The Counterintuitive Problem
Every research team wants to talk to experts. The logic seems unassailable: people who know the domain deeply will provide the richest, most detailed data about how things actually work. So you recruit senior practitioners, veteran users, power customers -- and then wonder why your transcripts feel thin.
The paradox is real and well-documented across qualitative methodology. Domain experts have internalized so much knowledge that they can no longer access the granular, step-by-step experience that makes interview data useful for design decisions. They speak in abstractions, skip over critical details they consider obvious, and provide compressed narratives that sound authoritative but lack the texture of lived experience.
This does not mean you should never interview experts. It means you need fundamentally different techniques when you do.
Why Expertise Degrades Interview Data
Automaticity erases conscious access. Expert practitioners have moved most of their domain knowledge from conscious processing to automatic execution. A radiologist reading a scan, a trader evaluating a position, a researcher designing a study -- they cannot articulate their process because it no longer passes through articulable cognition. When you ask them how they do something, they reconstruct a plausible narrative rather than reporting actual experience. This connects directly to research on the articulation gap, where the distance between experience and expression grows with expertise.
Curse of knowledge compresses communication. Experts assume you share their mental models. They skip foundational context, use jargon as shorthand for complex concepts, and omit entire decision stages they consider self-evident. A novice user will tell you "I clicked the blue button because I thought it would save my work" -- an expert says "I just saved it" without mentioning the three alternative approaches they considered and rejected.
Status performance replaces vulnerability. Experts feel social pressure to appear competent during interviews. Admitting confusion, uncertainty, or workarounds feels like a status threat. They present idealized workflows rather than actual messy practices. The principles of building rapport without contaminating data become even more critical here -- experts need specific permission to admit imperfection.
Narrative coherence overrides experience fidelity. Experts have told their professional story many times. They have polished narratives about why they do what they do, how they learned, what works and what does not. These narratives are compelling but post-hoc rationalizations rather than accurate accounts of experience. The narrative coherence bias is amplified in expert populations.
The Expertise Spectrum Problem
Not all expertise creates the same data challenges:
Procedural experts (people who execute complex workflows) struggle most with articulating their process. Their knowledge is embodied in action. Traditional interview questions about "how do you do X" will yield oversimplified descriptions.
Conceptual experts (people who understand why systems work) tend to theorize rather than describe. They explain what should happen rather than what does happen. Their data is analytically interesting but disconnected from actual behavior.
Experiential experts (people who have accumulated extensive domain exposure) provide retrospective summaries rather than episodic detail. They tell you about patterns they have noticed over years rather than what happened last Tuesday.
Each type requires different elicitation strategies.
Techniques That Bypass the Expert Filter
Incident-based questioning over general inquiry. Never ask an expert "how do you typically handle X." They will give you a textbook answer. Instead, ask about the last specific time they handled X. "Walk me through Thursday afternoon when you got that alert." Specificity forces experts out of abstraction and back into episodic memory.
Artifact walkthroughs over self-report. Put something tangible in front of experts and ask them to narrate their interaction with it. Screen recordings, documents they produced, emails they sent. The artifact serves as a memory anchor that pulls granular detail back into conscious access. This approach aligns with what we know about visual elicitation using artifacts for richer interviews.
Teach-back protocols. Ask experts to explain their process as if teaching a brand-new hire. The pedagogical frame forces them to unpack compressed knowledge, identify steps they usually skip, and surface the assumptions they normally leave implicit. "If someone started in your role tomorrow with zero context, what would they need to know to handle this?"
Deliberate naivete positioning. Position yourself as genuinely ignorant about their domain -- even if you are not. "I am not from this field, so please do not assume I know anything." This gives experts explicit permission to state the obvious, which is exactly where the useful details hide.
Process tracing with interruption. During a think-aloud protocol with experts, interrupt them at points where they jump ahead. "Wait -- you just went from looking at the dashboard to deciding to escalate. What happened in between?" Experts will skip the micro-decisions that constitute their expertise unless you catch and unpack each transition.
Counterfactual probing. Ask experts what would change their approach. "Under what conditions would you not do it that way?" Boundary conditions force experts to articulate the decision criteria they have internalized. Understanding the boundaries of the principles behind AI governance in enterprise contexts teaches us that expertise boundaries are where the most actionable insights live.
When Experts Are the Right Choice
Despite the paradox, experts are essential for:
- Understanding system-level dynamics that novices cannot perceive
- Identifying failure modes and edge cases from accumulated experience
- Validating whether observed novice behavior represents genuine patterns or learning artifacts
- Providing strategic context for tactical observations
The key is matching your research question to participant expertise level. If you need to understand moment-to-moment experience, novices provide better data. If you need to understand system dynamics, experts are irreplaceable. Many studies benefit from theoretical sampling approaches that deliberately vary expertise level across participants.
Designing the Expert Interview Differently
Allocate more time. Expert interviews need 75-90 minutes minimum. The first 30 minutes typically yield polished surface narratives. The depth only emerges once experts have exhausted their prepared answers and begin thinking in real time.
Use multiple modalities. Combine interview questions with artifact review, scenario walk-throughs, and observational components. Experts reveal different knowledge through different channels.
Plan for triangulation. Cross-reference expert claims with observational data and novice accounts. Where experts say one thing and novices experience another, you have found a crucial insight about the gap between theory and practice. This triangulation approach, explored in depth in research on how research triangulation strengthens product decisions, is especially valuable with expert participants.
Embrace productive confusion. When an expert says something that contradicts what you have heard from other participants, do not resolve it immediately. Sit with the contradiction. The resolution often reveals assumptions that neither party had surfaced.
The Balanced Recruitment Strategy
The best qualitative studies deliberately sample across expertise levels:
- 2-3 true novices for unfiltered, granular experience data
- 3-4 intermediate practitioners for the sweet spot of conscious competence
- 2-3 deep experts for system-level insight and pattern identification
This distribution gives you both the texture of lived experience and the architecture of domain understanding. The integration across these levels is where the strongest findings emerge.
Practical Implications for AI-Assisted Research
AI moderation tools face the expert participant challenge in amplified form. Without a human moderator's ability to detect compressed answers and probe beneath them, AI interviews with experts risk capturing only surface-level data. Teams using AI-assisted interviewing should build explicit expertise-detection logic that triggers deeper probing protocols when participants demonstrate domain fluency.
The expert participant paradox does not mean expertise is unwelcome in qualitative research. It means expertise demands different extraction techniques -- and the failure to recognize this produces studies full of authoritative-sounding data that contains less actual insight than a conversation with someone who started using your product last week.
