AI Can’t Replace a Researcher's Empathy
Sara Fortier kicks off her book on Design Research Mastery by addressing user research in the new AI era:
“Machines can never develop true empathy — so as practitioners, you need to embody that ‘empathy’ factor. It’s your job to represent the feelings, pain points and needs of your users, and you can only do that when you talk to them and learn from them.”
She’s right, and the proof is in every skillful user interview.
Anthropic recently ran what they’re calling the largest qualitative study ever conducted. 80,508 people, 159 countries, 70 languages, all interviewed by an AI. The scale is baffling, and some of the things that were surfaced are genuinely moving… yet, surfacing emotion and responding to it are different.
Sitting with it as a researcher, I kept thinking about what didn’t happen during those sessions. The pauses that no one followed and the feelings that were almost named but weren’t.
Those gaps aren’t a limitation of scale; they are the limitation of what AI can do in an interview with a person.
What you notice, and what you do with it
There’s a moment in almost every interview where a participant says something and stops. Not because they’re done but because their underlying thoughts are complicated, or they’re not sure they’re allowed to keep going.
When you feel that pause and explore it, that’s Theory of Mind in practice. While AI can follow a keyword, it can’t feel this opportunity to dive deeper.
Imagine a participant mentions, in passing, that they had to redo a year’s worth of work because the product changed. You sense something hanging in the air; it’s thick with tension, so you ask: “How did you feel when you needed to make those changes?”
What they say next about their exhaustion, frustration, and the cost they paid because of a change becomes one of your most critical findings. An AI may tag the original comment as a usability concern, but you sensed where it was headed and made space for the participant to flesh it out.
Do stay curious about what’s underneath. When something shows up in an exhale, a pause, a sentence that trails off, your spidey senses are the only thing that can catch it.
Don’t lead them there. “That must have been really frustrating, right?” isn’t following a signal; it’s planting one. Keep questions neutral, open-ended, and let them name their own experience.
What you bring to the work
Braun and Clarke’s reflexive thematic analysis puts it well. Rigorous qualitative analysis doesn’t happen when the researcher has been removed. Rather, it happens when the researcher’s perspective is acknowledged and applied with intention.
AI has no perspective to bring. No lived experiences of its own, no stake, no feelings of what matters. That may sound like neutrality, but in practice, it leads to missing critical insights that are deeply embedded.
Suppose you’re clustering findings from two user groups. One is adapting quickly to a new, more advanced toolset. The other is struggling, not because they’re less capable, but because they’ve always relied on the product to be straightforward. An AI will surface and categorize both. It won’t recognize that the second group is also at risk of being quietly alienated should this trend continue.
You recognize it because you brought something to work. Your experience, your judgment, your capacity to sit with quotes and know that it means more than just what it says is not a bias that should be removed; it’s the thing that makes the research worthwhile.
Do name what you brought to a finding when you present it. “I kept returning to this because…” is not a weakness. Contextualizing your interpretation and connecting it to patterns you’ve seen across studies reinforces your expertise and gives stakeholders something solid to trust.
Don’t conflate observations with interpretations, or present conclusions without showing reasoning. A finding delivered without a rationale can read as unreliable, or worse, suspicious. Transparency isn’t a qualifier on your confidence. It’s what builds it.
Carrying it into the room
The empathy you bring into an interview doesn’t stay there. It’s what lets you come out of a study and make a case, not just a report.
As Jennifer Fraser said at the Design Research Mastery panel, the role AI can’t do is align people. That work starts the moment you catch what’s unsaid in a session, and it carries through to the meeting where you make sure the right people feel its weight. Not through data volume, but through the right detail chosen because of a deep understanding rooted in empathy.
Do focus on the human moment before a metric. One quote can make the numbers actually mean something. That’s the move AI can’t decide on its own, and that’s the move that can change minds.
Don’t let the loudest or most convincing voice in the room decide for everyone. When a solution gets declared, or a problem gets dismissed, that’s the moment to use what you heard. It’s harder to argue a specific struggle, in a real person’s words, than a slide. Ultimately, design choices informed by research aren’t arbitrary, and it’s on you to make sure the room knows that.
Your edge isn’t going anywhere
AI is getting better at a lot of things. Scale and speed are real advantages, and for some research problems, they’re exactly the right tool. But depth is a different problem entirely, and feeling what’s in the room isn’t something AI can do yet. It can’t sense the deeper meanings, wrestle with what those meanings carry, or know what empathy-building quote may change everything.
You can.
And, in a landscape that’s moving fast toward automation, that capacity may become the thing worth protecting most. So, show up to your next interview knowing that what you bring is irreplaceable.