Agentic Patient 12: Early Diagnosis of Ovarian Cancer due to a Chatbot Use

[00:00:00] Dear listeners, welcome to Faces of Digital Health, a podcast about digital health and how healthcare systems around the world adopt technology with me, Tjaša Zajc. In 2026, few people still pretend that patients aren't using AI. They use it to find their way through the healthcare system, to make sense of test results and to prepare for appointments.

[00:00:29] In early 2023, however, which was a few months after ChatGPT was released, that was far from obvious and very few patients mentioned their use of chatbots to their doctors. Alvina Nadim didn't do that either. She is a professional engineer and change management advisor and she uproar ChatGPT the way an engineer tests a tool.

[00:00:55] She asked for possibilities, asked AI to explain its reasoning and kept adding to the same conversation for two weeks as her symptoms kept evolving. Ovarian Cancer came up as one of four explanations. She initially dismissed this idea.

[00:01:16] However, when talking to some family members, she did decide that maybe she should discuss this as a possible option with her clinicians. She used AI to present all the symptoms together in a way that her doctor could act on. This resulted in relevant testing and an early stage diagnosis of ovarian cancer.

[00:01:45] It's important to know that most ovarian cancers are found much later when treatment is far harder. So in this episode of the agentic patient, we talk about how Alvina used AI in 2023, how she sees patients using it now, given that she is a board member of the Quebec Cancer Coalition,

[00:02:10] a patient partner in research for ovarian cancer Canada and a member of the Public Advisory Council of the Health Data Research Network in Canada. Enjoy the show. Enjoy the show. And if you haven't yet, make sure to go to our website to see other episodes in the agentic patient series as well.

[00:02:33] And also check out our newsletter. You can find it at FODH.substack.com. That's FODH.substack.com. The newsletter only goes out once a month or every second month and contains in-depth articles based on the research done through the podcast. If you will enjoy the show, make sure to leave a rating or a review. Drop me a message.

[00:03:00] This is always very useful for me to just get some feedback and to also get ideas for future episodes. Now let's dive in today's discussion. Alvina, hi, and thank you so much for joining me on Faces of Digital Health and a special series called The Agentic Patient,

[00:03:24] where I talk to patients, researchers, clinicians about how AI is impacting their interactions in healthcare, the relationship between patients and doctors, and more importantly, how are patients using AI to help their care and also avoid the dangers that AI can pose.

[00:03:48] You specifically are a change management advisor, leadership coach, and professional engineer, which is not the focus of our discussion because you also have your own patient experience that was significantly impacted about AI. This is just something to keep the audience engaged because I actually want to start with something completely different.

[00:04:14] You recently came up with a new word, and that word is a no-lution. Can you just share what you meant by that and why this is something that struck a nerve with you? Yeah, sure. First of all, thank you for having me. I've been hearing so much about your podcast, and I was like, I have to meet this person, and finally we did. So no-lution, yeah, it's funny.

[00:04:42] It came out of just, I think, a little bit of frustration of, I feel like everybody right now has good intentions, and we're making things and trying to make things better. And I often feel that now in the world we live in, with everything is about clicks and what gets people talking, and I would say impulse to, if you see something, say the complete opposite, or really go against things so that you have these polarized conversations, and because that attracts views and clicks,

[00:05:09] but then I feel like it's just saturating our feeds, and therefore our awareness, because a lot of our awareness and attention is going to our social media feeds nowadays. And I just feel like it's leading to this no-lution idea, where people are just criticizing things and dismissing ideas. Because if someone's worked hard on something, maybe they've missed a few things. We don't always have all the pieces sometimes, but I think that instead of starting with a place of curiosity and wondering,

[00:05:38] hey, I wonder why this person went through trouble of coming up with this idea or the solution, there must be something, there must be a reason. I think one of the simplest examples could be, if we just look at the AI phenomenon in health in general, A lot of people started using AI secretly as soon as it became available.

[00:06:02] So just because the official healthcare system may not have been aware that patients are doing this, and the general public is doing this, doesn't mean it wasn't already part of their healthcare. So I want us to expand our awareness of like, when we say AI in healthcare, healthcare, that should include all of the shadow AI, like the secret AI usage, that started literally as soon as AI became available, right? Because if we ignore that part,

[00:06:32] then we're missing a huge piece of how people are really using it. And so coming back to the no-lution thing is that, whenever I've seen people talk about that, it gets quickly shut down. And I understand where it's coming from. Again, I'm an engineer. I get that. I wouldn't want anyone to go build a bridge and then drive my car on it. I get it. I want someone with the credentials to do it and safety and all that. It's all true. However, we can have that nuance of saying, this is true.

[00:07:00] And let me be curious about why that person did that in the first place. Stop poo-pooing on the ideas and just be more curious. And I feel like that's something that's really been like a pet peeve now to see conversations just being so polarized and not in that middle where there's a place to build something together. Yeah, absolutely. Let's move to your particular experience with AI and your patient situation.

[00:07:28] So can you take me back to, I think, 2023 when the chatbots, ChatGPT was just released and even at that time when it was still hallucinating more than it does today when we knew more about what the safeguards are that we're supposed to use when we potentially were more wary of how we use AI and what do we use it for apart from writing messages?

[00:07:58] How did you use it in your health journey? And how did that impact your patient journey? Yeah, so I always like to preface this with just a short thing on my background because it explains why this even happened in the first place. So I was working, like you said, in change management, so working with companies on helping them with digital transformation, adoption.

[00:08:23] So that exposed me already to just different types of technology and software and things like that. And so when ChatGPT was coming out in November of 2022, it was like a big deal for my work. And as a person, I was just curious about it too. So as soon as it came out, I was, you know, one of the early folks who downloaded it on my phone and started playing with it. And it's funny because at the time there wasn't even an app, but I just created a shortcut on my phone so I could use it like an app.

[00:08:53] But all that to say, I got really used to working with it. And if you remember, actually back in the day, it was closed off to a specific set of data. And I think it only went up until September 2021. So it was actually really super cautious. And if you made it do things that it wasn't comfortable, it would just stop working. It would say, oh, I don't have that. If you asked it a health question, it'd be like, I'm not your doctor. So it was more restricted.

[00:09:19] But I think what it did is that it helped me learn how to prompt in a certain way and things like that and really get into more like role playing. And so anyways, I'm doing this for work. And then, yeah. Let's just stop there for a while. Because you said that you learn how to prompt. Can you be specific? So what did you prompt? How did you change your prompts?

[00:09:46] And then later on in the discussion, we can talk about how you did that today. Sure. Yeah. So I think at first, I was just using it mostly for work. Like you were saying, like emails and ideas and things like developing better documents and things like that was very work-oriented and text-oriented. And then I started getting more curious about, okay, what are some other ways?

[00:10:11] And so what I learned is that you had to really be specific on who you are, what do you want, what do you need? And specifically, what role do you want the model to take on? So how do you want it to behave? So really be clear on who, when, what, how, and all that. And the more clear you were, the better it responded and the more aligned it was to what you were looking for.

[00:10:34] So because of that learning, I just got, I guess it just, as you use it, you learn how to use it. It's like this, you build on your own. It's like driving. At first, you're thinking about your brakes and you're not smooth, but eventually it becomes integrated into just, you don't really think about driving now, right? Like you're just, you know when to press the gas and when to press the brake, right? So it felt like that. So anyway, I'm doing this. And then I think it was like, I would say late January, early February,

[00:11:01] I started to just feel a little bit off, not like myself. I was feeling tired all the time. Didn't have much of an appetite. Just felt like this doom and gloom type of feeling. And I was just like, what is going on with me? And then I was just thinking, oh, maybe I'm tired. Maybe it's burnout or things like that. Just all kinds of stuff was coming up in my head. I do have two little boys. And at the time there were three and six in my head. I was like, oh, maybe it was like working parents life and whatever.

[00:11:28] But so I remember starting Googling things like, oh, women. At the time I was 36. I was like, oh, 36 year old. Could it be menopause, perimenopause, early menopause, hormonal issues? And I'm just Googling all this stuff. And it's just, as we Google, there's like a million things that come up. And I was just looking at that. I was like, there has to be a better rate, right? So I just started to think, oh, what if I asked a question to ChatGPT? Because that's what I was using back then. And I had never done that.

[00:11:57] I'd never asked it like a health-related question. And so I was, in a way, it was more of an experiment. I wasn't going in thinking I have something. It was more of an experimental thing of, oh, I've never had even the thought of using it for a health question. And so I asked it. At first, I was like, I'm feeling this and that. When I use I language, it stopped. It would say, I'm not your doctor. I can't answer health-related questions. So then I was like, oh, let's do role play then.

[00:12:25] So I was like, okay, you're a med student or a resident or whatever. And here's a case for you. Female, 36, gave it some context on the lifestyle. And then gave it a bit more context on what this person, me, was feeling and symptoms that they're reporting. And so I started to ask it, like, what would you do if this person came in? Explain your reasoning. What next steps would you take? Why? How would you approach this problem? Just show me your thinking and reasoning.

[00:12:55] And so it was really interesting to see it respond where it really took on this role and was really looking at it in a very kind of rational way. And I found that really interesting because even though it was answering me, there was a balance of not creating panic versus if you Google some of that stuff, you'll have so many links and what do you click, right?

[00:13:21] So it was just, it's not that interesting, the contrast of this concise response versus having to click through thousands of links. So I just found that interesting. But after that, like, as symptoms kept building, I just kept entering into that same chat and eventually narrowed it down. And it mentioned at one point, it mentioned ovarian cancer. Out of the four possible avenues to look into, it mentioned ovarian cancer.

[00:13:47] Again, at the time, I was not asking any questions like, do I have cancer or things like that. I wasn't looking for cancer, did not ever think of cancer, but it mentioned it. I dismissed it in my head. I was like, whatever, this is AI. Like, I'm not, it's not a doctor. But I think it's still, though, it subconsciously somewhere put that word and it made me more aware of, like, taking this more seriously.

[00:14:11] And what was interesting to me was that when I decided at one point to speak to a family member who's, she was a little bit older than me. And I was like, oh, I'm feeling these things and I'm bloated all the time. I don't have any appetite. I feel tired. I just, I'm sleeping so much. I don't understand. But I, my working hours are not that bad. And my husband is a super partner and we do this household and parenting work perfectly. Like, I don't have any complaints, but it's just, I don't know what I'm feeling. What is this?

[00:14:39] And out of her desire to make me feel better and calmer, she was like, oh, sometimes, like nowadays, like everybody's getting, has perimenopause, like, so early and stuff. So it could be just, like, the luck of being female type of thing. Like, it just comes with it. And I just, in that moment, like, I understood that she was trying to kind of calm me and stuff.

[00:14:59] But it was just that contrast between this history I had been building now, it was like two weeks of talking to AI with these symptoms that was saying this is something you should take seriously and go get checked. Versus a five-minute conversation with a human that loves me that was immediately trying to calm me by saying it's nothing. So just that contrast made me actually, is the thing that made me go get checked. Because I was like, oh, that's too much of a contrast.

[00:15:26] That's too, this thing, this AI model has this history that I've been building for so many days. I don't know, there's too much of a contrast to just this. Did you try to challenge AI? Because I think one of the confusing things that can happen is that you say something like, I think you're wrong. Or what could be the alternative reason? Or could you be wrong?

[00:15:52] And then it might immediately go into, yes, actually, that you're right. Maybe I am overcomplicating. Maybe it's not that serious, which leads to even more confusion. So at this point, did you try to challenge it? Or did you just go with what was building up? So you know what I find interesting is that in those early days, it wasn't as sycophantic as it is now.

[00:16:19] It would push back what you were saying more and it would stop earlier. So for example, like I was saying earlier, if I started talking about me, it would be like, I'm not a healthcare professional. It would just stop. I feel like as more people started using it, they made it more, just more conversational. So in the beginning, early days, it wasn't this conversation. It was really hard to make it do things. It was really like to really know how to prompt and really get it to understand you. To get it around its guardrails. Yes, exactly.

[00:16:47] It was really tough to make it do things and make it do things, answer you in a certain way. Because it wasn't agentic back then. We didn't have any agents. But there's something in what you said that I think is interesting because I think the difference was that because when I started using it, I started with a kind of like an experimental hat. I was already prompting it in a very, you're a tool and I'm testing you way. I was acting more like a tester than a user because I was curious about how would it respond? What would it do?

[00:17:15] So all of my questions were always like, what are the possibilities? What are some? So I wasn't asking, name me one thing and then prove me that you're right. It was more like, what could be possible and how did you get to this? So I was already kind of starting something with something large and then refining rather than give me that one answer and then we'll build proof that it's true. You know what I mean? Yeah, absolutely.

[00:17:40] It's one of the tips that one of the speakers in this series gave was even now to always just ask why. Like, why do you think this? Why would you say that? Why would that be true? And just keep drilling into the questions. But anyway, going back to your story. So you said you decided that this needs to be consulted with a professional.

[00:18:08] How did you approach that? What did you do with the responses that you got from AI? What was your attitude towards sharing that your concerns come from discussions you've had with AI? Yeah. So at this point now, we're in like March of 2023. So it's been like a couple of weeks of me having these symptoms and doing things and they're just progressing. Like more and more things are adding to my pile of symptoms.

[00:18:39] And so at this point, I'm like, okay, I got to go get checked. But what do I say? And if it's such a... Because of this dialogue, I was learning a lot about ovarian cancer as well. And I was learning how hard it is to diagnose. It's a very... There's no specific symptom. It's a constellation of symptoms that happen together that make people think that you have it. But even that constellation of symptoms, it doesn't always point to ovarian cancer. And some people don't have the mix of symptoms I did.

[00:19:08] I actually had a... I'll say the good mix in the sense that it was... It raised enough flags to get it detected. But usually people don't have these symptoms together. So in my case... And for me, as someone that's a patient partner with ovarian cancer, I'm always happy to share because I think it raises awareness too for women. So unfortunately, only around 10% of cases get detected at stage one because it's so hard to detect.

[00:19:34] And it's often one of the cancers that gets dismissed because of the symptoms, because they're so vague. But if you say these things together, if you're experiencing them together, and to know to say it together is a really big, important key. So in my case, specifically, I didn't have any appetite. I was feeling bloated all the time. I was extremely tired. I would be... Sorry, that's not true. I was hungry, but I didn't have... I couldn't eat. So it was like, I am hungry. But when I try to eat, I feel like I'm full.

[00:20:02] So feeling full without actually feeling like you satisfied your hunger. So it's a weird feeling. And I started to have cramps, like really intense cramps. Felt like menstrual cramps, but more intense. And it wasn't my normal. I wasn't someone that got menstrual cramps. I think the key is to know that there's your normal. And then when your normal isn't what you're experiencing, then it should raise a red flag. And then... And eventually, I started to bleed way ahead of my scheduled menstruation cycle. And so that was one of the big red flags for me.

[00:20:32] So it was this combination of all these things that made me wonder something. And then also, that's what I was feeding to Chad TBT at the time. And so then what I'm coming to now is, okay, where do I present to the doctor so that it raises the right red flags for them to get me the tests I need? So that's the conversation now I was having.

[00:20:57] Now that I had decided I'm going to go get seen, the conversation that I was having with the eye was, what symptoms do I need to present together for this case to be taken seriously enough to get the CA-125 test to get the imaging? So it really connected, like I was saying, that constellation of symptoms so that I could present it to know that I have to present it together. Because most women may not present them together. They might say, oh, I'm having really heavy periods all of a sudden. Or they might say, I'm having cramps, right?

[00:21:26] But in isolation, that might get you dismissed. And I don't think it's anyone's fault here, honestly, because if you bring just that one symptom to your provider, they don't have anything else to go by. Okay. Yeah. So, yeah. So it was, at this point, it was now serving more as like a, almost like a navigator to help me get access to the right healthcare as soon as possible, basically. So it was shifting roles, right? In the beginning, it was, okay, what's going on? It was like a live journal that's talking back to me and making sense with me. And it was like a thinking partner.

[00:21:55] And then it transitioned into, okay, how do we get you access to the care that you need? Yeah. Did you at any point worry about the fact that you are sharing very personal data with ChatGPT? Honestly, yeah. And how do you look at, how do you look at more broadly at the whole privacy aspect and a medical data sharing challenge? Yeah.

[00:22:20] So I think it's a good question, but at the time, I wasn't sharing anything specifically personal in the sense that I wasn't sharing any identifying numbers or some numbers or anything like that. I wasn't sharing any, that kind of information. So for me, it was as good as doing a bunch of Google searches using my Gmail account. You know what I mean? Like, is there a history of the searches I do? I don't know. Maybe. Yeah. You know what I'm saying?

[00:22:47] So I was just looking at it from, as I wasn't really thinking by it, to be honest. But there is a point though, when, so now once I'm seeing the doctor, by the way, in the beginning, I did not mention that I was using AI because it's way too early. Nobody was using AI like this. So let's go into that. Let's go into that situation slowly. You went to the doctor and then what happened? So I'll go to the doctor. Did you have a list of things that you knew that you had to say or? Yeah. I had my little post-it.

[00:23:17] I transcribed in my post-it. Okay. Here are the things I have to say based on the conversation I've had and to make sure I mention. And so I was like, yeah, I'm feeling this. Can we get some blood tests? Can you, can you do a pelvic exam? Can you? And so that's what happened. I got a blood test that they did the pelvic exam. And through, through that process, I was immediately sent for imaging. And then they, that's when they saw that. Okay. So what's, so I have to pause here a second.

[00:23:45] What's interesting is that in the beginning, like I said, these symptoms are so vague that they're like, oh, maybe it's fibroids. Because for a young woman at 36, no one's thinking ovarian cancer. It's not the median age, nowhere close. It's usually post-menopausal, like 60s. And I have no family history. So there's no way any medical professional is thinking cancer at this point, right?

[00:24:06] So they did say, let's go do imaging to check for fibroids and do like a blood test panel of just maybe, is it a miscarriage? Am I pregnant? Not realizing it. Like all the typical testing you might give a female coming in complaining of pelvic pain or something. So it's through the doctor's kind of curiosity around what's going on. Let me, let me request imaging.

[00:24:35] Imaging that moment is a pivotal moment in my story. Because if that did not happen at that time, this would not be a stage one diagnosis. I was barely a stage one diagnosis. So that imaging is what led them to see, oh, there's something on your ovary and your uterus is fine. There's something in your ovary. And so then that led to an MRI and et cetera. Yeah. Perhaps worth mentioning at this stage is that the ovarian cancer grows really quickly

[00:25:01] and can progress from early stages to an advanced stage within a year or so. And at the same time, it takes usually quite a few months on average to get diagnosis. So I guess it's between the time that the symptoms present to the time that you actually get a diagnosis and things progress. There's several months that a lot of women get the diagnosis. And also, I think it's something that we should change.

[00:25:29] If a female comes to complain about pelvic pain, I feel like the most common tests that are done are we'll look at your iron, we'll look at your hormones, like pregnancy hormones and things like that. I feel like it's too much associated with are you pregnant or not. Yeah. And it doesn't cost much more to add in the CA-125 tests, which is the cancer antigen 125 test,

[00:25:58] which is the only thing we currently have, which is not even that good for ovarian cancer. But it's still currently, unfortunately, the only thing that we have close to some kind of test. For some women, the numbers show normal, even though they have cancer. And so for some women, it shows really high, but they have something else. It's not because it's related to an inflammatory protein or something like that. So anyway, point is, like, why not just include that as something,

[00:26:26] considering that we don't have any other diagnostic or prevention mechanisms right now for ovarian cancer? I think at this point, it's important to mention that what we discussed was pre-2023. And after that, you became very active in the cancer space. So you were a board member of the Ovarian Cancer Canada. You are still a co-director of Cancer at Work, which is the leading national platform in Canada,

[00:26:56] supporting people with cancer in navigating the return to work. You're also serving as the patient partners committee president of the Quebec Cancer Coalition. So you're very active in the space of cancer patients in Canada. And that you've been doing that work since 2024. So when you look back at the last three years and not just your story,

[00:27:26] but everything that you are encountering through these organizations, what are you observing in terms of how is AI used in cancer care, in health care, in how patients use it? Yeah. Yeah. And we don't have to go on the broad spectrum, just on that patient, very close to patient use. Yeah. So like you're saying, there's such a broad spectrum of it. There's, but I feel like the most, like the low-hanging fruit, like the,

[00:27:58] probably like the majority of people that I've spoken to, the use cases are, it's, it serves as like a health system navigation tool. Like what I was saying earlier, where once I decided like I want to go get seen, it was like acting as a thinking partner or coach to help me figure out who should I be talking to? What is the best type of specialist for this? When like having all that information, because all that information exists,

[00:28:25] it's just that it's not connected in any way for someone who's not coming from a health care background to make sense of it altogether. So I think that when most people are using it, it's because it's like we don't have the capacity to go learn everything about every health care situation or every health care condition and then make sense of it. Right. So I feel like this is where AI is playing that role for most people is to help navigate their own health care journeys. I do think though there is some self-diagnosis stuff happening too,

[00:28:54] which is a bit like, I don't really think that's the safest thing to do. But if it's happening again, let's going back to the nullution thing. If it's happening, that should be a signal, right? Like why are people doing that? There's a reason. So, so I can't blame the person who's trying to help themselves. But yeah, I do think there's people who are looking at what's going on with me is a question. Why am I feeling this way? How do I feel better? What do I need to do to feel better? And then once they're in the health care system. Yeah. And feel heard. What do I need to say? How do I bring this to the doctor or whoever I'm going to see?

[00:29:23] And then once you're in, how do I navigate? How do I know I'm seeing the right specialist at the right time? Is this normal? Is this wait time normal? What's the process? So, so it's, I think it's acting a lot like a navigational assistant. One of the reasons why I started this series of Diogenetic Patient was because I really wonder

[00:29:45] how can we empower patients and also create AI literacy that helps patients navigate healthcare better and also express themselves better? Because sometimes the problem is that you just don't know how to explain something. I mean, even as a simple thing such as different types of pain. Yeah, that's like the pain skills. Like it's so hard.

[00:30:10] How do you know if it's a, if it's a throbbing pain, if it's a stabbing pain, if it's five different types of pains that the doctors will ask you if that's what you're feeling and you'll be like, I don't know what those terms mean. Pain levels. That's just, it's so different for people. Like I have a very high pain tolerance. I'm the kind of person that at night I'm like, oh, what happened here? I didn't see that bruise. Yeah. Yeah. But just going back to the challenge, you know, I'm sure that many, if there's doctors

[00:30:37] listening, when you said that basically you ask AI, how do I need to present my story to the doctor to be taken seriously, I almost feel as if many doctors would wonder, and this is maybe just what I'm wondering is like how many people are going to exploit that just to

[00:31:01] get exams and to which extent this is potentially going to increase the burden on healthcare. Like the worried, well, 3.0 type of a thing. So what do you think about that? Do you have any stories, any experience or any insights into the negative sides of patients using AI?

[00:31:28] I mean, I would say that's something that AI are not already existed, right? It's already something. We already have a huge problem. Forget AI for a second. Look at social media. It's full of people trying to sell to the worried well. All kinds of remedies and all kinds of stuff that has no scientific evidence. We already have an information crisis. So I feel like the problem really is not AI for me. It's we are in this jungle of information.

[00:31:57] And when I say information, it includes the good, the bad, the ugly, all of it. And I'm hoping, my hope is that if we do this right with AI, it could perhaps be the thing that filters out that noise for people. So I actually, I still, like I said, I still have my engineering experiment hat on when I, even as, even if I'm using it from the patient side, but I keep the hat on because it helps me navigate. Right. So one of the things I often do on purpose is I'll go screenshot something I saw on social media that I know is baloney.

[00:32:26] And I'll just go to the chatbot and be like, I saw this. What's, what is this? Let's say someone's selling, I don't know, some magic tea or something for cancer. What's this? I just saw this. What is this about? And then it's just me trying to see how it would respond. And it's pretty good at saying, no, there's no scientific evidence for this. You should not talk to your doctor about this. So it like, it's pretty good at steering you away from misinformation. So those are the kinds of things I'm really excited about.

[00:32:53] Can this become our like GPS when it comes to navigating information and this jungle of information we have? Because we don't have an information, lack of information. We have too much information now and we, our brains cannot process it alone anymore. There's just so much we're bombarded with. So that's how I'm looking at it this way. Because before I look, there's WhatsApp groups with aunties telling you what to do, the magic trees to drink. I had that as a cancer patient.

[00:33:22] I couldn't believe the recommendations I was getting from family to drink lemon tea and stuff like that. And I was just like, wow, like this is not just something I read about. It's happening to me that people are saying this to me. And they mean it's just they're misinformed and then they're sharing that misinformation. And that's not an AI, that's a human being. So this, I think this problem will continue to exist, but I'm hoping that we can leverage AI to improve that human experience of navigating all this misinformation.

[00:33:51] That's how I see it. Let's go to concrete things. So for example, for somebody that's trying to use AI today for their problems, what would be the top three things that you would advise them to be cautious of or how to use AI? Yeah, I think it's, I also understand that what I'm saying is not always easy to do because

[00:34:17] when, like when you're in a crisis, you're willing to jump out of a window, right? If your house is on fire, you're willing to jump out of a three-story building and break your leg because that means you survive. Right? So that's, so I feel when we're in survival mode, we do make decisions that put us at risk. But if we can still take a second to pause and reflect on what is it that I'm trying to solve and keep the prompts in that exploration phase, like think of yourself as something

[00:34:47] that's testing the tool and not something that's using the tool, not as a user, you know, not as someone that's using, but testing. I think that shifts a bit our way of thinking about the tool. It's like, I'm seeing what this can say. It doesn't mean it's definite. I'm still keeping in mind that this is a tool, right? It's not a human, it's not a doctor, it's not a professional. So having that kind of, I think that kind of lens on at all times as a tester. And so the prompts could look like what we were saying earlier, what are the possible things? Here are some symptoms.

[00:35:16] What are the questions I should be asking? And I think the best thing to do is link it back to a professional. So saying, okay, here's what I'm feeling, have that discussion, but then afterwards end with what should I present to my doctor? What are the questions I should be asking them? I think that's the best way to make sure that there's a human in the loop and that we are at the end, we are reconnecting ourselves to a human being who can actually help us. Because as of right now, the AI can't prescribe you anything or operate on you.

[00:35:44] So at one point, you do need to speak to a human to actually get better, right? It's not going to in itself make you feel better. So I think remembering what the desired outcome is, I want to feel better. And the only way for me to do that is to eventually speak to a real human, which is a doctor. Yeah, absolutely. I must say that personally, I went through various stages of just feelings when it comes to AI

[00:36:09] through my personal use as a chronic patient, everything from excitement, everything from feeling much better because I felt like I got clarity from AI. For example, when I got new medications and I asked, this is what I got prescribed. And because this is a medication that has a lot of side effects, can you explain to me what I can potentially expect in the upcoming weeks and what I should be mindful of?

[00:36:36] That was, for example, super helpful to also decrease the anxiety on is something going to happen or not. But then half a year later, I would be explaining my story. I would get a recommendation and then I would say, oh, but I also forgot to say that I have comorbidities. So this is also something you need to take into account. And then AI would say that changes your picture completely.

[00:37:02] And it would go into a completely different mindset, which to me was so confusing that I just decided I want to stay away from AI. So I don't know, what do you, what are the thoughts that come to you when I say that? And where do you see that we should still be mindful or how do we cope this uncertainty, which is huge when you have a big issue?

[00:37:29] As you said, when your house is on fire, you're just like your benchmarks and everything and the things you're willing to do change. And also the fear, obviously. So I find what you're saying really, like I relate to that, like I find it very relevant. I think that it's worth separating two things. So in my case, it was fire emergency, right? It wasn't a situation like, for example, in your case, we're saying it's chronic that you're living with and you're managing, right?

[00:37:58] For me, it was something new. There's a fire. I got to jump out. And so I think at that point, we're willing to take more risk to save ourselves wherever the cost. And then we tell ourselves, I'll deal with the cost later. So I think that's how it served in my case. But also, once I finished treatment, then it was like, now I'm managing this new version of me that is now looking at more like a chronic thing, a situation of looking now in menopause and how I managed that.

[00:38:27] So now it's a new state. And I think with that, like a LLM, like a regular AI, like ChatGPT or whatever people are using like Gemini or whatever it is, Claude, I don't think it's the best tool for that. And I think this is where we're actually stepping into the patient built tools. That's where it's getting interesting because people like us were like, this is annoying because it keeps forgetting the context or it keeps the context window of its history

[00:38:53] of remembering me properly is not working well because Brad is not designed for that, right? So that's where I'm seeing people in the patient space making, who have a tech background often, who end up making the tool and then they start making it for their community. So there are like different tools now that exist that manage specific conditions, let's say, because they're designed for that, right? So I feel like it's really, I feel like what we're telling the story of right now is, okay,

[00:39:20] emergency, it helps me get there through the door. But once I'm there and it's become, let's say, a chronic thing, now we see patient built tech and AI tools coming out of those needs that aren't being met by a general tool like this. So I think it's more of a generalist. Like, it's like, maybe we can, if we can make a parallel, it's like your family practitioner is a generalist and can point you to the right direction, but then eventually you might need a specialist whose specialty is that disease space, right?

[00:39:49] So I wonder if that's where we're going. What would be your one advice in terms of the caution that people need to have when they're using AI for their health-related issues? Yeah. It's, ideally, we don't share identifying information, like, and I know I'm guilty of it. I've done it. But again, I took that risk.

[00:40:17] So I think, for example, I got my test results and I was like, I don't understand this. Explain to me what this is. But should I have taken the time to maybe block out my information up on top, on the header, where I had all of the numbering and everything? Perhaps. You didn't do that. Not in the beginning. Because like I said, in that moment, that's when I realized that this was cancer and that the house was on fire. I was like, I need to jump out at whatever cost.

[00:40:44] So I think it comes down to the risk-reward kind of balance. And so at that time, the risk of not doing it felt really costly to me than the cost of doing it. But now, like in the after, it's like, was that the best idea? No. So at first, it was a thinking partner, like thinking through the symptoms with me and helped me figure out how do I get access to care.

[00:41:14] Then it became that like navigational assistant, if you will, the GPS in the healthcare system for me. But I also was acting also as something that was ensuring continuity of my story, right? In those days where I'm being handed over from different specialists and I'm experiencing major brain fog because of chemo, we have a problem when it comes to continuity of the person's story and data. And often the patient acts as like document control and project manager.

[00:41:44] Sometimes there are situations where my family doctor unfortunately didn't have access to what my oncologist was putting down in notes. And so I was acting as the relay, the person who is most vulnerable in this triangle. But it wasn't, it's not their specific fault. It's just the system is designed in the way that it's actually there to protect me, which is, anyway, it's like that because it's protecting me from my data from being shared between two hospitals. But I'm in the care of these two places. So it should flow, right?

[00:42:12] But anyway, so I become the carrier of information while I'm in this vulnerable state. I'm experiencing brain fog. How do you rely on my memory to remember everything? So that was interesting where AI became that tool that became a repository of, okay, here's what I, here's my appointment. Here's what we talked about. So it became this journal that carried the story for me forward, right? So that to me is a huge gap that it's signaling to the system. And a lot of people are using it like that.

[00:42:39] And even in what you shared a little bit, if you think about it, you are mentioning some of that, right? Like where your history, you're trying to map that history. I took this medication and here's other comorbidities and here's another medication. But you're looking for something that can hold that context and this continuity, right? So yeah, I think that's a big space. I think AI, that we people as patients are trying to fill that gap with AI. But as we saw, sometimes it runs out of the context window and it doesn't keep in context

[00:43:08] the stuff you may have said like last year. Yeah. One of the speakers that was on the series was also Sir Russ Reed Barrow, one of the first speakers in the series. He was explaining how he used three different tools to just take into account all the kind of the weaknesses of different systems. So for example, if you really want to have that longitudinal history, you need to just copy responses somewhere.

[00:43:37] Yeah, yeah, yeah. And just, you know, so yeah. So he combined different things. So I'm going to definitely share his website as well because he is sharing that openly on so on how can you use a system? Let me just also share the screen. How can you use the system to manage your whole story, keep on top of the information and everything?

[00:44:08] Yeah. This is available on the FcancerWith.ai if anyone is just... What's his name again? He's Russ. Oh, okay. Okay. Yeah. So with that, I'm going to conclude our discussion. I would like to emphasize that every patient story is individual. There is no one size fits all. There is positive things in the way that we use AI, but at the end of the day, if you're

[00:44:35] not a doctor, you don't have the medical background and there's no way that AI can give you what medical training does. So I would always say the same thing as AI does at the end of the day. It's the doctors that need to tell you what the end diagnosis could be. But let's keep sharing best practices so we can be better prepared and to also know how to communicate better when we are trying to express what is bothering us on the health

[00:45:05] side. So Alvina, thank you so much for coming on the show and sharing your story. Thank you. And I hope that your health stays intact in the future.

[00:45:17] Thank you.