Carrie Dugan, a user researcher, argues that insight produced in high-stakes research environments is never neutral. Drawing on a session with a 30-year-old cancer patient in Puerto Rico, she shows how trust, context, and ethics are not preconditions for research but are the research itself. She challenges practitioners to recognize who pays the real cost of the insight they produce and to carry findings in a way that honors the conditions that made them possible.
Key Takeaways
- Insight in high-stakes research is negotiated through trust, method, and constraint, not simply extracted from participants.
- Trust operates across four layers: participant disclosure, institutional perception, interpretation, and reporting fidelity. Each layer affects the next.
- Context is not background information. It is the operating system the data runs on, actively producing every response.
- Silence, hesitation, and the decision to continue a session after an emotional breakdown are all data points no discussion guide question would reach.
- When trust erodes, it does not announce itself. It surfaces later as incomplete data, findings that don't land, or decisions that don't reflect what participants said.
- Removing the human element from high-stakes qualitative research through AI or scaled methods risks losing the most meaningful signals entirely.
Questions & Answers
- Decision makers usually find comfort in large data. How do we marry this qualitative approach with quantitative research?
- Dugan emphasized that the marriage happens in the debrief, immediately after the interview. That is where the researcher's voice enters the data. When decision makers arrive with quantitative findings, the researcher's role is to bring the contextual insights alongside them so neither is treated in isolation. She noted that this context regularly gets missed, and it is the researcher's job to ensure it travels all the way through the process.
- How does AI and qualitative research at scale affect the kind of nuanced, high-stakes work you described?
- Dugan said this is her primary fear about the current direction of the industry. While AI has utility for overall theme generation and some current research applications, the signals she described, hesitation, silence, the choice to stay in a session, are the gold pieces most at risk of being lost when the human is removed from a nuanced conversation. High-stakes research is where that loss would be most consequential.
Session Notes
Opening: The Question We Don't Ask
Carrie Dugan opens with a provocation: who actually pays for the insight researchers produce? That cost does not appear in a research plan, but it is present in every decision, whether to push or pause a session, what findings survive into a presentation, and whether those findings are still recognizable to the person who gave them.
In high stakes research, someone is always paying for it. That's the conversation I want to start today.
The Case: A Session in Puerto Rico
Dugan describes a research session with a 30-year-old woman recently diagnosed with a rare cancer in Puerto Rico. The participant answered questions steadily at first, then stopped, looked away from the camera, and said:
It's so beautiful here. The sun makes everything feel wonderful except for me.
The participant needed to see a specialist on the other side of the island, reachable by one road, one lane. The weight of her situation broke through. The session was paused. When it resumed, she apologized and chose to continue in order to help other people she would never meet.
This moment crystallized the core argument: the data did not arrive cleanly. It came through grief, vulnerability, interrupted composure, and the hard limits of her access to care. She was not simply sharing information. She was trying to survive something that was actively shaping what she could say.
What High-Stakes Research Changes
High-stakes environments do not simply raise the consequences of research. They change what everything means across four dimensions.
- Questions become interventions. Every question has the potential to increase fear, create burden, or provide relief. Information gathering and impact on the participant cannot be separated.
- Answers are shaped by systems. What a participant says is evidence of a system acting on a person, not a clean report of preference or experience.
- Trust becomes the method. The quality of understanding depends entirely on whether a participant feels safe enough to tell the truth about what is actually happening to them.
- Context is the subject of study. Diagnosis, transportation, insurance, language, culture, and fear are not background. They explain behavior and produce the data.
How Participation Changes Under Vulnerability
When participants are vulnerable, they are not simply answering questions. They are making a real-time calculation about how much they can reveal without losing something they cannot afford to lose: energy, dignity, or stability.
- Incomplete information may be self-protection, not forgetfulness.
- Apparent inconsistency may reflect competing pressures.
- Surface-level agreement may be an attempt to avoid conflict.
- The most important signal is often what someone does not say, how fast they move past something, or where their language suddenly becomes vague or controlled.
The question stops being 'is this person telling the truth?' It becomes 'what conditions would make the truth possible here?'
Trust as Infrastructure: Four Layers
Dugan argues trust is not a warm-up or a prerequisite. It is a structure that operates across the entire research lifecycle. She identifies four layers where trust operates and where it can fail.
Layer 1: Participant Disclosure
Participants evaluate how much they can safely reveal. A patient may fear losing care access. An employee may fear retaliation. A citizen may fear scrutiny. The research design must create conditions under which truth is possible, not assume it will surface automatically.
Layer 2: Institutional Trust
Participants do not evaluate the researcher in isolation. They evaluate the institution behind the researcher. A participant may trust the individual interviewer while distrusting the sponsoring organization. Rapport alone cannot bridge that gap. Institutional trust determines whether participants experience research as an opportunity to be understood or a mechanism of control.
Layer 3: Interpretive Trust
The same statement can carry entirely different meanings depending on context, history, and power dynamics. A participant saying 'the process works fine' may be expressing satisfaction, resignation, exhaustion, or fear of criticizing the system. The transcript alone cannot distinguish between them. Without humility, interpretation becomes projection.
Layer 4: Reporting Fidelity
Research rarely travels directly from participant to decision. It passes through researcher, team, leadership, strategy, and action. At every handoff, findings can be simplified, reframed, or selectively emphasized, often without intent. A nuanced finding becomes a headline. A systemic issue becomes a usability problem. A trust issue becomes a communication issue.
Can the reality participants shared survive the journey from conversation to decision? If it cannot, the research has failed regardless of how rigorous the methodology appeared.
How the Layers Connect
Each layer feeds the next in a chain of consequence.
- If participants don't trust the system, disclosure changes.
- If disclosure changes, interpretation changes.
- If interpretation changes, reporting changes.
- If reporting changes, decisions change.
Trust failures rarely announce themselves in a single session. They erode earlier and surface later as incomplete data, findings that don't land, or decisions that don't reflect what participants actually said. By the time the failure is visible, it has been traveling across sessions and handoffs for some time.
Context as the Operating System
Methods are filtering systems for reality, not windows into it. Interviews shape what counts as experience worth naming. Surveys convert ambiguity into categories. Logs capture what is measurable but not necessarily what is meaningful. In high-stakes research, what gets filtered out is not just nuance. It is the signal.
In the Puerto Rico session, everything the participant said came through a cancer diagnosis, one road between her and the care she needed, the weight of breaking down in front of a stranger, and the choice to stay in the study anyway. Those are not background details. They are a system acting on a person, and every response she gave was produced inside that system.
Closing: The Shift in How to Think About This Work
Dugan closes by naming the assumption that needs to break: that there is a clean layer of truth underneath, waiting to be extracted if the research is designed well enough. That assumption holds in low-stakes environments. In high-stakes environments it does not.
- What people say is always situated, shaped by stress, risk, power, and what feels safe to express in the moment.
- Silence carries meaning. Hesitation is data. The decision to stay in a session after breaking down in tears tells you something no discussion guide question would reach.
- There is no unmediated truth waiting to be uncovered. There are only accounts, partial, constrained, and deeply contextual.
- The researcher's job is not to extract those accounts but to interpret them carefully, with full awareness of the conditions that produced them.
Insight isn't discovered. It's negotiated through trust, through method, and through constraint. It's only possible because someone under pressure you may never fully understand decided you were worth telling the truth to.
Insight creates responsibility, not only to report accurately, but to carry what was learned in a way that honors the trust extended, the vulnerability risked, and the context that made the truth speakable at all.
A Note on AI and Scaled Qualitative Methods
In discussion, Dugan expressed concern that the industry's move toward AI-driven qualitative research at scale risks losing the very signals that make high-stakes research valuable. Theme generation and surface-level synthesis may be serviceable uses of AI, but the gold nuggets she describes, the hesitation, the silence, the choice to continue, require a skilled human presence that scaled tools cannot replicate.
Transcript
Read the full transcript
All right, welcome back from the break everyone. We are at the top of the hour, so I think we're going to get rolling here with our final presentation of the day. Can't believe it. Uh, the day has flown by. Um, let's see. So, um, cleanup hitter for today, our final pres presenter is going to be Carrie Dugan. Um, I'm going to be hosting her uh her session, but I'm going to step away and let her kind of do her thing. She's going to be talking to us about the price of insight, the overlooked realities in conducting research in vulnerable communities. Um, I think it's a very interesting topic. I'm very excited to hear her dig into it. Uh, so I'm going to shut up. I'm going to go dark and I'm going to let uh Carrie kind of do her thing. >> Thank you, Bill. Hello, everyone. I am going to share a deck. I have just bear with me for a moment. Awesome. I'm Carrie Dugan. I'm a user researcher and I've spent the most meaningful part of my career in environments where stakes are real, where you learn about people and what you do with it actually matters. Here's a question I don't think we ask enough in this field. Who pays for the price of insight we produce? The price doesn't show up in the research plan, but it's there. It's in every decision whether we uh choose to push or pause, whether to report it or what we found or what the organization can absorb, whether the finding that survived in the presentation is recognizable to the person who gave it to you. That's what I mean by the price of insight. And in high stakes research, someone is always paying for it. That's the conversation I want to start today. I was conducting a research session with a woman in Puerto Rico, 30 years old, recently diagnosed with a rare cancer. She kept her composure at first, answering questions, steady voice, trying to make sense of what was happening to her in real time. Then she stopped, looked away from the camera. It's so beautiful here. The sun makes everything feel wonderful except for me. She needed to see a panelist on the a specialist, I'm sorry, on the other side of the island. One road, one lane, one specialist. a trip that should have been simple but wasn't when you're sick and running out of certainty. As she explained it, her voice started to crack and then she burst into tears unexpectedly, no longer able to keep it together inside the structure of an interview. We stopped the session. When we came back, she apologized and despite her situation, she wanted to continue to help other people. I remember thinking how often we design research to extract insight cleanly from moments that are anything but clean. And how rarely we talk about what it costs to get it. The data wasn't just responses. It came through grief, interruption, vulnerability, and limits of access to care. She wasn't simply sharing information. She was trying to survive something that was actively shaping her what she could even say in the moment. I wasn't just uncertain about what I would hear. I was uncertain whether I should be the one asking at all. That's when I understood something I hadn't been able to articulate before. Insight is not neutral. It's taken from real people inside real constraints, health, geography, systems that fail them. Moments of emotional collapse. In high stakes environment, insight is not discovered. It's negotiated through trust, through method, and through constraint. She didn't hand me understanding. We arrived at it together inside conditions that neither of us fully controlled. And if any one of those things had been different, if she hadn't trusted the process enough to come back, if she if the method hadn't given her room to stop, if the situation hadn't pressed so hard on what she could even say, we would have gotten something cleaner and much less true. Here's what I've learned about high stakes research. It doesn't just raise the consequences, it changes ev what everything means. In that session in Puerto Rico, every question I had to I had the potential to increase fear, create burden or provide relief. questions stopped being information gathering tools and rather they became interventions. Her answers weren't just responses. They were shaped by cancer, by financial pressure, by fin by family obligations, and by what health care she could even access. What she said was evidence of a system acting on a person. Trust stopped being something you establish in recruitment email or at the beginning of a session. It became the method itself. The quality of what I learned depended entirely on whether she felt safe enough to tell the truth about what was actually happening to her. context, her diagnosis, yes, but also transportation insurance language culture, fear. That wasn't background information anymore. It was the subject of study. The system was explaining her behavior. Then there's ethics. In high stakes research, ethics isn't a form you sign at the beginning. It's a responsibility you carry through the whole thing and after the interview ends. In low stakes inter research, people tell you what they prefer. In high stakes research, people show you what they must do to survive. When someone is vulnerable, participation changes. They're not just answering questions. They're making a calculation in real time. Much of how themselves can be how much they feel safe in to enter that conversation without costing them energy they don't have, dignity they're holding on to, or stability they can't afford to lose. That changes your job as a researcher. You're no longer trying to elicit answers. You're trying to notice what the situation is actually allowing to surface. Many times, the most important signal isn't what someone says. It's what they don't say. It's how fast they move past something. It's where their language suddenly gets vague or careful or controlled in a way that it wasn't a moment ago. It changes what valid even means. An answer isn't just right or wrong, accurate or inaccurate. It has to be understood against the conditions under which it was produced. Vulnerability doesn't sit outside of the data. It reshapes the data as it's being formed. In routine research, trust is a warm-up. You build rapport. People relax, the session flows. We tend to think of research as a single thing. You either have it or you don't. But in high stakes research, trust isn't a moment. It's a structure. And it operates at multiple levels at once, some visible and some not. In high stakes research, trust becomes the actual method. In the Puerto Rico story, the risk was never that we would not get data. The risk was that we would get data without understanding. We could have left with the answers to every question on the discussion guide and still miss the story that made those answers meaningful. Trust changed what became visible. the fears, the constraints, relationships and circumstances shaping every decision she made. Without it, we would have collected responses. With it, we gained understanding. That's why trust is not what you do before the research. It is the research. Interviews shape what counts as experience worth naming. Surveys turn ambiguity into categories. Logs capture what's measurable, but not necessarily what's meaningful. Every method is a filtering system for reality, not a window into it. In high stakes research, what gets filtered out isn't just nuance, it's the signal. And trust failures don't look like trust failures here. When trust breaks down, it doesn't announce itself. It shows up as incomplete data, findings that don't quite land, decisions that don't reflect what participants actually said. By the time it's visible, it's actually been eroding for a while across sessions synthesis handoffs. The failure looks like a research problem, but it's a trust problem that has traveled. Trust rarely breaks in this session. It erodess much earlier. To understand where trust fails, you have to understand where it lives. And it doesn't live in one place. It operates at every stage in what participants feel safe in revealing in how institutions are p perceived in how findings get interpreted and reported. That's what has to be treated as infrastructure not instinct. Think of it in layers. The first question most researchers ask is whether they can trust part what participants tell them in high stakes environment. That's the wrong starting point. People in these situations aren't simply reporting experiences. They're managing risks the researcher cannot immediately see. A patient may fear losing access to care. An employee may fear retaliation. A customer may fear financial consequences. A citizen may fear scrutiny. Every response from a participant is a calculation. How much can I reveal here to this person right now without it costing me something I can't afford to lose? Which means what might look like incomplete information might be self-p protection. What appears inconsistent may reflect competing pressures. And what sounds like inre agreement may be the attempt to avoid conflict. The question stops being is this person telling the truth? It becomes what conditions would make the truth possible here? That's not a subtle reframe. It's fundamentally a different research problem and it changes everything about how you design the session. Hold the space and interpret what you hear. People do not evaluate researchers in isolation. They evaluate the institution behind the researcher. Participants are constantly asking rarely out loud, even if it's a blinded study, who benefits from this study? What will happen with what I share? Will anything actually change? Is this organization trying to help me or manage me? A participant may trust you personally while distrusting the organization you represent. When that happens, rapport alone cannot solve this problem. Trust is not only interpersonal, it is also institutional. An institutional trust often determines whether people experience research as an opportunity to be understood or a mechanism of control. Even when participants speak openly, another challenge happens. Can we trust the interpretation of what we heard? Meaning is never fully contained in words. The same statement can mean different things depending on the context, the history, power dynamics, and constraint. A participant saying the process works fine may be expressing satisfaction or resignation or exhaustion or fear of criticizing the system. The transcript script alone cannot tell you which one. Researchers are not simply collecting answers. They're constructing explanations and those explanations depend on the assumptions brought into analysis. Trust in interpretation requires humility. Recognizing that participants are experts in their own experience, but researchers are responsible for understanding the conditions that shaped it. Without that discipline, interpretation becomes projection. Research rarely travels directly from participant to decision. It passes through stages. Participant, researcher, team leadership strategy action. At every stage, information can be simplified, reframed, or selectively emphasized. Often not intentionally, but because organizations are complex and attention is limited. A nuance finding might become a headline. An uncomfortable finding might become a recommendation. A systemic issue becomes a usability problem. A trust issue becomes a communication issue. Reporting fidelity asks one critical question. Can the the reality participants shared survive the journey from conversation to decision? If it cannot, the research has failed regardless of how rigorous the methodology appeared. All four layers point to the same truth. Trust is not a prerequisite that allows research to begin. It operates across the entire life cycle of inquiry influencing disclosure, interpretation, communication, and action. And each layer affects the next. If participants don't trust the system, disclosure changes. If disclosure changes, interpretation changes. If interpretation changes, reporting changes. If reporting changes, decisions change. We often treat context as background, this setup before the finding. But in high stakes research, context isn't surrounding the data, it's producing it. Everything she said came through a cancer diagnosis. A single road between her and the care she needed. the weight of breaking down in front of a stranger and the choice to stay in the study anyway. That's not background. That's a system acting on a person and every response she gave was produced inside of it. That's what makes trust an active research condition. It doesn't just make people comfortable. It changes what's possible to know. In vulnerable participant research, the usual assumption, freedom, comprehension, comfort are already compromised before you ask the first question. Context isn't background. It's the operating system the data runs on. Halfway through the session, something shifted. The questions I had prepared suddenly felt like the wrong instrument for what was happening in front of me. We hadn't come to study disruption, but disruption was the only honest subject in the room. What uncertainty does to a person isn't abstract. It recognizes what they can think about, what they can plan for, what they can hold in attention. She wasn't giving me a healthc care experience. She was showing me what the co it cost to keep functioning when the ground has moved beneath you and hasn't stopped moving. That's what made context visible. Not more detail about the same thing, a completely different thing. The individual experience dissolved into something larger, a set of forces, medical geographic financial emotional, that had been shaping her long before we spoke and would continue shaping her long after. We didn't discover that. We were allowed to see that the price of insight isn't paid at the end of analysis. It gets paid every time complexity gets simplified to fit a timeline. Every time a finding travels from one handoff too many and arrives at something the participant wouldn't recognize. Every time the system that produced the insight, it's treated as background instead of the subject. Insight creates responsibility, not just to report accurately, but to carry what you learned in a way that honors the conditions under which made it possible. The trust extended, the vulnerability risked, the context that made the truth speakable at all. The participant who told you the truth that one day might be affected by what you did with it, that's the price. And in high stakes research, someone is always paying it. I'm going to stop sharing and I'm wondering if there's any questions uh from the audience that I can answer. But before I do that, I I want to close this talk. I want to leave you with the shift that changed how I think about this work. We enter research believing there's a clean layer of truth be underneath that if we ask the right questions, design the right process and analyze carefully enough, insight is what we extract at the end. That assumption holds true in low stakes environments. In high stakes, it doesn't. What people say is always situated, shaped by stress, risk, power, and what feels safe to express in the moment. Silence doesn't or silence carries meaning. Hesitation is data. Even if the decision to stay in a session after breaking down in tears, that tells you something no question on a discussion guide would have ever reached. There is no unmediated truth waiting to be uncovered. There are only accounts partially constrained and deeply contextual. And our job is not to extract them. It is to interpret them carefully and with full awareness of the conditions that produce them. Some of you might be thinking this sounds like ethnography. And you're right. It shares the same instinct. Context matters. Presence matters. Meaning is situated. But ethnography has time, immersion, the ability to let understanding accumulate slowly across weeks or months. High stakes research rarely has that. You have one session, one moment of access with someone already under pressure before they arrive. The field isn't stable. It's actively destabilizing. And the ethical weight is immediate in a way that ethnography is designed to avoid. A question asked at the wrong moment, a finding that travels into a policy decision, the consequences don't wait for you to finish your analysis. That's the price of insight in these environments, not just research methods. The recognition that you are interpreting situation accounts under conditions you didn't create and cannot fully control. And what you did with that insight will outlast the conversation. Which brings us back to where we started. A woman in Puerto Rico, 30 years old, one road, one specialist, the sun making everything feel wonderful. Except for her. She didn't hand me insight. She allowed me to be present for something real under conditions that made the truth extraordinarily difficult to reach and stayed anyway to help people she would never meet. In high stakes environments, insight isn't discovered. It's negotiated through trust, through method, and through constraint. It's only possible because someone under pressure you may never fully understand decided you were worth telling the truth to. >> Very well said. Um I'm just checking now for any uh any questions. I'm seeing some comments. Uh looks like both Lynn and Karen are definitely uh underscoring and and you know agreeing with just about everything you have to say. Um Lynn says, "No question is just validation of the way I've approved qualitative research for my entire career. Uh hesitation is data." Um and another comment uh these same drivers explain people willing to enroll in clinical trials who will not benefit from the therapy from the therapy itself. Let's see. Sesh has uh just a great presentation. Agree about getting to the real insights in the context of uh respondent in highstakes situations, but decision makers usually find comfort in large data. How do we marry this approach with quantitative research? >> That is really a a really good point. I think that it happens really early after the interview. It's I I don't think I've ever felt um where a debrief was so important because that is where you get your voice into that data. So people come to the table with that data with that quantitative data. You are bringing those insights. I think that definitely can get missed and do get missed and it's that context that marries that um that data and I think it's your job as a researcher to make sure that those are seen all the way through the process. >> Yeah, I think your topic is a particular interest in the current research environment, right? Um, what we as a an industry are, you know, kids in on Christmas morning about right now is AI, everything, qualitative research at scale, all these >> new and shiny tech methods that that we can start applying to research. But I mean this is a great example of you know just because you can should you like you know when you start to remove remove the human from what is a very nuanced you know conversation approach um you know we run the real risk of of losing out on a lot of what it is that you you described today. Right. >> That that is really my fear of what's going on today. Um and and you see that really in high stakes research much more than just validation even more than discovery work. Um, and while AI is is good for, you know, overall theme generation or how we're using it right now in research, I I just think that I fear that these things are going to get lost and these are really the gold pieces, >> the gold nuggets. >> Yep. That's and and you need someone who's actually skilled at mining for the for for for such gold. Um I guess something that I thought of and even in the you know call it low stakes research environment. I think this is just a fantastic reminder of at the end of the day the participants are human and you know maybe not even that you're aware of or that was even a part of the construct of the research but you know potentially vulnerable in their own right what baggage or external you know things folks are bringing to the table when they're sitting down with you for that interview. I mean the best example I can think of of late is you know my firm does a lot of research among um on the subject of like home improvement and among uh contractors you know construction laborers and this is a population that is you know heavily uh Latin American Hispanic immigrant and the difficulty that we've had of late in you know getting folks to to open up in light of you know some of the the immigration crackdown issues that are happening externally and this is I mean research about you know power tools which should have nothing to do with you know these other factors but but it does and it's things that we all have to take into consideration >> it's so true everything in affecting our environment politics our health our families the weight of everything is what we bring to everything and when these are escalated is really hard to separate even if you are trying It's human >> indeed. >> Yeah. >> Um let's see. And Shirley came through with uh just comment. Each of these sessions raised profound research insights for my examination. Touched my soul. Um and that seems to be the echo of of a lot of the folks. Um yeah, very well articulated. Um greatly said, much appreciated. I'm just flipping over checking all the chat channels to see if I haven't missed any questions. And I think that's all we have at the moment. >> Um Carrie, yeah, that that was fantastic. Uh really fun. >> Much appreciated. Um and you're our last presenter of the day. So >> thank you. >> Really want to thank you. Thank all of our presenters. Uh thank all the attendees for for coming and showing up and you know making this you know what continues to be an engaging interaction. And that's kind of the intent of all this is just, you know, bunch of research and insights nerds get together, talk a little shop and and get a little bit small smaller, smarter w with every interaction. Um, so thanks y'all. Uh, presentations, if you missed any or had to step out for meetings, will be available in the next few days on our website uh for on demand download and viewing. Uh, and we'll hope to see you around next time. Carrie, thank you. >> Thank you very much. Thanks so much. >> Bye. >> Thank you.
