation. Or when you brush your teeth, a reminder can be sent to practice communication affirmations in the mirror right after. 

  3. Your Speech Therapist will be providing some degree of education and explanation during sessions. At times, this information can be challenging to absorb or remember. An AI software can help explain concepts in a more layman's terms to further support your understanding.

  4. Providing feedback, moderating success: An AI system can help you keep track of your progress, provide real-time feedback and be able to reflect on success and recommend taking a break or suggest doing a less demanding task if success is reduced.  

Time-based Tasks: If you're in a crunch for time, you can ask AI to recommend a speech or language task to complete within your time frame, so that you feel you got enough repetitions in to support progress and use your limited time effectively.

 
 

At WELL SAID we work with  a variety of populations. Some of these include those with brain injuries, aphasia or are neurodiverse. Research suggests that AI can support these populations in various ways.


Head Injury 

AI algorithms can tailor rehabilitation programs to individual needs by analyzing all patient data such as medical history, symptom history and changes and recovery progress.


“AI-powered tools play a crucial role in cognitive rehabilitation for TBI patients, aiming to enhance cognitive functions such as memory, attention, and executive functions”  (TBI, 2024).

Advanced speech recognition and natural language processing technologies are able to interpret and translate impaired speech patterns and offer exercises and difficulty level changes based on performance. (TBI, 2024) AI can also support caregivers such as providing symptom tracking tools, education resources, real-time feedback and support, and overall easing the burden of care. 


Aphasia

Word finding, and expressing thoughts as intended can be challenges for people with aphasia (PWA). Research by Purohit, et al. (2023) assessed how ChatGPT can identify intended words for PWA:

PWA Speech Sample

“ and they're gonna go uh dancing and eating probably there in the king's house then dancing. and of course Cinderella has something and then they they take it off. this affect her throw away that stuff and she goes crying she can't go ball at all.”

AI Response

“Ah, I see. It seems like you are referring to the story of Cinderella. In the story, Cinderella is invited to a ball at the king's palace but her stepsisters and stepmother mistreat her and take away her beautiful dress and accessories”

(Iannone and Giansanti 2024)


Neurodivergence: ASD and ADHD
(The ADHD Centre, 2023,Iannone and Giansanti 2024)

Those who are neurodivergent have unique challenges that impair areas such as  task initiation and completion, organization and planning, memory, listening ability, social interactions, and understanding and expressing emotion, among others. Ways in which this population can optimize on AI may include: 

  • Create and organize your calendar

  • Set reminders 

  • “Hey Siri” usually takes less energy than writing it down or using your phone

  • A virtual assistant can send you reminders for appointments, when to leave, create a to-do list and organize them in order of importance.

  • Time management can be supported through helpful suggestions on how to break down tasks into smaller steps, which can make it easier to keep your attention  

  • AI can help you understand a subject better as it simplifies concepts or summarizes information. 

  • Generate a response for an email or text you received. Have it edit what you have written 

  • Adapt to user preferences in real-time such as personalizing accessibility and adapt to evolving needs  

  • Take you through an emotion regulation practice such as box breathing. 

  • Model a conversation relevant to a lived experience. 

  • Help you express your thoughts and feelings.

Try these prompts with AI: 

  • Help me create a daily schedule. 

  • Create reminders for tasks and appointments.

  • Set a medication reminder for a specific time.

  • Add ‘buy groceries’ to my to-do list and prioritize my tasks for today.

  • Optimize my morning routine to make it more efficient and less stressful.

  • Recommend techniques to improve my focus while working.

  • Provide time management tips to help me stay on track 

  • Help me set SMART goals for the week and track my progress.

  • Set reminders for short breaks during the day to practice mindfulness or breathing exercises.

New research on the use of AI in autism screen and diagnosis has shown  how DT-MRI scans can extract images indicating levels of brain connectivity related to their understanding of the brains of children with autism (Khudri, 2023) Khudri notes, “We envision an autism assessment that begins with DT-MRI followed by an abbreviated session with a psychologist to confirm the results and guide parents on next steps.” 

 

Limitations and Considerations

As Kbhudri implies above, we have a responsibility to ensure effective and safe use of artificial intelligence. One major consideration and limitation of AI  is the lack of empathy and nuanced understanding that human therapists posses There is a necessity of human validation and support to ensure AI technologies provide best outcomes, cause no harm while understanding and optimizing on  the complexities of each individual.

Generative AI is still under development and is currently limited to pulling information and content from only the past five years. It’s imperative to be informed of any AI’s terms and conditions, awareness of bias, understanding the basis of how it works and how your information is stored. 

Speak to your Speech-Language Pathologist to learn more about how AI can be used and optimized in speech and language therapy.

 

To speak with a psychotherapist or one of the speech-language pathologists at Well Said: Toronto Speech Therapy, schedule an initial consultation by clicking the link below or calling (647) 795-5277.