Using Artificial Intelligence to Demonstrate Communication Disorders:

Two examples of artificial intelligences for medical use reside in the Activity/Daily Living Room. One bot, Brandon, simulates a fourth grade student, while the other, Megan, is a nurse with a severe disfluency. Together, they help participants utilize real world interview questions in a way that helps demonstrate the principles of the S-24 Erickson Scale of Communication Attitudes. What does this actually mean?

The original Erickson Scale presented the patient with a series of true or false questions addressing their concern with how stuttering affected their daily lives. This format was later modified in 1974 by Andres and Cutler. My name is James Zeigler, and I am the project director of the virtual clinic. I added my own modification to the Erickson scale to better serve my class.

The first question on the Modified S-24 Erickson Scale of Communication Attitudes is: "I usually feel that I am making a favorable impression when I talk." The patient would then select the true or false column. Jim changed the syntax and the form of the intransitive or transitive statements into an interrogative form: "When you talk, do you usually feel that you make a favorable impression?" By asking the same question in a variety of ways, the clinician uses the patient's improvised conversational skills to diagnose potential causes or stresses that trigger disfluency.

Now, let's get started.

To engage a particular bot, an avatar touches the AI to receive one of several possible notecards containing randomized questions. The first and second question on all notecards contain the introductory statement, which is used to activate the AI to your commands, and a closing statement, which disengages the AI.

We're going to introduce ourselves to the AI named Brandon by typing or copying the introductory question on the notecard and pasting it into open chat: “I'm here today to know more about you. I might ask you questions that sound like I am repeating myself.” The script is intended to sound conversational and realistic. Brandon answers, “Hello, Avatar Name. Me, too. I like your shirt. That's okay as long as I d-don't have to repeat the fourth grade!”

The d-don't is displayed in chat as the letter “d,” followed by a dash, followed by the word don't. To understand what disfluency a punctuation mark represents, we turn our attention to this chart on the wall. We see that the dash represents an instance of stuttering called a repetition.

The scripted dialogue in a notecard is necessary due to the limitations of the chat bot technology. Without exact working, the bot may reply with silence, or with an answer that has nothing to do with the question being asked. We're also currently exploring ways to give the chat bots voice capability.

Let's ask Brandon another question: “When you're talking to a bunch of kids, do you find it pretty easy to look at your audience?”

Brandon replies, “I d...do I guess.” The d...do is displayed in chat as the letter “d,” followed by three dots, followed by the word do. Using our chart, we see that the dots represent a stuttering symptom known as a block, with each dot equal to one second of blocking. In this case, three seconds.

Now we'll ask Brandon a third question. “Do you dislike introducing your friends to other kids?”

Brandon replies, “I like to have lots of friends and we all hang out together. New kids are fun because you find out cool stuff from them.”

Brandon's reply contains no punctuation marks that would indicate a disfluency. Based on his age, we can presume that Brandon does not stutter, and any occurrences of disfluency at this point are natural and more a cause of his young age. We'll now type our closing statement and end our conversation with Brandon.

Let's see what Megan has to say.

We'll type our introductory statement: “I'm here today to learn more about you and your speech.”

As we can see, Megan answers with, “Hello, Avatar Name. I'm glad t-t-t-to m-m-m-m-meet you. I hope you ca****n help m-m-m-e.” Just based on this initial question, we can see that Megan's disfluency is severe. Using the chart, we see several instances of repetition, as well as four asterisks, which indicate a four second prolongation of the “a” sound.

We'll ask her another question: “Do you usually feel that you are making a favorable impression when you talk?”

Megan replies, “...I don't think people take m-m-m-me seri****ous bec-c-c-cause of my speech. Not a good one, no.” Using our chart, we can see that Megan starts off with an instance of stuttering known as a block that lasts for three seconds before she is able to say I. She then does a repetition of the “m” in me, followed by a four second prolongation of the “i” sound in seriously, and then another repetition of the “c” in because.

Let's ask Megan about stuttering and her job: “Do you feel you talk well enough for the kind of work you really like to do?”

Megan answers, “This ..is why I***** am here. I have had some p-p-patients request another nurse because they think I'm dumb or something.”

As we conclude our interview with Megan, we understand, in her own words, her motivation for seeking treatment. Her answers also help us frame the context of any potential psychological barriers--such as low self-confidence or a fear of public speaking--that may interfere with Megan's ability to overcome her fear of stuttering. We can count the number of stuttering instances in her speech to locate problem words or letters. We can also look at the relationship to problem words with nearby non-problem words, and to see if the point at which a problem word appears in a sentence makes a difference. This simple interview is really quite complex.

Additionally, users can rate the level of severity of the stuttering on a standardized rating scale.