Traceability Is Key To Building Trust in AI Tools: Rhett Alden, PhD, Chief Technical Officer, Health Markets and Raman Kaur, APN-c, BSN-RN, VP of Elsevier Health Education

Lindsey Smith 

Hi, I'm Lindsey Smith, welcoming you to Raise the Line from Elsevier, an ongoing exploration about how to improve health and healthcare. 

One of the most common concerns raised about the use of AI in healthcare is the so-called black box issue: a seeming lack of transparency about how AI arrives at a conclusion. 

We're going to focus on those concerns today by turning to two Elsevier colleagues who are well-placed to guide us through the complexities involved, including the concept of traceability and what role it plays in how AI tools are built and deployed. 

Rhett Alden is the Chief Technical Officer of Health Markets at Elsevier, a role that allows him to drive product development, overall technical strategy, and operational efficiencies in support of the health markets business. He has spent his entire career developing solutions and leading technology strategy in healthcare and life sciences. 

And we're also joined by Raman Kaur. She is the vice president of technology at Elsevier Health Markets, focusing on driving AI, simulation, and platform strategy across global health education portfolios. Among other expertise, she brings years of experience as a nurse practitioner to her work. 

Thank you so much for joining us today.

Rhett Alden 

Hi, thank you. I'm really glad to be here.

Raman Kaur 

Thank you, great to be here, Lindsey.

Lindsey Smith 

So before we get into AI, I'd like to just start with each of you personally. What originally drew you guys both into healthcare, and what's kept you passionate about this work over the years?

Raman Kaur 

So I can go first, Lindsey. I started in healthcare as a nurse and a clinical primary care nurse practitioner. So my entry point was very direct working right at the bedside with patients and seeing how decisions impact real lives. From there, I had the opportunity to work across different parts of the system -- with hospitals, with large insurance organizations -- which gave me a broader view of how fragmented healthcare can be, especially when it comes to information and decision making. 

I think what stood out across all those environments for me was the same problem:

clinicians and teams are spending too much time trying to find, validate, and trust information. That's what pulled me into technology and innovation actually. So I've spent several years now building solutions to improve that siloed system. And now I have, thanks to a lot of the innovative efforts, even some patents under review to try to make this space even more transformative. 

So I would say, to round out what keeps me here, is the impact and scale of change we can make. So if we can improve how knowledge is delivered and trusted, we can improve decisions, we can improve learning, and ultimately patient outcomes. So if we get this right, we're not just improving workflows, we're improving how people take care of other humans, including us.

Lindsey Smith 

Absolutely. And let's go over to Rhett Now.

Rhett Alden 

In some ways my journey is similar but also a little different. I started as a researcher focused in biophysics -- a lot of protein dynamics, site mutagenesis work and DNA, etc. -- and I ended up getting a PhD in biophysics and moved into biotechnology. Initially, I worked at a company that was looking at chromatin structure and kind of cell differentiation, which was really interesting. But as my career evolved, I really wanted to get closer to patients and closer to impacting people staying kind of in early discovery. 

So, I moved more and more into health technology and in finding ways to really engage and improve people's lives. And that's really why I've stayed in healthcare my entire career. You know, I've done work in sequencing DNA as well as transcription work and some medical device work for chronic disease patients. Right now I'm at Elsevier really focused on how do we unlock the knowledge that is embedded in the corpus of information that we have and how do we really liberate that to improve patient outcomes. 

So in the end, my mission is very similar to Raman's. Maybe my journey is askew, but I think we're very aligned on why we're here...to improve the efficiency of healthcare and also make patient’s lives a little bit better.

Lindsey Smith 

I always enjoy hearing people's story and how they end up where they are. Rhett, I think your work in research and biotech and medical device creation brings a really deep understanding of what it takes to build those real solutions. And, Raman, your background with healthcare and technology I think is really a unique blend that gives you guys both really compelling perspectives. 

So, let's shift into AI a little bit here. One issue that comes up again and again is the so-called black box. How do each of you explain that challenge and what does it truly mean for clinicians and patients on the front lines?

Raman Kaur 

Lindsey, that is a very interesting question and often talked about. So in healthcare, the black box problem isn't abstract. It's not just a conceptual thing. It's very real. Clinicians and students are making decisions that impact patient outcomes and building their own knowledge in real time, and they need to trust what they're seeing. 

A lot of AI tools will give you an answer, but they don't really show where the answer came from. Usually it's a summary. So, an amalgamation of several sources, right? That's risky, especially for learners who are building foundational knowledge. So for me, the issue is just not about accuracy. It's the appropriateness and trustworthiness of the content and being able to know where the content is actually coming from. Can I actually trace this back to evidence I recognize and trust, or no?

If you can't do that, you shouldn't be using that answer in a clinical or a learning context from my standpoint, but I would love to hear Rhett's thoughts on it.

Rhett Alden 

Yeah, I agree with what Raman said and I'd like to draw an analogy to medical devices that I brought up earlier. You know, in one example in my work, we developed a DNA test and a pathology test for non-Hodgkin's lymphoma. That test really helped with early diagnosis of a really hard to detect lung cancer type.

The challenge there is no one would use that if it was a black box. They need to understand how it's analyzed, what parameters are used, what the intended use of the product is, and what the failure rate or false positive or false negative rate might be. These are really critical parameters for any clinician, and really anyone in health care.

Now, if we draw that analogy to black box and AI, it really comes down to what I would call epistemic trust, which is trust in the knowledge. Knowledge is really something that is derived through consensus -- consensus of clinicians and consensus of researchers. It's not something that should be taken lightly. And that consensus comes from referenced articles, it comes from plenary journals, it comes from trusted experts in field. 

When we think about black box, that's not there, right? That foundation or grounding isn't present. So it's really critical for anyone using AI, particularly in healthcare, that they have that epistemic grounding of knowledge provenance and where those claims are derived from.

So that I think is where I would juxtapose this idea of black box concerns versus full transparency and claims validation.

Lindsey Smith 

That's a really helpful way to frame it. Raman, you mentioned the system can give you an answer, sometimes a really accurate one, but it can’t always show you how it got there. And Rhett, that was a really great example for clinicians. I could see how that could kind of create some tension in medicine, since it's built on explainability. And if I'm a clinician, I'm going to need to trust the recommendation. Whether it's about a diagnosis or risk or treatment, I really need to understand the why behind it. 

I want to come back to something that Raman said. You mentioned traceability, which might be a new term for some people in our audience. Can you break that down on what traceability actually means in practice and what steps it includes?

Raman Kaur 

Certainly. So traceability is simple in concept, but powerful in practice. It means that the AI can be traced back to a specific source of evidence. A response has a traceable, tangible way for us to know where it's coming from. I believe it can be broken down in three parts, actually: where did the information come from; how was it used to generate the answer. If I know that, then can I go back and verify it myself, right? That entire loop of feedback. 

So in healthcare, that usually means linking back to peer-reviewed articles, clinical guidelines, trusted medical content. So instead of just an answer, you get an answer with evidence, you can inspect and build context for yourself. From a user standpoint, this changes a user's behavior in not necessarily just consuming AI for summarization, but actually trusting AI because it proves traceability to you. So it moves people from passive consumption to active verification, which is what you want in clinical setting.

Lindsey Smith 

That makes a lot of sense. So, if I'm understanding this correctly, that traceability is about having a clear paper trail so that the system isn't a black box and can be trusted and reviewed and validated. Is that fair? 

Raman Kaur 

Absolutely.

Lindsey Smith 

Awesome. Let's move over to Rhett for this next question. Elsevier has many AI-powered products, but for this conversation, let's go ahead and keep our focus on ClinicalKey AI, which is designed to deliver fast, accurate, evidence-based answers at the point of care. Can you tell us a little bit more about what it does and how it fits into a clinician's normal workflow?

Rhett Alden 

Yeah, absolutely. So one of the challenges in being a clinician, particularly in the 21st century, is that patients have become more complex, partly because of our ability to treat them. It's not uncommon to have patients that might have cardiometabolic issues. They may also have cardiology issues and are coming in for yet another problem. These chronic diseases make managing a patient very, very challenging. So, many clinicians are looking for tools to help them look at contraindications, drug interactions, and alternative strategies for managing a particular patient. 

CKAI is really intended to help resolve some of those challenges in finding information about individual patient criteria, and the way it does that is through a combination of a query entry by the clinician, but also pulling information from the medical record itself. In building that context, we're able to give context-aware responses, as well as transparency back to source articles to help validate that.

So what we're finding in the field is that clinicians are using it to validate. You could think of it as a second opinion. Then in some cases, they’re looking from a more exploratory perspective for complex patient problems, where they're dealing with a lot of comorbidities and complex criteria. And then the third piece is really research and even looking at translational medicine, looking at more experimental procedures and other things like that.

So, there's a number of different use cases that are being pointed to for CKAI. And the intelligence in the product is really around understanding the intent of the query and then also, how do we validate every claim and assertion in the response? Because that's one of the key things, as we said earlier, around transparency. Claim validation is super critical so that every claim can be traced back to a reference and a peer-reviewed reference.

Lindsey Smith 

Two things that kind of stood out to me as you were talking was that focus on workflow. If it's not practical in real life, adoption just won't happen, so I think making it practical is so important. And the second point on that trusted content and evidence-based answers and citations is more important than ever in today's day and age. We'll be sure to put the ClinicalKey AI link in the show notes for our listeners to check it out if they haven't already. 

As a follow-up question, I want to shift over to Raman. Let's make this real for listeners. Given your background, can you walk us through a realistic office visit scenario and explain how traceability helps build confidence for both the clinician and the patient?

Raman Kaur 

Sure, Lindsey, and I'm chuckling a little bit because Rhett and I did not align ahead of time on this, but he mentioned cardiometabolic or cardiac disease, and that's the example that I have so our thoughts are very much aligned. 

So, let's walk through an example. Imagine a clinician – it really could be MD, DO, nurse practitioner, or PA -- seeing a patient with a new onset of chest pain. That's very concerning, a very high pressure situation. Say, for example, they use an AI tool to quickly understand possible causes, differential diagnosis, most valid conditions in that particular patient's demographics, or even next steps of what should be done as per the guidelines. A black box system will just give them a summary, lacking traceability details, of what should happen and that's actually not enough in that type of setting. 

A traceable system will show the recommendation, the supporting studies or guidelines that allow the clinician to validate what they're looking at in front of them in real life in a matter of seconds. That changes the confidence that clinician has in that tool, which is traceable. So now the clinician isn't trusting the AI, they're trusting the underlying evidence they're consuming from the AI-assisted platform. 

So why does that matter? That matters because when you're making life altering decisions for other humans under a time constraint and under pressure, you really need to be able to have a system that will put you in the best successful outcome. And it is a grim reality of our medical system today that we are very overburdened and the difference between a black box system and a traceable system is everything.

Rhett Alden 

Maybe I could just add a little bit to that. I mentioned earlier about the integration with the EMR to build more patient context. In this example, you would also be interested in knowing the ethnicity of a patient because some drugs produce different responses based on ethnicity just due to genetic variances. This is something that we can help in pulling in because the clinician may not be aware of that or may not be taking that into consideration in terms of drug selection and dosage. 

So, this is just another piece where we can be more than just a response engine, but actually be one that can help establish a larger context to help give a more precise answer for....I won't call it personalized medicine, but I think of it more as precision medicine to provide a more full context around that individual patient.

Lindsey Smith 

That example really brings this to life and I think it shows how traceability shows up in everyday care and something that that genuinely helps build trust and confidence in that moment. 

Raman, let's go back to you for a moment. From your perspective as a nurse practitioner, how much do clinicians actually need to see behind the scenes of an AI recommendation before you think they feel comfortable using it in practice?

Raman Kaur 

I think the goal for clinicians is not to have a full explanation of AI. I think the goal for them is to understand and be able to explain the evidence behind the recommendation. I don't believe that the clinicians need to be burdened with exact algorithms, exact retrieval methods. What they want to know is where the information is coming from and can I trust it? Because if you think about it for clinicians, their decisions tie directly to their medical responsibility, the license that allows them to treat these patients. So for them, the information source is extremely important. 

And the secondary level is how does the information apply in that particular context or scenario? If they're just getting information and it's not pertinent to the question they're asking, that's another problem, too. So the role of AI should really be to make that flow of information easier through key sources, through clinical reasoning, and of course, keeping context in mind, like Rhett was saying a little while ago. Knowing who and what the patient is. 

So I would say that with the traceability factor, I think the physician or any clinician for that matter can have a very strong clinician-patient conversation instead of feeling like they don't have the right resources just because they don't understand the what behind the system...that they're not equipped with that information, right? The system should be empowering them.

Lindsey Smith

Absolutely. I appreciate that practicality and imagine there's a balance between the transparency of the information source and the overall usability. And balance is especially important I think for people who are just starting out in the field.

Raman I'm gonna have you take this question as well. We have a lot of students and early career professionals listening in, and I wanted to ask you what your advice would as they begin using AI-powered decision support tools in their training and early practice.

Raman Kaur 

I'm happy to answer this from my own experience as a clinician that was emerging through the ranks. So of course, I didn't have AI -- I would have loved to have that -- but early in my career, I used tools constantly. Tools like UpToDate, articles, Google, WebMD, because you are still building your judgment.

So, I would strongly suggest for the students of today or early career professionals to note that what matters is not just getting the answer quickly. I know we all want to be there first and instant gratification is such a big thing, but it's really important to understand why that answer is right for the patient. AI can accelerate that process, but it also can shortcut your own learning if you're not careful, and that's a very big focus area for us in developing education AI tools. So my advice would be use AI to explore and pressure test your thinking, validate your thinking, but don't use AI to replace it. 

So for example – Rhett knows I love giving examples -- if I were seeing a patient and used AI to look at differential diagnosis, I wouldn't stop there. I would ask, well, what evidence is this based on? Does it actually apply to this patient's demographic and their history? What am I missing? And that's how you build that clinical judgment under pressure scenarios. The people who benefit most from AI are not the ones who are relying on it blindly, but the ones who intellectually spar with the AI.

Lindsey Smith 

That's incredibly valuable guidance. So, just to play that back for our listeners, use AI to explore but also validate your individual thinking. We're going to go over to Rhett now for this one. Looking ahead, what is next for Elsevier in the technology space that you're most excited about?

Rhett Alden 

You know, we did a Clinician of the Future survey in 2025 that found 80 % of clinicians are using AI and less than 30 % say they trust it.  I think that's really telling. So when we think about Elsevier's role in this journey, I think it's twofold: one is to establish what it means to be trusted as a purveyor of information and knowledge for clinicians and students; and second, is to move beyond saying we have an AI solution to one in which we're saying that we're helping clinicians as an assistant to solve problems and address challenges that they face at point of care. 

We want to evolve this model where clinicians appreciate and understand what it means from an AI ethical perspective to have that level of epistemic trust and transparency, but then also continue to build integration within clinical workflows so they can realize the value that these tools can bring. So I think when we look at it from Elsevier's perspective, it's those two full pieces along our journey.

And I'd like to add that from a student's perspective or even a clinician's, you know, they're trained in medical ethics -- things around do no harm, justice, beneficence – so they also need to be clear about their position on AI ethics. I think this is another piece where we're trying to educate clinical boards and hospital systems and help them look at what kind of markers and what kind of traits should they be looking for in these solutions, because in the end, our goal is to do no harm as well as we deploy these solutions.

Lindsey Smith 

Right. That is definitely exciting and I think it speaks to how quickly that AI space is evolving. 

As we start to close out this episode, is there anything from either of you that we didn't cover that we should have? Let's go ahead and start with Raman, and we can finish up with Rhett.

Raman Kaur 

Sure, Lindsey. This has been a great conversation, but I think one final parting comment I'd like to make is first, this AI movement is just not a technology shift, I believe it's a trust and responsibility shift. And there is a layer we don't talk about enough about this AI movement, which is sustainability and the global impact of AI. These systems are very resource intensive. We hear about this in the news all the time. They have a real global footprint in terms of energy use and carbon impact. 

So, as we build these systems as technologists, we also have a responsibility to be thoughtful about how often we generate queries, how efficiently we run our models, where we're creating unnecessary load. In healthcare, where we always talk about improving lives at scale, the responsibility extends beyond users into the broader system, right? So getting AI right means building systems that are not only trustworthy, but also responsible in the ways they operate globally.

Lindsey Smith 

Sustainability is a great one, thanks.

Rhett Alden 

Yeah, I think sustainability is super important. I think the other piece that we didn't necessarily talk about is bias. There is bias everywhere in systems and particularly in AI, because a lot of LLMs are trained on the internet. Bias is inherent in our social structure. It's also inherent in some of the research and journal articles, particularly older ones -- either studies that were predominantly done on white males or kind of the misrepresentation of diseases in females because they weren't studied enough.

So, we need to be really cautious in deploying AI systems, not only for students, but also for clinicians to make sure we can mitigate those factors. It's something that I think Elsevier is starting to take a stronger role in, in terms of really looking at how do we mitigate bias and really have an equitable solution for patients.

Lindsey Smith 

Those are two great and really important topics. We'll have to have you guys back on the podcast to get into that as well, but I think that's a great way to close out today's episode. 

This has been a really thoughtful conversation. Thank you both for sharing your perspectives and for helping us better understand what trustworthy AI tools like CKAI can look like in healthcare. Thanks so much.

Rhett Alden 

Thank you so much for having us.

Raman Kaur 

Thank you, Lindsey. Thank you so much.

Lindsey Smith 

And that's a wrap on today's episode. As we've heard, building trust in AI for healthcare isn't just about powerful algorithms. It's about transparency, traceability, and thoughtful design that keeps clinicians and patients at the center. A big thank you to Rhett and Raman for sharing their insights and helping us better understand how these principles come to life in real world tools like CKAI. 

And thank you for listening to the Raise the Line from Elsevier podcast. If you found this conversation valuable, be sure to subscribe, share the episode with a friend and check out our other discussions on improving health and healthcare. 

I'm Lindsey Smith. Remember to do your part to raise the line and strengthen the healthcare system. We're all in this together.