I’m sharing my thoughts on how we can learn to ethically use Chat CPT for speech therapy. Chat GPT will likely become an increasingly reliable and useful tool for SLPs as the technology advances. Despite serious drawbacks, I think SLPs may find ways to use Chat GPT in an ethical and responsible manner for certain tasks. In this article, I share my experience and ideas about the pros and cons of the tool.
Full disclosure: I did try having Chat GPT write this blog post for me, but it was a D- effort. It still took me at least 8 or 9 hours to research and write this. Also, I did enough fact-checking of Chat GPT’s responses to sort the use cases into “successful” and “questionable”, but I didn’t look up every reference. So please don’t base any clinical decisions on any references in this blog post.
Free DIRECT download: Reading passage and questions generated by Chat GPT4 (patient handout). (Email subscribers get free access to all the resources in the Free Subscription Library.)
Outline:
- How we already use AI in our daily lives.
- What is Chat GPT?
- How might we use Chat CPT for speech therapy?
- Limitations and cautions.
- Successful examples.
- Questionable examples.
- Chat GPT may be useful to SLPs but should be used with caution.
- Related Eat, Speak, Think posts.
How we already use AI in our daily lives
Many common activities rely on AI-assisted tools. Many people are aware of some common tools that use AI, including voice-activated assistants like Alexa, social media, Google Maps, Spotify, and YouTube.
AI is also used in online shopping, banking and finance, and email services to enhance customer experience, detect fraud, filter out spam, and provide customer service.
In healthcare, AI plays a key role in diagnosing conditions, identifying risks, planning treatment, and in drug discovery.
AI is already being used in many ways not mentioned here, and it will likely become increasingly common as the technology advances.
What is Chat GPT?
Chat GPT is a natural language processing model that we engage with in a conversational way. I’m not an expert by any means, but my understanding is that Chat GPT is not a glorified Google. It does not simply access a database of information and retrieve information based on your input (called a ‘prompt’).
Instead, the model uses artificial intelligence to answer complex questions and assist with many language-based tasks, such as:
- Writing an essay or a resume.
- Creating an app.
- Writing code or Excel formulas.
- Summarizing articles or books.
- Translating material.
The model isn’t truly “intelligent”. It doesn’t understand the meaning of what it’s saying in the way that we do. But it has access to an astounding amount of information that it’s able to analyze to produce helpful answers.
As SLP’s, we can experiment with Chat GPT to find ways to ethically and responsibly use the tool in our practice. We must always keep in mind that any information we share with Chat GPT is stored and analyzed by OpenAI, so we should never share protected health information (PHI). And we must always take responsibility for making the final decisions about any aspect of patient care.
How might we use Chat GPT for speech therapy?
AI-assisted tools such as Chat GPT are evolving. I’ve found a few current uses for Chat GPT3 (free) and Chat GPT4 (paid), and I’ve also found errors. The most shocking error I found so far was Chat GPT4 making up an entire therapy protocol and citation.
Tools like Chat GPT will be increasingly useful as the models are improved. Here are tasks I think Chat GPT will be able to help us with, now or in the future:
- Medication review.
- Creating stimuli for therapy and home exercise programs.
- Brainstorming solutions for specific problems.
- Finding research (Chat currently doesn’t have access past Sep 2021.)
- Summarizing one or more research papers.
- Learning how to assess and treat specific conditions.
- Developing plans of care for specific patients.
- To create study guides/mnemonics for self-learning.
Limitations and cautions
Currently, Chat GPT has significant limitations, including:
- Quality of the output depends on the quality of the prompt.
- Does not have access to information more recent than September 2021.
- May omit important information.
- Can generate different answers each time it’s given a prompt.
- May produce completely false information that looks legit, complete with a bogus reference.
It’s imperative that SLP’s use any digital tool ethically and responsibly. At a minimum, we should:
- Never share protected health information.
- Not assume that Chat GPT is 100% reliable.
- Never rely on Chat GPT as the sole source of information for any diagnosis or treatment.
- Always ask Chat GPT to cite its sources so that we can fact-check.
Successful examples
I’m going to share several examples that I consider were successful or successful enough to be useful now. I asked Chat GPT to cite its sources, but I didn’t check them (other than one that didn’t look right, which I describe below).
The point of this blog post is to start a conversation on the use of AI-assisted tools, not to make an argument for or against using Chat GPT in clinical practice. I’ve only been using it for a month, and most of that is for private use. I’m sure I will use it immediately for some of the uses below.
Developing materials for therapy and home exercise programs
This may be the best use of Chat GPT for SLP’s at the current time. We can use the model to develop personalized word lists, sentences, reading passages, and other written materials.
For example, I found it very easy to use Chat GPT4 to write a paragraph and a set of questions that I could use for reading or listening comprehension, or short-term memory.
Here is my prompt to Chat GPT4:
Write a paragraph about the history of Italian cuisine with at least 5 sentences, written for adults in easy to understand plain English (ex. around a 60-70 on Flesch-Kincaid scale).
Here is Chat GPT4’s response:
“Italian cuisine has a long and rich history that goes back thousands of years. It started with simple dishes made by the ancient Romans using ingredients like olives, wheat, and grapes. Over time, trade with other countries brought new ingredients, such as tomatoes and potatoes, which became important parts of Italian cooking. In the Middle Ages, Italian chefs started to use more herbs and spices, adding flavor to their dishes. Today, Italian cuisine is famous for its delicious pasta, pizza, and gelato, enjoyed by people all around the world.”
I thought that sounds pretty good! It offers specific details that we can use to assess comprehension and recall.
Now I realize that we can’t trust the model to be 100% accurate, so if it matters, we should fact check our material before sharing with patients.
Using Chat GPT to generate questions
Either way, we’ll want to ask questions. Here’s my prompt to Chat GPT4:
Now write 5 yes/no question, 5 multiple-choice questions, and 5 open-ended questions about the above paragraph. For the multiple choice questions, provide 4 possible answers where one is correct, one is close to being correct, and 2 are wrong. For the open-ended questions, include 2 questions that require inferential thinking.
Here is the model’s response:
Yes/No Questions:
- Did Italian cuisine originate thousands of years ago? (Yes)
- Were olives, wheat, and grapes important ingredients in ancient Roman dishes? (Yes)
- Did the Middle Ages have no influence on Italian cuisine? (No)
- Is Italian cuisine known for its sushi? (No)
- Is pasta a famous part of Italian cuisine? (Yes)
Multiple-Choice Questions:
- Which civilization is most closely associated with the beginnings of Italian cuisine? a. Ancient Romans b. Ancient Egyptians c. Mayans d. Ancient Greeks
- Which ingredient was NOT part of ancient Roman dishes? a. Olives b. Wheat c. Grapes d. Tomatoes
- When did Italian chefs begin using more herbs and spices in their dishes? a. During the Renaissance b. In the Middle Ages c. In ancient Roman times d. During the 20th century
- Which ingredient became an important part of Italian cuisine due to trade with other countries? a. Rice b. Tomatoes c. Corn d. Coconut
- What is NOT a famous Italian dish? a. Pizza b. Gelato c. Tacos d. Pasta
Open-Ended Questions:
- What are some of the key ingredients used in ancient Roman dishes?
- How did trade with other countries influence the development of Italian cuisine?
- What are some examples of famous Italian dishes that people enjoy today?
- (Inferential) Why might Italian chefs have started using more herbs and spices in their dishes during the Middle Ages?
- (Inferential) How might the popularity of Italian cuisine have contributed to its global spread and influence?
Brainstorming solutions to specific problems
One of my patients wanted ideas for activities they could do while sitting. I used Chat GPT3.5 to help me come up with ideas for functional and recreational activities, that I then consolidated into a personalized list that I provided to my patient.
Here are the two lists that Chat GPT came up with:
- Sorting and folding laundry
- Reading books or newspapers
- Solving puzzles (e.g. crossword puzzles, Sudoku)
- Playing card games or board games
- Painting or drawing
- Writing letters or journaling
- Listening to music or audio books
- Watching movies or TV shows
- Using a computer or tablet for email or browsing the internet
- Doing gentle stretches or seated exercises
- Engaging in conversation with a caregiver, family member or friend
- Doing crafts such as knitting, crocheting or beading
- Playing simple musical instruments like a keyboard or harmonica
And the other:
- Dusting surfaces and items around the house
- Wiping down counters, tables, and other surfaces
- Organizing cabinets and drawers
- Sorting and organizing paperwork (e.g. bills, receipts, mail)
- Sewing or repairing clothes
- Sharpening pencils or tools
- Planning and making a grocery list or menu for the week
- Paying bills and managing finances
- Polishing silverware or brass items
- Tending to indoor plants
Summarizing single research papers
Summarizing books, papers, and other documents is a relative strength for Chat GPT. ZDNET offers tips for how to summarize documents. They caution against using the tool to summarize books you haven’t read, as the model may generate inaccurate summaries.
I haven’t used the model to summarize research papers, outside of my failed attempt below in which I asked it to summarize across multiple research papers.
But going forward, I can see using the tool to summarize articles I read. If the summary is accurate, I can refer back to it in the future. If it’s not accurate, I could coach the model into fixing it.
Having fun: A poem about the cranial nerves in the style of Edgar Allen Poe
Once upon a midnight dreary, While I pondered, weak and weary, Over many a strange and curious theory Of the cranial nerves that we all carry.
From the olfactory to the optic, The nerves are numbered, specific and cryptic, With each a purpose unique and terrific, Their functions make us human and terrific.
The trigeminal with its three divisions, Controls the face and all its expressions, While the facial nerve with its many actions, Helps us smile and convey our emotions.
The glossopharyngeal and vagus nerves, Keep our throat and voice box in reserve, And the accessory nerve with its subtle curves, Helps us turn our head with a gentle swerve.
The hypoglossal nerve beneath the tongue, Controls our speech, movement, and song, And the vestibulocochlear helps us hear, Sounds and vibrations, far and near.
These cranial nerves are a marvel to see, A network of cords that sets us free, To smell, to taste, to hear, to speak, And move our bodies, strong and sleek.
So let us cherish these nerves so dear, For they make us human, without fear, And help us navigate this world so clear, With all its wonders and all its cheer.
Questionable examples
I’m sharing a few questionable examples. I haven’t tried to have Chat GPT identify high risk meds or meds that affect swallowing or cognition often enough to know if it’s reliable. I’m uncertain how useful the model will be for analyzing specific web pages for information relevant to SLPs. But the final example is clearly a problem: the model created references out of thin air.
It’s possible that I may have gotten more accurate responses if I had worded my prompts better. And the model is being improved, so it may be more reliable in the future. It’s a learning process!
Medication review
Here is a use case that we may wish to take with a grain of salt. We can use Chat GPT to analyze a list of medications to identify high risk medications, as well as those that may affect swallowing or cognition. I recommend that we use the tool with caution, as I did notice a discrepancy, which I describe below.
Here is a prompt I gave Chat GPT4:
Use the following categories for high-risk medications: antipsychotic, anticoagulant, antibiotic, opioid, antiplatelet, and hypoglycemic. For the following drugs, list the ones that are high risk. Also list any that could affect swallowing or cognition. Tylenol, Metformin, Eliquis, atorvastatin, nitroglycerin, thorazine, Cogentin. Cite your sources in APA format.
Here is Chat GPT4’s response:
High-Risk Medications:
- Antipsychotic: Thorazine (chlorpromazine)
- Anticoagulant: Eliquis (apixaban)
- Antibiotic: None of the drugs listed are antibiotics.
- Opioid: None of the drugs listed are opioids.
- Antiplatelet: Nitroglycerin (discussed below)
- Hypoglycemic: Metformin
Drugs that could affect swallowing or cognition: Thorazine (chlorpromazine), Cogentin (benztropine)
The sources Chat GPT4 cited didn’t seem to be accurate, and the links were broken. I believe the identification of Thorazine, Eliquis, and Metformin as high risk meds is correct. Nitroglycerin is discussed below.
Asking for more details about side effects
As an SLP, I’m interested in learning more about the side effects of Thorazine and Cogentin. When I asked Chat GPT4 to give more details, here is what it wrote:
Chlorpromazine is an antipsychotic medication that can cause sedation, dizziness, and confusion, which can negatively impact cognition. It can also cause anticholinergic side effects, which can affect swallowing by reducing salivary flow and increasing dry mouth. Anticholinergic effects can also contribute to cognitive impairment and delirium in older adults. Therefore, chlorpromazine use can increase the risk of aspiration pneumonia and dysphagia in older adults, especially those with pre-existing cognitive and swallowing disorders (APA, 2019).
Benztropine is an anticholinergic medication that is used to treat the extrapyramidal symptoms (EPS) caused by antipsychotic medications such as chlorpromazine. While benztropine can be effective in reducing EPS, it can also cause anticholinergic side effects such as dry mouth, blurred vision, constipation, and confusion. These side effects can affect swallowing and cognition, especially in older adults with pre-existing cognitive and swallowing disorders (Cogentin prescribing information, 2021).
Questioning a discrepancy between answers
I noticed when I ran this prompt on Chat GPT4, that nitroglycerin was included as an antiplatelet drug. Last week, using Chat GPT3.5, I ran a similar prompt that didn’t identify nitroglycerin as a high-risk drug.
So I asked Chat GPT4 about this:
Do you know why you identified nitroglycerin as an antiplatelet today, but Chat GPT3 free version did not do so a week ago?
Here is the response:
“As an AI language model, I don’t have access to the previous interactions between Chat GPT3 free version and users, so I can’t speak to the specific reasons why nitroglycerin may not have been identified as an antiplatelet in a previous conversation. However, it’s possible that the previous version of the model did not have access to the same updated medical knowledge and information that I do, as my knowledge cutoff is more recent.
“It’s also possible that the previous version of the model did not recognize nitroglycerin as an antiplatelet because nitroglycerin primarily works as a vasodilator, which means it widens the blood vessels to increase blood flow and decrease chest pain. However, nitroglycerin also has antiplatelet effects by increasing the production of cyclic guanosine monophosphate (cGMP), which leads to decreased platelet aggregation (Mayo Clinic, 2022). Therefore, it’s important to note that the classification of drugs can be complex and can depend on multiple factors, including the specific mechanism of action and the intended therapeutic effect.
“Regardless of the reason for the discrepancy, it’s always important to double-check any information provided by AI language models or any other source to ensure its accuracy and relevance to your specific situation.”
I checked with Dr. White, my co-author on posts about dry mouth and osteoporosis drugs, and he said that “nitroglycerin is not a potent anti-platelet drug.”
Analyzing websites for information useful to SLPs
This may be considered an emerging area of strength for Chat GPT4. I wouldn’t consider the model’s responses to be as complete and accurate as I would want for treatment, counseling, or education purposes. This is likely to change over time, as I become more skilled in writing prompts and as the model improves.
I gave this prompt to Chat GPT3.5 yesterday, and to Chat CPT4 today, and I got different answers.
Of the 25 most common inpatient conditions listed on the following site, which of them may be associated with impairments of swallowing, voice, speech, language, or cognition? Be specific in your response. https://www.timeofcare.com/the-most-common-inpatient-conditions/
Here is Chat GPT3.5’s answer:
Several of the 25 most common inpatient conditions listed on the website may be associated with impairments of swallowing, voice, speech, language, or cognition. Here are some specific examples:
- Stroke – Stroke can cause a range of communication and cognitive impairments, including aphasia (loss of ability to understand or express language), dysarthria (difficulty speaking), dysphagia (difficulty swallowing), and cognitive deficits such as memory loss and confusion.
- Pneumonia – Severe cases of pneumonia can cause complications such as aspiration pneumonia, which occurs when food, saliva, or vomit is inhaled into the lungs due to dysphagia.
- Heart failure – Chronic heart failure has been linked to cognitive impairment, including problems with memory and attention.
- Chronic obstructive pulmonary disease (COPD) – COPD can lead to cognitive impairment and depression, which can negatively impact communication and quality of life.
- Acute myocardial infarction (heart attack) – Acute myocardial infarction can cause cognitive impairment, including problems with memory, attention, and executive function.
- Sepsis – Sepsis can cause delirium, a state of confusion and disorientation that can affect communication and cognitive abilities.
- Acute kidney injury (AKI) – AKI can cause cognitive impairment and delirium, which can affect communication and quality of life.
“It’s important to note that these conditions can affect different individuals in different ways and to different degrees, so the presence or absence of cognitive, swallowing, or speech and language impairments can vary widely. It’s also important to note that a thorough evaluation by a healthcare professional is necessary to diagnose and manage any speech, language, or swallowing impairments in patients with these conditions.”
A quick scan of this response shows a couple of omissions. Stroke omits voice impairment. In the original article, COPD exacerbation was the listed diagnosis, which can affect swallowing. And delirium could increase risk of dysphagia.
Here is Chat GPT4’s answer:
“I’m sorry, but I cannot access external websites as an AI language model. However, I can try to provide you with information on some common inpatient conditions that may be associated with impairments of swallowing, voice, speech, language, or cognition.
- Stroke: A stroke can cause various impairments depending on the part of the brain affected. Common impairments include dysphagia (difficulty swallowing), aphasia (difficulty with language), and cognitive deficits.
- Traumatic brain injury (TBI): A TBI can lead to a range of issues affecting swallowing, voice, speech, language, and cognition depending on the severity and location of the injury.
- Dementia: Dementia is a progressive neurological condition that affects cognitive function, which can in turn impact speech, language, and communication abilities.
- Parkinson’s disease: While not strictly an inpatient condition, Parkinson’s disease can cause dysphagia, voice changes, and cognitive impairments, which may necessitate hospitalization for treatment or management.
- Multiple sclerosis (MS): MS can lead to various neurological symptoms that can affect swallowing, speech, language, and cognition.
- Brain tumors: Depending on their location, brain tumors can cause impairments in swallowing, voice, speech, language, or cognition.
“These are just a few examples of conditions that may be associated with the impairments you mentioned. It’s important to note that each patient’s experience will vary depending on the specifics of their condition, the severity of the impairment, and any co-existing medical issues.”
I don’t know why Chat GPT3.5 could access the internet to analyze the website, while Chat GPT4 could not.
Summarizing multiple research papers with Chat GPT3.5
I asked Chat GPT3.5 to summarize ASHA’s Evidence Map on Primary Progressive Aphasia. This is actually the 3rd iteration of my prompt, as the first two didn’t provide specific enough information.
NOTE: I’m sharing Chat GPT’s responses as a point of discussion; I’m not researching and writing about PPA in this article. We know the model can be unreliable, so please don’t base any clinical decisions on the information presented here.
Summarize the research papers listed on ASHA’s Evidence Map for Primary Progressive Aphasia (https://apps.asha.org/EvidenceMaps/maps/landingpage/269b08de-6017-ed11-8135-005056834e2b) being specific about assessments and treatments that SLPs can use. Include references in APA format.
Here is Chat GPT3.5’s response.
“The ASHA Evidence Map for Primary Progressive Aphasia (PPA) includes research papers published between 1990 and 2020 that investigate various aspects of assessment and treatment of PPA. The following is a summary of the research papers listed on the evidence map, focusing on assessments and treatments that speech-language pathologists (SLPs) can use:
Assessment:
- The use of the Arizona Battery for Communication Disorders of Dementia (ABCD) is supported for the assessment of language deficits in PPA (Gorno-Tempini et al., 2004).
- The Progressive Aphasia Severity Scale (PASS) is a reliable and valid measure of language impairment in PPA and can be used to track changes over time (Blandin et al., 2018).
- The Test of Early Language Development (TELD-3) and the Clinical Evaluation of Language Fundamentals-Fourth Edition (CELF-4) can be used to assess language abilities in individuals with PPA (Murray et al., 2019).
Treatment:
- Intensive language therapy programs, such as the MIT language program, can improve language abilities in individuals with PPA (Wilson et al., 2010).
- Constraint-induced language therapy (CILT) can improve language production and comprehension in individuals with PPA (Chapey et al., 2013).
- Communication partner training, such as the ASHA FACS program, can improve communication and reduce caregiver burden in individuals with PPA (Boyle et al., 2017). (The ASHA FACS is not this! And this is the reference I can’t find.)
“Overall, the research supports the use of various assessments and treatments for individuals with PPA, including the ABCD, PASS, TELD-3, CELF-4, MIT language program, CILT, and communication partner training.”
Again, the references don’t seem to exist.
Chat GPT3.5 generated a false paper that sounds plausible
Ok, so if you read the treatments listed for PPA, you may have noticed the ASHA FACS program. I know the ASHA FACS as something different, so I tried to find the reference provided.
The Aphasiology paper that corresponds to 31(8), 901-918 is not the paper Chat GPT3.5 summarized. In fact, I couldn’t find any paper by this name or with these authors.
Please describe ASHA FACS communication partner training program (Boyle et al 2017) in detail.
Here is Chat CPT3.5’s response:
The ASHA FACS (Facilitating Augmentative and Alternative Communication and Supports) program is a communication partner training program developed by the American Speech-Language-Hearing Association (ASHA) to support individuals with primary progressive aphasia (PPA) and their communication partners.
The program is designed to improve communication, reduce caregiver burden, and enhance quality of life for individuals with PPA and their families. The ASHA FACS program consists of six modules that cover various aspects of communication partner training, including:
- Introduction to PPA and communication partner training
- Augmentative and alternative communication (AAC) supports
- Environmental modifications
- Communication partner strategies
- Advanced communication partner strategies
- Putting it all together: individualized communication plans
Each module includes a detailed description of the content, learning objectives, and activities. The program also provides resources for communication partners, such as handouts, videos, and case examples.
The ASHA FACS program emphasizes the importance of individualized communication plans that are tailored to the specific needs and abilities of the individual with PPA and their communication partner. The program also emphasizes the use of AAC supports and environmental modifications to enhance communication and reduce frustration.
Research has shown that communication partner training programs, such as the ASHA FACS program, can improve communication and reduce caregiver burden in individuals with PPA and their families (Boyle et al., 2017).
Overall, the ASHA FACS program is a comprehensive and evidence-based communication partner training program that can support individuals with PPA and their families in improving communication and enhancing quality of life.
Reference:
Boyle, M., Tippett, D. C., & Festa, E. K. (2017). Communication partner training for individuals with primary progressive aphasia and their partners. Aphasiology, 31(8), 901-918.
The outline of the therapy program sounds appropriate, so I was hoping to find where it came from. Here is the prompt I gave:
Can you give the link to that reference (Boyle et al 2017) so I can find the paper on the internet?
And here is Chat GPT3.5’s answer:
Certainly! Here’s the link to the paper:
Boyle, M., Tippett, D. C., & Festa, E. K. (2017). Communication partner training for individuals with primary progressive aphasia and their partners. Aphasiology, 31(8), 901-918. https://www.tandfonline.com/doi/full/10.1080/02687038.2017.1301438
The link is not good. When I pointed that out to Chat GPT3.5, it apologized and gave another link, which also was not good.
Summarizing multiple research papers with Chat GPT4
I ran the same prompt through Chat GPT4. I think it’s fair to say that if I were to look up and read every reference below, that it is quite possible some don’t exist or are mis-attributed. But I’ll share the response, so that you can see for yourself how plausible it looks.
When I ran the same prompt on Chat GPT4, asking for it to summarize the research on ASHA’s Evidence Map for PPA, here is what it said:
“I’m sorry, but as an AI, I cannot access or browse the web in real-time. However, I can help you summarize research papers on Primary Progressive Aphasia (PPA) and list some assessments and treatments that speech-language pathologists (SLPs) can use based on the knowledge I have up to September 2021.”I’m sorry, but as an AI, I cannot access or browse the web in real-time. However, I can help you summarize research papers on Primary Progressive Aphasia (PPA) and list some assessments and treatments that speech-language pathologists (SLPs) can use based on the knowledge I have up to September 2021.”
- Assessment: Speech-language pathologists can use several tools to assess PPA. Some of these tools include:
- The Western Aphasia Battery (WAB) (Kertesz, 1982)
- The Boston Diagnostic Aphasia Examination (BDAE) (Goodglass et al., 2001)
- The Progressive Aphasia Severity Scale (PASS) (Rogalski et al., 2011)
- Treatment: Although there is no known cure for PPA, some treatments and interventions have been shown to improve the quality of life and communication skills for individuals with PPA. Some interventions that SLPs can use include:
- Speech and language therapy: Focusing on enhancing communication and compensating for language difficulties (Henry et al., 2013)
- Augmentative and alternative communication (AAC) devices: Using high- or low-tech devices to supplement verbal communication (Beukelman et al., 2011)
- Cognitive-linguistic therapies: Addressing underlying cognitive deficits that contribute to language impairments (Croot et al., 2009)
- Group therapy: Providing opportunities for social interaction, emotional support, and practicing communication strategies (Cummings et al., 2012)
Here are the APA formatted references:
- Beukelman, D. R., Garrett, K. L., & Yorkston, K. M. (2011). Augmentative communication strategies for adults with acute or chronic medical conditions. Baltimore, MD: Paul H. Brookes Publishing Co.
- Croot, K., Nickels, L., Laurence, F., & Manning, M. (2009). Impairment- and activity/participation-directed interventions in progressive language impairment: Clinical and theoretical issues. Aphasiology, 23(2), 125-160.
- Cummings, L., Patterson, K., & Code, C. (2012). Communication in primary progressive aphasia: A literature review. Aphasiology, 26(7), 947-976.
- Goodglass, H., Kaplan, E., & Barresi, B. (2001). The assessment of aphasia and related disorders. Philadelphia, PA: Lippincott Williams & Wilkins.
- Henry, M. L., Rising, K., DeMarco, A. T., Miller, B. L., Gorno-Tempini, M. L., & Beeson, P. M. (2013). Examining the value of lexical retrieval treatment in primary progressive aphasia: Two positive cases. Brain and Language, 127(2), 145-156.
- Kertesz, A. (1982). The Western Aphasia Battery. New York, NY: Grune & Stratton.
- Rogalski, E., Cobia, D., Harrison, T. M., Wieneke, C., Thompson, C. K., Weintraub, S., & Mesulam, M. M. (2011). Anatomy of language impairments in primary progressive aphasia. Journal of Neuroscience, 31(9), 3344-3350.
I wanted a specific list of therapy approaches, so here is my prompt:
What are the specific treatments speech-language pathologists can use to enhance communication and compensate for language difficulties in adults with Primary Progressive Aphasia? Be specific and please cite your sources in APA format.
Chat GPT4’s response:
“Speech-language pathologists (SLPs) can use a variety of treatments to enhance communication and compensate for language difficulties in adults with Primary Progressive Aphasia (PPA). Here are some specific interventions:
- Errorless Learning: This intervention involves minimizing errors during treatment to facilitate learning in individuals with PPA. Errorless learning can be used for teaching new vocabulary, improving word retrieval, and practicing functional phrases (Fillingham et al., 2005).
Fillingham, J. K., Hodgson, C., Sage, K., & Lambon Ralph, M. A. (2005). The application of errorless learning to aphasic disorders: A review of theory and practice. Neuropsychological Rehabilitation, 15(3-4), 290-318.
- Semantic Feature Analysis: Semantic Feature Analysis (SFA) is a technique that involves identifying and discussing features of a target word or concept to improve word retrieval (Boyle & Coelho, 1995). SFA can be adapted for use with individuals with PPA to facilitate the retrieval of nouns and verbs.
Boyle, M., & Coelho, C. A. (1995). Application of semantic feature analysis as a treatment for aphasic dysnomia. American Journal of Speech-Language Pathology, 4(4), 94-98.
- Script Training: Script training involves teaching individuals with PPA to use structured, predictable sequences of language to communicate in specific contexts. This method can help improve their ability to engage in conversation and express their needs effectively (Youmans et al., 2005).
Youmans, G., Youmans, S. R., & Hancock, A. B. (2005). Script training treatment for adults with apraxia of speech. American Journal of Speech-Language Pathology, 14(4), 295-305.
- Melodic Intonation Therapy (MIT): MIT uses the musical elements of speech, such as pitch and rhythm, to improve expressive language in individuals with PPA. By emphasizing the melody of speech, MIT can facilitate word retrieval and speech production (Norton et al., 2009).
Norton, A., Zipse, L., Marchina, S., & Schlaug, G. (2009). Melodic intonation therapy: Shared insights on how it is done and why it might help. Annals of the New York Academy of Sciences, 1169(1), 431-436.
- Augmentative and Alternative Communication (AAC) devices: SLPs can introduce AAC devices, including communication books or boards and electronic devices, to help individuals with PPA convey their messages and enhance their ability to communicate (Beukelman et al., 2011).
“Beukelman, D. R., Garrett, K. L., & Yorkston, K. M. (2011). Augmentative communication strategies for adults with acute or chronic medical conditions. Baltimore, MD: Paul H. Brookes Publishing Co.
“Remember that treatment approaches should be individualized and tailored to the specific needs of the person with PPA. Collaborating with the individual, their family members, and other healthcare professionals can help optimize the intervention plan.”
As you can see, Chat GPT3.5 and GPT4 listed different therapy approaches.
Chat GPT may be useful to SLPs but should be used with caution
In summary, I think Chat GPT will prove to be increasingly useful to SLPs as the models are improved. As it stands now, I think the model is most useful for develop materials for therapy and home exercise programs.
I also think Chat GPT is helpful for brainstorming ideas, and it has a strong reputation for summarizing individual documents.
Chat GPT does have serious drawbacks, which will hopefully be overcome in the future. The two largest flaws currently is that it doesn’t have access to the internet past September 2021, and it can generate false information that looks legitimate.
Regardless of how well the model works, SLPs should always gather information and make decisions in an ethical and responsible way. If SLPs chose to use Chat GPT, they should never violate patient’s privacy. They should also fact-check Chat GPT’s responses, and not rely on the model for diagnosing or treating patients.
Despite these drawbacks, I think Chat GPT can be a useful tool to SLPs now, if we use it with caution.
What do you think? Please share your comments below, or feel free to contact me.
Related Eat, Speak, & Think posts
- 8 strategies for easier reassessments.
- 5 ways I improved work-life balance as a home health SLP.
- 7 tips to benefit your home health patients and minimize your workload.
Free DIRECT download: Reading passage and questions generated by Chat GPT4 (patient handout). (Email subscribers get free access to all the resources in the Free Subscription Library.)
Featured image by geralt from pixabay.
Lisa earned her M.A. in Speech-Language Pathology from the University of Maryland, College Park and her M.A. in Linguistics from the University of California, San Diego.
She participated in research studies with the National Institute on Deafness and other Communication Disorders (NIDCD) and the University of Maryland in the areas of aphasia, Parkinson’s Disease, epilepsy, and fluency disorders.
Lisa has been working as a medical speech-language pathologist since 2008. She has a strong passion for evidence-based assessment and therapy, having earned five ASHA Awards for Professional Participation in Continuing Education.
She launched EatSpeakThink.com in June 2018 to help other clinicians be more successful working in home health, as well as to provide strategies and resources to people living with problems eating, speaking, or thinking.
Thanks for such a thorough deep dive into the application of Chat GPT! Really interesting read. I have been playing around with it for generating treatment stimuli and interestingly I have found that it (at least GPT3) is terrible at knowing what a syllable is. When I asked for a list of 6 syllable words it included anywhere from 4-6 syllables!
Oh, interesting. I posed this problem to Chat GPT 3.5:
I write a blog for speech-language pathologists. I suggested Chat GPT is a great tool for creating treatment stimuli. One of my readers commented that in her experience, Chat GPT makes a lot of mistakes when asked to make word lists of 6 syllables, so I wanted to try it. Is there a prompt we can use to make sure we do get lists of words that only contain 6 syllables to use for therapy?
Chat GPT 3.5’s response:
Certainly! To generate a list of words with 6 syllables specifically, you can use the following prompt:
“Please generate a list of words, each containing exactly 6 syllables. These words will be used as treatment stimuli for speech-language pathology therapy.”
By providing the specific requirement of 6 syllables in the prompt, you can increase the chances of getting a list of words that meet that criteria. Additionally, you can mention that you’re seeking accurate and suitable words for therapy purposes.
Feel free to use this prompt, and I hope it helps you generate the word list you need for your blog and speech-language pathology therapy.