Provider Pulse Ep. 24: From Siberia to Harborview - Yuliya Speroff and the Art of Medical Interpreting (Part 2)

In Part 2 of our conversation with Yuliya Speroff, Medical Interpreter Supervisor at Harborview Medical Center, we further explore the profound human work of medical interpretation. Yuliya pulls back the curtain on what interpreters witness every day—the cultural gaps, missed cues, unintended misunderstandings, and the invisible labor that makes cross-linguistic care possible.

We discuss the delicate balance interpreters in maintaining objectivity and accuracy while also acting as cultural brokers and advocates when necessary. We also explore how interpreters are uniquely positioned to provide insights into the patient experience. 

Yuliya shares memorable stories illustrating when meaning is lost, how family members unintentionally censor information, and why even well-intended provider phrasing—like “hit it out of the park” or “we’ll keep you comfortable”—can fail across cultures.

The episode also examines some of the risks and promises of AI, ambient note-taking, and machine translation in clinical encounters. Yuliya outlines why even tiny translation errors can have catastrophic consequences, and why interpreters’ expertise is essential in validating new technologies before they become standard in patient care.

Other themes you’ll hear:

• Why idioms and cultural references are so difficult to interpret  

• How power dynamics prevent interpreters from giving provider feedback  

• What trainees and future physicians need to know about working with interpreters  

• Why increasing diversity among clinicians improves communication for all patients  

• Why human interpreters remain irreplaceable, even in an era of rapid AI advancement  

You can find the Greek medical interpreter skits on the TikTok account of @yiannispac

Through humor, candor, and deep expertise, Yuliya shows that interpretation is not just word-for-word exchange between languages—it’s a practice of empathy, cultural insight, and relational care.

This episode is a must-listen for anyone who works with interpreters, cares for diverse patients, or is curious about the future of cross-cultural communication in healthcare.

Visit EthnoMed.org for additional resources. Follow us on YouTube and Instagram @EthnoMedUW

Chapters

• 0:00 Yuliya Opening Quote: When "Big Picture" Does Not Translate
• 0:41 Introduction, Recap, and Episode Overview
• 1:17 Yuliya Speroff, Interpreter vs Translator
• 2:24 Cultural Brokering: the fine line between accuracy and offering context
• 4:23 Advocacy in Medical Interpretation
• 7:27 When Meaning Does Not Translate: how interpreters process the many misunderstandings they witness
• 11:56 Interpreters as Students of Human Interaction
• 17:12 Family Members as Interpreters, an Anecdote
• 19:52 Teaching the Clinician: Idioms and Cultural References, Diversity in Medicine Improves Communication
• 22:43 AI and Ambient Note Technology
• 32:39 The Black Box Problem of Ambient AI
• 36:41 The Future of Education and Advocacy
• 38:29 Closing Reflections: Why Interpretation is a Form of Care
• 41:50 Outro

Transcript

Yuliya 10/24/2025

Provider Pulse Episode 24: From Siberia to Harborview (Part 2)

[00:00:00] 

Yuliya: So there was one conversation I remember interpreting the provider said to the patient, you know, you are undergoing this cancer treatment, but I think it's time to look at the big picture. There's a team that can help you look at that, And the patient had no idea what was being offered.

 they were trying to be kind, they were trying to avoid saying these difficult and scary things. And so they talked about the big picture and making you comfortable, which again sounds nice.

Make you comfortable? What does it even mean? 

Duncan: Does it mean that you're going to get them a [00:00:30] better bed, you are bringing them an extra pillow, or are they meant to understand that you're going to stop active treatment and provide pain control and other comfort measures? Right. Welcome to the EthnoMed Podcast. In today's episode of the Provider Pulse Series, we continue our conversation with Yuliya Speroff, Medical interpreter supervisor in the interpreter services department at Harborview Medical Center.

In the first part of our discussion, Yuliya shared her journey from growing up in Siberia to discovering [00:01:00] her passion for language and finding a professional home at Harborview

in the second part, we turn to the nuances of the medical interpreter's role and hear Yuliya's advice for providers on how to become more effective partners in cross-cultural communication.

Yuliya: My name is Yuliya Speroff I'm a medical interpreter supervisor at the Interpreter Services Department at Harborview Medical Center, and I'm also a certified Russian interpreter. Medical interpreter is a [00:01:30] very unique role in that most people don't know that this job exists. Before I came to the US I didn't. It's not that I didn't know that this job existed, I didn't really think about it much. maybe if they were born here, they don't really think about this, which I think is, you know, very normal. Right? and even some other professionals who maybe have worked with us may know us under a different name, the wrong name. So translator, you know, I'm sure you know the difference.

Interpreters help with spoken or signed communication, whereas translators work on written [00:02:00] document translation. So a translator might be a person who translates clinical trial documentation or a book of poetry, right? They may or may not also be an interpreter, and vice versa, an interpreter may or may not be able to do document translation.

So different jobs. But, it is a unique profession and I came to it in a very roundabout way, and it's a little unorthodox.

CHAPTER 1: CULTURAL BROKERING

Duncan: We start our conversation by looking at the delicate space between accuracy and context. The moments when [00:02:30] interpreters bridge, not just language, but culture itself through a process known as cultural brokering. 

Yuliya: Some of the things that we learn in our training is that it's not our job to manage speakers. Our job is to try and replicate a situation where two people speak the same language and potentially share the same cultural background. So we're trying to level the playing field, so it's not necessarily our job to tell a doctor how to do their job. It's not our job to tell a [00:03:00] patient how to be a good patient, or how to advocate for themselves. Depending on the code of ethics that you follow, there might be room for things we call cultural brokering. This is where you feel like there is a disconnect based on cultural differences.

You might nudge speakers in that direction. And we never speak in absolutes. Like a very common example that we think of is Ramadan, which is fasting, where not only do you not consume food and drink, but often you don't consume [00:03:30] anything by mouth from sunrise till sundown. And that includes medications often or things like smoking.

And so, let's say you prescribed a medication to a patient and you said you must absolutely must take it on schedule, or really bad things will happen. And I think, oh, well this patient is from a country where they're likely to be following Ramadan. But, I don't know for sure. I also don't know for sure whether they're planning to adhere to strict fasting or whether they'll make exception for medications, which can [00:04:00] also happen.

And so I may just bring this up and say, you know, this is the interpreter doctor, Ramadan starts, this is what it means. Would you be interested in asking the patient about that. And then I might say to the patient, you know, this is the interpreter, so I just told the doctor about Ramadan. I'm not sure if that's relevant, but maybe it's a good discussion to have.

Right. So I'm not speaking in absolutes. I'm hedging and bringing this up.

CHAPTER 2: ADVOCACY IN ACTION

Duncan: Yuliya discusses the role of advocacy in medical interpretation.

Yuliya: Another part of the interpreter's role, [00:04:30] which is called advocacy, that's often misunderstood. Advocacy is interpreters speaking in their voice to prevent or maybe potentially right a wrong. A lot of what people traditionally think of advocacy is just sharing information.

So let's say I'm with a patient, I've been interpreting for them throughout the day. I know that they have an IV line placed in their arm and it's there. And then, they go to an office visit and the medical assistant says, Hey, so let's take your vitals. Let's check your [00:05:00] blood pressure Is your right arm okay? And the patient goes, sure. And I'm like, oh, wait, wait, wait a minute. I was just with them. They just got an IV put in there. And I interpreted the nurse saying, don't let anyone do anything on your right arm. And so maybe the patient forgot, right? Maybe they forgot these instructions, maybe they forgot that it was on their right arm, right?

And so I might speak up as an interpreter and say both to the patient and the nurse, you know, this is the interpreter. This is, what just occurred. Would you like to talk about this? So this would not be advocacy. [00:05:30] However, if the nurse said, oh, you know, it's fine. We'll, keep doing it then, you know, I would have to raise the alarm and say, this is the interpreter, please stop.

You know, let's, go get somebody else here because I don't think it's safe. So there are ways to appropriately intervene as an interpreter, but most of the time it is not only that you cannot, but it is not your job. So if the patient and the provider are angry at each other. Your job is to interpret all of the nuances. If somebody's using bad language, your job [00:06:00] is to make sure you're interpreting that bad language, which a lot of trainee interpreters balk at. They're like, what do you mean? I have to use the F word that seems so rude and just unusual in a professional environment.

But if you think about it, if a patient is swearing, there are a couple of things to consider. There is communicative autonomy, which is the idea that everyone has the right to say what they want, when they want, how they want, right? Like you, Dr. Reid, nobody probably follows you around in your daily life telling you, [00:06:30] oh, that's rude.

Don't say that. Oh, that's silly. You know, don't talk about that. And then

Duncan: They probably should.

Yuliya: They probably should. Yeah. Well, sometimes we have spouses that fortunately do that for us. And then there is an idea that everything can have clinical significance. So maybe if somebody is swearing, what does it mean?

Are they in so much pain that that's the only way they can think of to convey that? Is it that maybe they're intoxicated and that's something that's not typical for them? Is it that they got hit on the head and that's [00:07:00] also, you know, not typical for them? Finally, is it the extent to which they're angry, in which case it's a safety concern.

Because if a patient says, if you don't do what I effing want, I'm effing going to F you up, and I say, oh, the patient is a little upset and they're using bad language, you don't realize the extent to which the situation is escalating, right? That's a huge safety concern for you, and I'm doing you a disservice by quote unquote protecting you.

CHAPTER 3: WHEN MEANING DOESN'T TRANSLATE

Duncan: I ask Yulia about how she deals with interpretation [00:07:30] for visits, where she notes that there are problems with communication, but she's unable to intervene.

So what I'm hearing from you is that there are a lot of nuances in terms of interpretation. Where it goes into these areas of advocacy, pointing things out. 

There must be so many interactions that you see where things could have been done better from both sides. What do you do with all of that? 'cause you, you are a student of human interaction And you understand the patient's perspective probably better than anyone else, and the physician's perspective, [00:08:00] and you have been seeing these interactions some of which go well and some of which don't. Where does all that frustration where you're not allowed to cross that line, where does that go?

Yuliya: You know, this is a really good question. Some of the things that happen in an encounter that doesn't go well, some of these things, there's nothing you can do about, you can just learn to navigate them. One common example I'll give you. I just reposted on LinkedIn, a video by a physician assistant from New York, and he had a funny skit where it talks about you are [00:08:30] provider, you ask the patient a simple question and the patient goes, oh, the first time I, tried seafood, this is what happened.

And they go on for two minutes and the doctor goes, well, I was asking you if you have any allergies. And so that will happen all of the time, whether the interpreter is present or not, where you ask a very straightforward question and somebody tells their whole life story. So as an interpreter, what do you do?

You interpret that life story word for word, right? So some of these things you just navigate. Some other things I can think of is, let's talk [00:09:00] about providers language. So one thing that can happen is really complex, very high level language, which again, might happen with English speaking patients too.

I think US doctors have a tremendous amount of training and you may not remember that not all of us do. And so that's just the way you speak. I often actually find that trainee doctors maybe because it's so fresh and they've just learned it, they will use all of that language with patients.

Again, not realizing that's not how people speak. So that is something I can [00:09:30] address, but when I train staff and providers and remind them, please be mindful of how you speak. Again, just because somebody doesn't speak English doesn't mean they're not going to have health literacy.

They might be extremely educated. As a patient, they might be very knowledgeable about their condition, more knowledgeable than you. You know, I'll never forget when I was interpreting and there was a term I didn't know and I said, this is the interpreter. I'm not familiar with the term ALL.

And the patient turned to me and said, oh, acute lymphocytic [00:10:00] leukemia, duh. And it wasn't quite as exaggerated, but you know that's a good example of that. Another extreme is being very vague. when doctors are trying to be kind.

So there was one conversation I remember interpreting the provider said to the patient, you know, you are undergoing this cancer treatment, but I think it's time to look at the big picture. There's a team that can help you look at that, would you like to meet with them? And the patient had no idea what was being offered.

 Now, I had interpreted for them before and I had overheard [00:10:30] some conversations, which I understood that the patient was not responding to the treatment and the team was questioning whether they should proceed with the treatment since it wasn't benefiting the patient or whether they should consider, kind of entering palliative care, end of life care, and ceasing the curative treatment.

But that's not how it was communicated. You know, they were trying to be kind, they were trying to avoid saying these difficult and scary things. And so they talked about the big picture and making you [00:11:00] comfortable, which again sounds nice. I dunno if you've ever seen TV show the scrubs. There was a scene where, a young resident, said, oh, why do we always say that?

Make you comfortable? What does it even mean? And then in his head, he imagines a box of kittens that you bring to the patient and you turn it over and they're all of these kittens, and he was like, that's what it means to be comfortable. Right. And so when you say like, oh, we'll make you comfortable, what does it mean to the patient?

Does it mean that you're going to get them a better bed, you are bringing them an extra pillow, [00:11:30] or are they meant to understand that you're going to stop active treatment and provide pain control and other comfort measures? Right. And so you have to really think about the patient's understanding of the concepts that may not exist in other countries.

Does palliative care exist in other countries? What is hospice? Is there such a thing as hospice services provided at a home, or is it a place rather than a service?

CHAPTER 4: STUDENTS OF HUMAN INTERACTIONInterpreters as students of human interaction

Duncan: what I'm increasingly realizing is that all of these interpreters [00:12:00] are students of human interaction, not just students, but they're become experts of human interaction. What is being understood and what's not. And you are bearing witness to conversations that are very effective and ones that are very ineffective. And I can only imagine how frustrating it is to see these ineffective ones. And my guess is most providers are like me, remain blissfully ignorant about their own shortcomings in using interpreters. And I'll go back to when I started at International Medicine Clinic.

I did one [00:12:30] year of hospitalist, but I was outta residency and I asked the interpreters, how was it? You know, how did the visit go? And they said, that's fine. So I could never get any feedback.

And I said, you know, I want to get better, what can I do? And they're like, oh no, it's good. And then I talked to the other providers and they said, you know, there's such a power differential that you're never gonna get any feedback from there. But then when I'm talking to you, it sounds like a very lost opportunity.

Because there's this wealth of knowledge of interactions of what works and what doesn't that the providers are not aware of. And I wonder if there's a way to even more [00:13:00] systematically go about it. 

The lack of feedback from interpreters to providers, in part due to power differential

Yuliya: That's a good question and, okay, Dr. Reid, first of all, I have interpreted for you and you're a good partner. You pause appropriately to allow me to interpret accurately and completely. You use fairly straightforward language, so you're a joy to interpret for. To answer your other question, honestly, I don't know.

I have talked to you about before the sheer amount of effort I put into getting to trainee doctors. every year I go on the website of UW Medicine, [00:13:30] I look up all of the programs that have incoming residents or interns or fellows, and I email them one by one and I say, Hey , i'm Yuliya Speroff. I'm from Harborview Interpreter Services. You have incoming trainee doctors. May I please talk to them? May I please take 15 minutes or 30 minutes or 60 minutes of the time in your orientation? 

I will say there are some programs that we, built a relationship with that they will proactively reach out. they build me into their schedule, so they [00:14:00] will always invite me, but a lot of the time it's me and sometimes I don't get a response.

 Can I get to them even earlier, like right now, somebody reached out to me they are working for a post-bac program where they prep students to get into medical school, and so they're going to have me as a guest speaker for their, students early next year.

Even though it's going to be like on a Saturday morning, I'm so happy to do it because I'm getting to students even before they get training to get them to start thinking about it. [00:14:30] Because I will say from the trainings I have been doing, there are a lot of future doctors with lived experiences of having to interpret for their loved ones, which never should happen. It's on the hospital, it's on the clinic to provide professional interpreters. We never want the family members or particularly kids to carry that burden. They have that lived experience, they have that understanding. And so when I tell them, oh, you know, inability to speak English is a social determinant of health, they're like, yeah, we know.

Tell [00:15:00] us the phone numbers for your department. We know your department is important. We want to partner with you. Just tell us how, tell us, what number to call. Tell us what email to send translations to. That's what we care about because we know the rest. So that is one thing that is very heartening for me to see.

But on the other hand, again, I know that when it's something that we haven't experienced, that we haven't faced, it's really hard to imagine it. And so it's just not on folks' radars that Oh, you might have a patient at some point, let's say [00:15:30] even as a doctor, maybe you speak another language or a couple of languages.

Well, you're also going to see patients who speak other languages too, right? So how are you going to communicate? What role are you going to play in making sure you understand your patient and they understand you?

Duncan: And I think the outreach you've been doing is huge and I think that's, a great first step for a lot of them. And it sounds like you've identified the people that get it, and I think that's another reason to increase the diversity of medical providers because that lived [00:16:00] experience, they understand. But I guess my question is, is just the lack of feedback and I think it's the hierarchy. And the power structure, because I sit a lot with the, residents that are in international medicine clinic, and when I see them using the interpreter, there's a lot of small things that can be improved.

But I think it just requires that longitudinal feedback. So I think we could even brainstorm. ' 

I'd be interested in other novel ways of doing it. You said some nice things about my working with interpreters, but there's definitely [00:16:30] things I can do better. What is a way that we can sort that out and figure that out and illustrate it in interesting ways, and I think you have so many interesting anecdotes.

 And then there's such a richness of experience with the interpreters that I think we can take advantage of. And it might be empowering as well. 

Yuliya: I, you know, I love that idea about feedback, but you are right. I'm a supervisor, right? Part of my job is from time to time to reach out to providers and say, Hey, you know, this is something that was reported to me. Let's talk about this. Let's talk about [00:17:00] why this is not the best practice.

So one common example is, you know, family members interpreting, right? you have a professional in-person, interpreter assigned to you. Please work with that interpreter rather than with a family member.

Duncan: Yuliya turns to a personal memory, interpreting for her own parents, to show why family members often struggle in that role. When her parents met her fiance's family, the children tried to bridge the language gap themselves revealing just how easily family members can slide into the role of censorship.

Yuliya: My parents [00:17:30] they do not speak English. both of them I think learned some German in school. Long forgotten. And this really came into play when my husband and I were getting married. So my husband is from the U.S. He speaks some Russian, so he was able to communicate with them. But when we were getting married and his whole family came to Russia for the wedding, they were really relying on us to interpret for them.

And this is an example why family members should not interpret. So our parents have fairly similar backgrounds. They desperately wanted to [00:18:00] talk to each other and get to know each other.

Their children are getting married, they're now family too. And I remember having this conversation around the dinner table and my mom was like, you know, can you tell them that I don't like long flights? And ask them if they have pickles in the US. And I was like, mom, of course they have pickles in the US and the long flight part of the conversation is over.

It's, it's too late now. And then I hear my husband across the table having a very similar conversation where his mom was like, ask them if strawberries grow in Siberia. He's like, of course they do. [00:18:30] What a silly question. I'm not going to interpret that. And so we were just censoring them and they were looking at each other and just like smiling.

They're like. I want to talk to you, but I can't because my family member is censoring me. And I think that happens a lot in healthcare, right? Like, you're asking too many questions, let's just get a prescription and go, you know, I'm not going to interpret all of that. Or, you know, these are bad news. I don't want you to be upset, so I'm not going to interpret that.

So, they don't speak English. I wasn't the best interpreter for them when they wanted to talk to their [00:19:00] in-laws.

However, you're right about the power differential, you know? I know you are responsive to feedback.

You are actively seeking feedback. Is everyone, is everyone open to somebody else telling them, you know, these are some of the things that, I would really appreciate, would make my job as an interpreter easier. You know, it would be great if you would remember to address the patient directly so that you don't say interpreter tell him, I'm going to examine him. Ask him when the pain started. You know, it's much easier for me to [00:19:30] interpret if you just said, Hey, you know, I'm going to examine you. When did your pain start?

Duncan: when you look at these interpreters, it sounds like you didn't even necessarily know what you were signing up for because you're signing up for actually being a student of the human condition.

And of human interaction, you're signing up for being an interpreter of body language, of nuance of all of these things, whether you want to or not. 

CHAPTER 5: TEACHING THE CLINICIAN

Duncan: Yulia describes another tricky area for interpretation, the use of cultural references.

Yuliya: And that's something that I do interpret a training in [00:20:00] too. How to interpret cultural references, how to understand them. And I don't think it's realistic to tell providers or patients when you speak, don't use cultural references.

Don't use idioms because it's so baked in. Yesterday I was interpreting and in my head I was counting how many idiomatic expressions the provider used and they were so natural they were such big part of that conversation. I don't think the provider was conscious of it or was thinking, Ooh, I [00:20:30] have an interpreter.

Let's make her cry by using 15 idioms in the next five minutes.

The issue with translating idioms, role of code-switching

Duncan: Yulia recalls moments of confusion when American sports and cultural references slipped into conversation-- reminders of how language carries culture in subtle ways.

Yuliya: When I first moved to the us, I don't know American sports. So when a provider says, Hey, we got your test results and you hit it out of the park, what does that mean? Is it good or bad?

Is hitting out of the park? Good. To me, it sounds bad. It should be in the park. Why are [00:21:00] we hitting things out of the park? It's not good, but in reality, what the provider is saying, your test results are good or maybe better than expected. And if I don't understand that, I will not interpret it correctly. But again, is the provider trying to be opaque?

No, in fact, they're happy for their patient and they're saying, you know, I'm going to emphasize how good the results are by using this colorful expression to show that I'm happy for you as the patient.

Duncan: But it could be [00:21:30] useful to be aware when you're using idioms, particularly ones that are related to a sport that is only practiced in and parts of Asia.

Yuliya: I, yes, ideally. But how realistic is that when you are speaking, you know, how cognizant are you of every phrase you're using?

Duncan: Well, going back to if you grew up in a bicultural background. Very right. Like I don't talk to my mom about touching base or hitting outta the park, right? she's not from this country. So I don't use any of those cultural idioms.

Right. Do you talk about those idioms to [00:22:00] your parents? Exactly. So then you already do that. So I don't think it's a huge ask, honestly. And I think that's another reason you need to get that diversity in providers, because I talk to these providers and I'm like, whoa, the, the language you're using is a little bit difficult to follow.

Would you talk to your mom like that? And they're like, no. So then I was like, just use the language that you would talk to your mom. Don't use these idioms. And then they start speaking in a way that's much more easily interpretable 

Yuliya: That makes so much sense. I'm going to use that example from [00:22:30] now on because right. You know, if you told your mom, if you said, Hey, I'm sending you to the surgeon, but he's the Michael Jordan of surgeons,

Duncan: She'd be like, she's like, why do you talk to me like that?

Yuliya: Who's Michael Jordan? Right? 

CHAPTER 6: AI AND AMBIENT NOTE TECHNOLOG Y

Duncan: One thing, can we wrap up talking about this artificial intelligence, AI is a thing that is advancing quickly, but everything we talked about today was the richness of, interpretation.

Interpreters as humans that have inner feelings, that can receive [00:23:00] trauma, that have knowledge of idioms, of culture, all of these things. And now when we talk about AI, we're talking about breaking it down to units of meaning. What are your thoughts just broadly, at any level, about AI machine translation and where things are going in terms of in-person interpreters?

Discussion about AI

Yuliya: Yeah, that's a good question, a slightly scary question. AI, as of today, in October of 2025, and things may well be different even in a few months. But as of now, AI is a [00:23:30] tool. It is a, useful tool. But let's think about this from the medical perspective. AI is used in medicine for example, to look at images and detect, breast cancer, maybe signs of a stroke. When it does that, a human has to validate it correct, so they have to look and say, yes, their finding is correct, and then a human has to convey this news to the patient, explain to the patient what it means.

And take it from there. Right? Go over treatment options. So it's a [00:24:00] tool. It's not replacing a doctor where it diagnoses a patient, where it tells the patient the news and says they're there. Do not cry. We, it will be okay. Although I don't know why I'm exaggerating that nowadays the voices can get pretty good.

 And so that's how I think about it in terms of interpreting and translation. It is a tool. I use it to do written document translations. I use it to prepare for interpreting assignments, when I think, oh, I'm going to interpret in rheumatology, I haven't been to a rheumatology [00:24:30] appointment in a while.

Can you give me 10 examples of common disorders? Can you give me 10 common tests, 10 common medications? Give me some questions a doctor is likely to ask, and then I'm going to look up how to say all of that in Russian and I'm ready to go. So it's a helpful tool. Now, in the same way that it cannot replace a doctor, I do not believe it can replace an interpreter.

First and foremost, it's just not there yet. In terms of capabilities, there are so many examples. Let's think Spanish. [00:25:00] Spanish is a language that worldwide is very prevalent. So there are a lot of materials in Spanish so AI can train on. So it's getting better. But even with Spanish, I was giving this training to some residents and I was talking about how Google translate and other machine translation is just not there and it's against hospital policy to use it without verifying it.

 And an attending physician was sitting there and they said, you know. I just tested it in Spanish and I speak Spanish, so I'm [00:25:30] able to validate it. And there's a very common instruction that's given by doctors and other providers that says, hold this medication. Hold this medication until the surgery, hold this medication until you see your primary care provider, meaning pause, do not take it until whatever else happens.

So in Spanish it translated it as kind of maintain like hold. Yeah, keep doing it. Which is exactly the opposite. And it can have catastrophic consequences. It's just one word. [00:26:00] It might be 99% accurate. What's in that 1%? Is it something silly like an idiom that doesn't make sense? Or is it a catastrophic error?

One word mistranslation causes, somebody to bleed out during surgery because they didn't stop taking their blood thinners. And that's for Spanish. What about all the other languages? Where there's less written materials. You know, we ran this experiment and we found so many [00:26:30] inconsistencies and inaccuracies.

One example that comes to mind, it was supposed to say, keep taking your antibiotics. In Tigrinya it said keep taking your anti soul medication. I would be scared and if I were a patient, you know, I would not be taking my anti soul medications. And another example I give to providers, you can absolutely use these modern tools like AI, but be equitable.

 Think about the standards you apply to your English speaking patients. Let's say you're using [00:27:00] some sort of AI tool to help you write discharge instructions or some other instructions. You would say, please write up a regimen for whatever, but then you are going to verify to make sure it didn't say a thousand milligrams instead of a hundred, that it didn't hallucinate something.

You have to do this. This is your legal liability, right? The liability rests with you. And so when you are doing this for, let's say an Amharic or Somali speaking patient, when you say, these are my discharge instructions, translate them into [00:27:30] Amharic. You don't have anyone checking the Amharic or Somali or Spanish or Tigrinya to verify that it didn't hallucinate anything, right?

And you are not willing to do that for an English speaking patient and to assume that legal responsibility. So you shouldn't be willing to do that for translations as well.

Duncan: So you're taking this argument saying, well, it's not perfect and it makes errors, and the errors can be catastrophic, even if it's a very small percentage, because I think that's the argument they're gonna make is like, oh, this technology is getting better and better. But then I think your counter [00:28:00] argument is that, well, even if it's getting quite good, if it's making one critical error in a word that could be catastrophic.

And so that's why you still need human, verification. Then the other thing is that these large language models are probably never going to capture all of the nuance that you're asking an in-person interpreter to do. Or a language interpreter over the phone in terms of translating idioms, understanding cultural context, actually becoming a advocate.

Yuliya: Absolutely. I think you know [00:28:30] all of the examples I gave you of the interpreter perceiving a misunderstanding, perceiving what it means for a patient to say, I will try. Maybe in some cultures you don't directly contradict a person in the position of power. And as a physician, as a clinician, you are in the position of power.

And so you will say, I will try. But what it really means is I'm not going to do anything that you suggested because I don't think it'll be helpful for me. And so [00:29:00] a skilled interpreter might at that moment say, this is something that you may want to explore more, and ask more questions about.

Duncan: And I think this is where the impasse is occurring because I think the people that believe that AI is going to take over and replace interpreters and translators are the ones that don't understand the nuance of interpretation. And I think part of it is because when these communications are not done well, there's no forum for us to show exactly what happened.

So I think that can be a big advocacy thing that we do. 

Yuliya: [00:29:30] Absolutely, because you might believe that you just had a good conversation and you walk away Going back to my early example, you talked about the big picture. You talked about keeping yourself comfortable. The patient said, great, I love being kept comfortable, and you walk away thinking and you chart, discuss palliative care with the patient.

Patient agreed. Meanwhile, the patient had no idea what she talked about and they said yes to, you know, to be 

Duncan: So 

Yuliya: you. So,

Duncan: what is the mechanism that we can show, right? Because otherwise the [00:30:00] physician is like, oh, I did a great job.

Yuliya: yeah, absolutely. So I would hope that you have a skilled interpreter with you, whether it's in person or over the video, over the phone. And I will say for phone and video interpreters, it's not that they're any less professional or trained or skilled, is that there's often like a physical barrier.

Because when I'm there in person, I can follow you out of the room and I can say, Dr. Reid, you know, do you have a minute? I wonder if you thought about the following, would you be [00:30:30] interested in, maybe not now, but you know, discussing this with the patient again, this is my concern and I may be wrong, but this is what you know I'm concerned about.

Duncan: Or just body language too. I mean, everyone's been on Zoom meetings and then you meet in person and it's a quite different experience, isn't it? Absolutely. Meeting Zoom. And it's the same idea that a telephonic interpreter, they just can't pick up on the cues, you can't pick up on the cues from them.

Whereas when they're in person, they can see your hands, they can see the patient pointing at you. Yeah.

Yuliya: Yeah, [00:31:00] absolutely. Yes. definitely some elements of nonverbal communication will be lost with the phone. And that's where, again, this is a tricky thing. Usually what I say as, an ambassador of my department is that an interpreter is better than not interpreter. Having an in-person, especially staff interpreter, is invaluable because it's not only an experienced and professional interpreter, but they have institutional knowledge.

They know how Harborview works. Maybe [00:31:30] they have worked with you many times, so they kind of know how you work as well. So that's always going to be kind of the golden standard. Even better, a provider who speaks your language and who understands your culture, but at the end of the day, you have to communicate with someone when you don't speak the same language.

Telephonic interpreting is available 24 7. It has some very skilled and professional interpreters. Actually, I have had experiences when I called interpreters and they were my students that I trained, so I know [00:32:00] they're good. I trained them. Same with video, right? we have the visual cues now.

Even though as an interpreter, you're at the mercy of providers, if they turned you to face the wall, you know, you could ask them to change it back. But they may or may not do that. So you are still kind of present in the room. And again, if this language is not available in person or we just don't have anyone right now because it's the middle of the night or because all 10 of our Spanish interpreters are assisting all [00:32:30] these other priority appointments.

Right. that's great. Use whatever you have, whether it's phone, whether it's video, whatever is approved. Right. Not Google Translate 

Duncan: So now we have this ambient AI software that records a visit it automatically transcribes in 25 different languages and makes a summary in English without any interpreter verification. What is the feeling in the interpreter community about this technology?

Yuliya: Yeah, that's a good question because if we think [00:33:00] about everything I just shared about the ability of machine translation to handle not only Spanish, but all these other languages. It's a bit of a black box, right? Because we don't see the transcript in the language.

Let's say it's a Russian, we don't see the resulting conversion from Russian into English. We only see that summary. So if an error does happen, it's hard to trace it. Did it happen at the transcription stage? Did the Russian get transcripted [00:33:30] inaccurately, or did it get converted inaccurately, or did it get summarized inaccurately?

So to me, it sounds like there are multiple opportunities for inaccuracies and as an interpreter, I don't really have an opportunity to examine that. And I know that that's something that a provider is verifying and validating, but they too don't see inside of that black box, 

Duncan: And I think the justification that they're using is that these are interpreted visits. So this ambient [00:34:00] AI is recording both in the language and the interpretation, but then it's impossible to figure out what ends up in the final note, and that there might be a combination.

And what I'm hearing from you and my guess is it's probably from a lot of people in the interpretation community, is that it would be beneficial to have interpreters verify the transcription technology and possibly the summaries.

Yuliya: You know, I think, that when something is not part of your experience, it's not part of your knowledge, you don't really [00:34:30] think about this. So I think it's a tool like any other tool can be incredibly helpful. But, when you and I talked about this, you call this edge cases, right?

What if it works for the majority of English speaking individuals, but what about all these other cases? And for all these other cases, there should have been a person in the room who said, wait a minute, what happens if we have this patient? 

Duncan: it sounds like we're well situated with all of these in-person interpreters [00:35:00] that we could potentially validate these technologies that are being used.

Yuliya: I wish there was somebody in the room from the very beginning who said, okay, even, you know, it converts from this other language into English. Do we know that this conversion is accurate? Do we know that it's accurate in all of the languages? Have we checked that or are we hoping that it's accurate?

And I don't know, it could have been the case and they could have done all of these steps that I'm now wanting to do. but yeah, I wish that this was not an [00:35:30] afterthought, but from the beginning stages of the development of the product.

Duncan: But I think there's additional products that are in the pipeline when we think about the AirPods from Apple, I think they have a simultaneous interpretation from Spanish English. So my guess is there's gonna be a big push for these, and it's probably going to be up to us working together to try to validate and figure out whether these are appropriate for edge case scenarios.

Yuliya: Correct. And again, you know, these things sound great for particular [00:36:00] purpose. I just traveled to Serbia, and one of the surprisingly difficult things to figure out were menus. I thought food would be easy, right? As a Russian speaker, there are so many Serbian words I understand, but not menus.

They were indecipherable. So I used Google Translate and their picture translation feature. It worked great. It's still, some of it was nonsense. The stakes weren't terribly high. Worst thing that could happen is that I ordered something and I didn't like it. Worst thing would be maybe I [00:36:30] thought I was ordering lamb and I got, you know, a pork 

Duncan: John? Well, it could be allergy

Yuliya: Well, that too, right? Did again, didn't think about it because I don't have any food allergies. But the stakes could be higher. Yes.

Duncan: then coming back to this AI thing, maybe the issue is these providers are like, I've been providing care to non-English speakers with an interpreter, and I do a great job. And the interpreter might be thinking in their mind, well, I actually do all of this work behind the scenes to make sure that these go well because this provider in particular [00:37:00] doesn't do these things well.

But in their mind they do it perfectly. And that's why they say, well, what is this interpreter actually doing? Right? And how is this machine not doing the same thing? And it's because they're not aware of all of these things. So how do we make providers aware of all the nuance behind this and of what's actually occurring and the edge cases of where our machine learning or machine translation couldn't work and pointing that out.

Because I think we are needing to do education from very beginning. They're taking interpretation for granted because they have been [00:37:30] insulated from any of the feedback.

Yuliya: Yeah, that's an excellent point. And as you're saying this, I'm having thoughts in my mind of doing simulations where you interview a patient with Google Translate and then we analyze all of the things that didn't get understood. I

Duncan: I think. So I think that's how we do it. And if it works perfectly, then it's great.

And if it doesn't, then you say, what is the risk that's acceptable to providers and what are the safeguards? but I think we [00:38:00] need to be the people that are educating about what is the nuance of interpretation that's been invisible to you because you have excellent interpretation staff that has been covering your butt for the past years and you didn't know it.

Yuliya: Yeah.

Well that's the thing though, isn't it? When somebody is skilled at their job, they make it look very easy. I think that's the thing. When somebody does something really well, you're like, it doesn't seem like, it's taking a lot of effort. Must be easy when don't realize that it's years of training 

Yeah. We take a [00:38:30] leaf out of, the physician assistants book who has engaged in tiktoks. I will say too, that's the person I 

Duncan: Mm-hmm. 

Yuliya: He's the only provider that, at least that I have seen, maybe others But he produces this content for other healthcare providers, and he has a skit about a doctor who is the worst nightmare for a medical interpreter.

He says medical interpreter. You don't understand how huge it is that he doesn't say medical translator, but uses the correct term.

And so what [00:39:00] the skit is, you have a doctor who uses incredibly high level vocabulary and he keeps talking and talking and you see the interpreter time to time going, and then the doctor keeps talking.

And of course, in reality, I will say that an interpreter is going to stop you at that point and, and at least try to stop you. Right. And might say, doctor, this is the interpreter. The doctor might, may well be all over

Duncan: By telephonic they might not be able interpret Exactly. Yeah. Interrupt. Yeah.

Yuliya: It's so much harder to interject yourself without being [00:39:30] obnoxious, because in person I can, raise my finger and can raise my eyebrows at you and I can do all these things, which you may or may not heed.

But still. And so he has this skit, right, which shows he understands the work of a medical interpreter. He understands how to partner with one as a provider, and he shows all of the things you're doing wrong. So I think, you know, he can be our inspiration.

Duncan: background? I wonder. 

Yuliya: so, he's a Greek speaker. 

So he's acutely aware of language barriers. He's acutely aware of [00:40:00] cultural differences. He talks about growing up with immigrant parents, 

Duncan: so, 

yeah.

And that's who our target is here is the Minority Association of pre-medical students. Why? Because you're gonna get better care for a greater diversity of people if you have providers that look like the patients. ' cause they'll understand issues of language. They'll understand issues of inequity, issues of racial profiling.

Right. You don't have to take a course on it. They'll be like, I lived through it. I know.

Yuliya: If you're thinking, well, you [00:40:30] know what, I am a person who is born in the us, I'm an English speaker. Why do I care if my doctor has that kind of lived experience and, that background,

 You know, if you have a doctor who is aware of these cultural nuances, aware of these language barriers, they're likely to be a much better communicator in English as well.

that's why we want doctors in medicine with all of these diverse experiences so that they can help everyone. Including, folks who may think, I don't [00:41:00] have any, for lack of better word, special needs, right? I don't have any cultural barriers or language barriers. So I can speak to any doctor, but hey, wouldn't you also benefit from a doctor who is very skilled at communication with everyone?

Duncan: So interpretation isn't just about word for word exchange from Russian to English, and you pointed this out. It's about this whole cultural background that you got from careful study of friends of American cinema And very [00:41:30] similarly, even if you're speaking the same language, it's that cultural background which might differ

from where you grew up, what class you were from, what part of America, or even if you're from a different English speaking place. And that's why that actual cultural context still matters.

Yuliya: Absolutely. Okay.

Duncan: Thank you so much for your time, Yuliya. I.

Yuliya: My pleasure. Thank you for having me on here.

CHAPTER 7: CLOSING

Duncan: When Yuliya talks about interpretation, she's describing more than a job. She's describing a form of care. Each pause, each carefully chosen word, enhances [00:42:00] medical interactions. At Harborview Medical Center, interpreters like Yuliya remind us that every conversation is a chance to build trust and that clear language and accurate interpretation are essential for effective medical care. You've been listening to the EthnoMed podcast provider Pulse series from Harborview Medical Center. Our deepest thanks to Yuliya Speroff for sharing her story and for the work she does every day to build bridges across language and culture.

Thank you for listening to this episode [00:42:30] of the EthnoMed Podcast. If you enjoyed this episode, please share it with a friend. See you at our next episode.

************* END OF PART 2 ***************************************************‌