My Work with Indi Young

The day Indi asked why I was quiet

In the middle of one of Indi Young's group practice sessions, she stopped and asked whether I had anything to add. I had been quiet for a while.

My camera was off. I was listening through an earbud while handling an elder-care interruption in the room around me while in a Zoom meeting.

I was present and attentive, but not in the clean way professional life likes to imagine presence. I could hear the group moving cards across a shared whiteboard doing affinity analysis. I could follow people testing whether two stories belonged together. I could hear Indi asking us to explain our reasoning rather than quietly rearranging the board. At the same time, another part of my life required immediate attention. I was helping my uncle to the bathroom.

So I told her over my earbud where I was. I was listening. I was caring for someone. I was doing both.

That moment has stayed with me because it contains more of this story than I understood then.

In May of 2025, I had stepped away from a thriving full-time career to help care for an older relative. I carried real regret about the professional life I had interrupted. I did not regret the care itself. The regret came from watching a career still full of energy lose its conventional shape while the world of artificial intelligence was opening faster than I had ever seen a field open.

And yet there I was, inside one of the most demanding forms of qualitative training I had encountered, listening through an earpiece while caregiving made its claim on the room. How could I be doing both? It felt unreal.

I thought I had walked away from the work.

I had not.

I had walked into a different way of doing it.

The career I thought I had interrupted

People who have not spent much time caring for an older person may not know how much of care is sitting and waiting.

There are urgent moments. There are appointments, physical tasks, decisions, and forms of vigilance that never fully switch off. But there are also long stretches when the responsible thing is simply to remain nearby.

That time does not behave like leisure. You cannot wander far. You cannot always enter a meeting with your camera on and a tidy desk behind you. You learn to work in intervals, to hold a thought while listening for movement in another room, and to return to it after an interruption without resenting the person who needed you.

Those hours gave me space to do research.

I continued consulting. I continued building. I was still following the frontier of artificial intelligence, but I was no longer doing it only through the demands of a formal role. I could spend days on questions that would have been hard to justify inside a tight quarterly budget plan. I could examine how large language models handled interview transcripts, where summaries became falsely confident, what provenance disappeared inside a clean synthesis, and which parts of qualitative judgment could be supported by agents without being surrendered to them.

I have probably put more hours into qualitative AI research during this period than at any other point in my career.

The market is ready for that work now. Organizations can already generate themes and summaries quickly. The more complex need is becoming visible: how to know whether the system understands the evidence, whether a claim remains connected to the person who made it possible, and where a human being still has to accept responsibility.

My interest in those questions did not begin with Indi. I ran my first paper-prototype research in 2007, and I began reading Indi's Mental Models the following year. Through later roles in software delivery, information architecture, infrastructure, data systems, and ResearchOps, I kept carrying a human-centered practice even when my title belonged to something else. At Nationwide, I worked inside the conditions of a large enterprise, where research had to survive business unit diversity, access permissions, compliance, vendor platforms, shifting strategies, reorgs and the simple fact that people move to other jobs.

Before I formally studied Data Science That Listens (DStL), I had already spent years watching good research lose context as it traveled. It happens more often than not. Too often, stakeholder assumptions controlled the frame before a method had a chance to reveal anything unexpected. Indi's work systematically resists that control.

What Indi’s practice gave me was a more exact way to see what had been lost.

Being invited into the work

During this period, Indi invited me into her small team of researchers.

I do not treat that invitation lightly.

Her work asks more of a researcher than attentive conversation. It asks the listener to notice when their own experience has entered the room, when a familiar topic has become a shortcut, and when the desire to help has arrived before understanding. It asks analysts to stay close enough to another person's interior reasoning that the resulting patterns do not become prettier versions of their own assumptions. Indi is relentless about this.

Indi has spent years building a practice for doing this. I have not encountered a more adept qualitative data analyst. I may never meet one better than her in my lifetime. She’s rare.

That assessment comes partly from watching her work with a transcript and partly from watching her work with other analysts. She can detect the moment a phrase has drifted away from the participant. She can hear when a group of statements shares a topic but not a way of reasoning. She can see when a label sounds coherent because the analysts want coherence, not because every person represented in the group belongs there.

She is transparent in the practical sense. In our group Slack discussions, she answers questions where other learners can see the answer. She corrects the framing when the framing is wrong. She will explain a distinction again, but she will also tell me when she has already explained it and I have failed to carry it forward accurately.

Her generosity is not softness.

It is access to the work, including access to discipline.

I entered the Listening Deeply practice in the spring of 2026. Later, I joined the June practice group for Finding Thinking Styles. The self-paced material gave me the language and the edge cases. The live group made the discipline physical.

There were about ninety short interview transcripts from people deciding whether or not to go out to a performance. Each researcher was assigned a small set of transcripts to analyze and a small group of people to represent. The task was not to become an expert on concerts, theaters, or ticket buying. It was to stay close to how particular people reasoned while making a decision.

That difference sounds manageable once the resulting cognition has been compressed into a clean phrase beginning with an active verb.

On a shared board with nearly a hundred cards and several analysts speaking for people they had come to know through transcripts, it was not manageable at all.

It was alive.

Representing people, not cards

The practice group changed the moral texture of analysis for me.

A card on a whiteboard looks like data. It has a title, a color, a position, and perhaps a few lines of text. It can be moved with a mouse. Once several cards sit together, the arrangement begins to look like a finding.

But each card stood for a person whose reasoning one of us had agreed to represent.

The responsibility was not simply to place the card where it seemed semantically similar. We had to explain why it belonged. We had to return to the person's supported cognition when the title tempted the group toward a superficial match. We did this weekly for nearly two months on one study. If someone was absent, another researcher could adopt their cards to work with, but the card did not become ownerless material available for anyone's intuition.

In one session, I adopted a card from an absent teammate. The person described in it was drawn toward the versatility and creativity displayed in a live performance. To represent that person responsibly, I could not merely say that they liked the same sort of event as someone else. I had to understand what the experience meant inside their decision.

Another teammate showed me what this could sound like. Instead of describing a participant from outside, she spoke provisionally in the first person, from that person's point of view. The shift was small and immediate. The card stopped being a category candidate and became a logic someone in the room had to answer for.

That first-person advocacy was not permission to invent. It had to remain supported by the transcript and the cognition already identified. The purpose was not theater. It was to keep the analysis close enough to the person that a broad label could not erase the reason they belonged.

This is one of the places where my agentic work began to change.

Before the practice group, I could imagine several agents comparing text and negotiating clusters. After the practice group, that architecture felt insufficient. A useful agent would need bounded responsibility for particular people. It would need to advocate from evidence before seeing everyone else's conclusions. It would need to preserve disagreement, show why a move occurred, and distinguish a first-person discussion artifact from a source quote.

The human practice group was not performing a slower version of clustering.

It was distributing responsibility.

Learning through a provisional board

An early pass through the board was intentionally provisional.

That gave us permission to be wrong without making wrongness meaningless.

We moved our assigned cards from beside our names into tentative groups. We were supposed to speak when we moved them. We were not supposed to copy a card into two groups simply because we could see two possible relationships. The board needed one canonical location so that ambiguity would become a conversation rather than a hidden duplication.

My own placements changed.

A story about agreeing to an outing first appeared to belong with a group organized around making an experience meaningful for someone else. As the group talked, a different relationship became visible, one more connected to social connection and letting another person shape the plan. I moved the card.

Another one of my cards remained alone because none of the emerging groups carried enough of its reasoning. Leaving it single felt unfinished. It was also more honest than forcing it into the nearest available meaning.

The early board was messy. Several groups were held together by one broad shared idea while the rest of the cognition diverged. Energy looked like a pattern until the group asked what the energy allowed each person to do. Care for another person looked like a pattern until child exposure, partner enjoyment, and facilitating family connection began pulling in different directions.

The method did not reward the first plausible answer.

It made us live with provisional ones.

In later sessions, group descriptions became invitations and tests. A new phrase could make a previously unrelated card belong. The same phrase could expose why another card did not. We formed subsets rather than flattening a difference. We moved cards instead of defending old placements. We left some cards single. We created a group around heritage and social connection that felt meaningful, then deleted it when the relationship did not hold strongly enough.

That deletion mattered.

The group session was not wasted work. It was evidence that a compelling theme can still be the wrong affinity. The method had protected the participants from an explanation the analysts found attractive.

By the final sessions, I had begun to understand that human affinity analysis is not consensus generation. It is disciplined provisionality. Analysts make a possible sense public, test it through the people they represent, revise the shared structure, and preserve the fact that another interpretation once looked plausible.

Indi guided the process without taking every answer away from us. She asked questions, noticed where a group might split, recorded language as it emerged, and intervened when the method was drifting. But the goal was not to have her silently repair the board into correctness.

The work had to become ours without ceasing to be accountable.

The generosity of correction

I test methods by pushing on their edges.

Sometimes that produces a useful question. Sometimes it exposes that I have misunderstood the center.

At one point I sent Indi a worked example from a transcript. I called the output a form of verb-framing rather than using Indi's term, verb-forwarding. I also connected it too early to another analytical layer and explained why certain words should be excluded using my own emerging logic.

She corrected all three premises.

The unit already had a name inside her method. The analytical sequence mattered. And the reason for excluding a word was not the reason I had supplied.

This was not a semantic disagreement. I had begun turning my interpretation of the practice into an account of her practice.

That correction forced a distinction I now try to keep visible. Indi's method is hers. My attempts to build deterministic guards, agent roles, review gates, and verb-forward summaries may be inspired by what I learned from her, but inspiration does not make an extension canonical. If I change the reason for a rule, change the sequence, or name the unit differently, I may be building something adjacent.

Adjacent can be valuable.

It has to be honest.

Another exchange sharpened this further. I asked broad questions about the lineage of the method, the completeness of its classifications, who had taught parts of it, and whether there might be a path for practitioners to become teachers and stewards.

Indi first corrected what I had failed to remember. Other people had taught parts of the work. Other practitioners had materially influenced how the method moved from observable tasks toward cognition. She had explained pieces of this before.

My larger questions did not disappear, but they could no longer rest on a false account of what she had already said.

This is part of why I describe her as generous. She has been increasingly patient with my pokes and prods, but patience does not mean approving the frame of every question. She protects factual and methodological accuracy before joining me in speculation.

That has improved my questions.

It has also taught me something about mentorship. A mentor does not only enlarge the student's confidence. A mentor gives the student's confidence something solid enough to strike against.

What a practice group can reveal

The Thinking Styles course helped me see the end-to-end shape of the analysis.

A human conversation produces more than topics. A transcript contains explanation, opinion, remembered scenes, emotion, internal judgment, and personal rules tangled together. The analyst has to decide what kind of thing a passage contains before trying to compare it with another passage. Weakness early in that chain cannot be repaired by elegance later.

The practice group made those dependencies visible without turning them into abstractions.

When a participant had not offered much cognition related to the purpose, there was less to analyze. The answer was not to manufacture density. When a summary had lost the participant's reasoning, the group had to return closer to the source. When several people performed the same action for different reasons, behavior could not hold them together. When two people made opposite choices through the same interior concern, the visible act could not keep them apart.

One exchange near the end gave me a test I still use. If the cards were rewritten only as what each person did, would the group break apart? If so, the action was not the binder. The shared reasoning was.

This is demanding qualitative work because the analyst has to resist several kinds of convenience at once.

Topics are convenient. Sequence is convenient. Demographics are convenient. A polished phrase is convenient. The first group that seems coherent is convenient.

The person is not obligated to be convenient.

Indi's work is designed around that refusal.

It does not eliminate researcher bias. No honest method can promise that. It makes some forms of bias harder to perform invisibly. The analyst has to expose why a card moved, why a group coheres, why a person belongs, and why an attractive interpretation was rejected.

The practice group did not make any of us neutral.

It made every one of us answerable.

The question I bring from enterprise

My gratitude for Indi's work does not a repetition of her career.

Her method did not emerge from a career spent inside an organization of twenty thousand people. Mine did.

I have watched ownership fragment across business units. I have watched platforms become enterprise standards before researchers believed they served the work. I have seen ten years of research data become simultaneously valuable and difficult to govern. I have worked with compliance, IT, engineering, product teams, vendors, and leaders whose time horizons rarely align.

Those conditions produce questions the method alone does not answer.

What happens when an enterprise carries many germinal questions at once? Who owns each resulting model after a reorganization? How does an agent know which model answers the present question? What happens when two models overlap, or when one question grows from another? How does a repository preserve not only the finding but the limit on the finding?

Indi has confirmed that most germinal questions produce separate skylines, while some overlap and some run in sequence. That gives me a methodological boundary. It does not yet give an enterprise an operating system.

That is where I may have something to offer.

My contribution is not a better way to listen than Indi. It is experience with the machinery that surrounds listening after the session ends: repositories, permissions, provenance, temporal history, data models, retrieval systems, organizational ownership, and now AI agents capable of producing fluent answers from several studies at once.

The enterprise problem is not merely scale in the sense of more transcripts.

It is scale of responsibility.

A model can be perfectly preserved and still become dangerous if nobody remembers which decision it was built to inform. An agent can cite every source and still answer the wrong question. A business unit can fund good research and leave the enterprise with an orphaned model when the sponsor moves on.

The deeper I move into Indi's method, the less satisfied I become with enterprise systems that preserve topics while discarding purposes.

The question has to travel with the answer, especially when a person or an agent prompts a repository for insights.

Building agents that do not erase the method

My work with agents began as an efficiency question.

Could a system read transcripts, identify interior cognition, produce useful summaries, and help researchers group them? Could it reduce the mechanical labor without reducing the evidence?

The practice group changed the architecture.

The system I now imagine separates machine capability from responsibility. Deterministic layers hold the board state, evidence links, versions, permissions, and movement history. Probabilistic agents propose interpretations, challenge groups, speak from bounded participant evidence, and look for unsupported inference. Human researchers decide what enters the model and remain responsible for where it applies.

The AI system can render cards into Miro. It can preserve every revision. It can notice that a label changed and invite a bounded recomparison. It can hold deleted groups as negative cases instead of cleaning them from history. It can continue comparing evidence after my attention has tired.

It cannot become the person who sat with the participant.

It cannot bear the relationship created by having listened.

This is why I think of the researcher's role as progressive agency. At each stage, AI should increase the researcher's capacity to notice, compare, question, and decide. The human cannot be reduced to one ceremonial approval gate at the end of a machine-run process.

A useful system does not automate the researcher out of the tedious parts and accidentally automate them out of understanding.

It removes board maintenance while preserving knowledge work. It makes interpretations visible before making them efficient. It keeps the source close enough that a confident category can still be challenged by the person inside the transcript.

That is not DStL as Indi teaches it.

It is my attempt to build technical infrastructure worthy of what her method protects: the integrity of another person's reasoning against the quiet return of our own bias.

Indebted and independent

I am indebted to Indi in several ways.

For her authorship of Mental Models in 2008.  For her invitation in 2025, to the years of practice she made available for me to study, to the precision of her corrections, and to the unusual transparency of being able to ask difficult questions in a group where the answers remain visible.

I am indebted to the discipline itself. It changed how I understand qualitative analysis. It made me less willing to accept a theme simply because it sounds sensible. It gave me a way to explain why the same behavior can contain different human worlds, and why different behaviors can emerge from the same interior concern.

Indebtedness does not erase independence. It makes the boundary more important.

I have an obligation to distinguish what belongs to Indi from what belongs to me. I have to challenge without caricaturing, extend without absorbing, and credit without implying endorsement.

I have offered critiques of the method. I have pushed on its classifications, its teaching model, its certification path, its relationship to other research traditions, and its readiness for enterprise and agentic systems. Some of those questions remain open. Some were weakened by my incomplete understanding. The asymmetry matters: I am still learning a craft in which Indi has decades of mastery. I have also begun to shape work that is genuinely mine.

That is what a living relationship to a body of work should permit.

The world needs more opportunities to encounter Indi's work. DStL deserves practitioners who can explain why it matters, researchers who can test it carefully, and builders who can carry its discipline into new systems without turning it into a prompt template.

Evangelism without critique becomes branding.

Critique without gratitude becomes extraction.

I want neither.

I want stewardship strong enough to include friction. For me, that is part of a resilient life: gratitude that can survive disagreement.

The work I did not leave

I return to that practice session when Indi asked why I was quiet.

I was listening through an earpiece. My camera was off. Elder care had interrupted the professional surface of the meeting, but not the work itself.

On the whiteboard, people were trying to represent lives that were not their own. In the room around me, I was responsible for a life that could not be reduced to an enterprise project plan. Through my headphones, Indi was asking us to slow down, explain why two things belonged together, and remain willing to move them apart.

Caregiving had made me afraid that I was leaving a thriving career at exactly the wrong moment.

Instead, it stripped away a distinction I had carried too rigidly between serious work and the intervals spent waiting for life to need me. Those intervals were not outside the work. They were where another kind of attention became possible.

The waiting became study. The interruptions became part of how I understood attention. The regret did not vanish, but it stopped being the only interpretation available.

I still consult with enterprises. I still build. I am working more deeply in qualitative AI than I ever have. The difference is that I no longer measure that work only by whether an organization has given me a role large enough to contain it.

Indi's invitation gave this period a methodological home in DStL, the name she coined for the practice.

Her practice gave me a discipline to learn, a boundary to push against, and a standard I cannot meet through fluent language alone.

I thought I had made a reckless mistake by walking away from enterprise work.

What I had walked away from was one version of where the work was allowed to happen.

The work was still there.

I had to become quiet enough to hear it. Waiting. Listening deeply.

Thank you, Indi, for helping me hear it. And thank you to the enterprise work I thought I had left. It waited for me to return differently.