Splicetoday

Digital
Nov 16, 2026, 06:26AM

The Interview Is What Happens Between the Questions

What we need from interviewers is courage.

63e4bde3f732f1ce9879ddac8a34f5a59c barbara walters 1978.rvertical.w570.jpg?ixlib=rails 2.1

Imagine a reporter interviewing a musician about the end of a long career. She asks why he stopped touring. He says the travel become exhausting. It’s a sensible explanation, the kind that fits neatly into a headline. The reporter could move to the next prepared question, or she could notice that he stopped speaking just as the word exhausting left his mouth. Perhaps what happened on the road was more complicated than fatigue. Perhaps he’s already said everything he wishes to say.

The next question can’t be selected intelligently from a list prepared yesterday. The reporter must listen to the answer she actually received. She must decide whether the pause invites a gentle follow-up or deserves to be left alone. The interesting part of the interview begins at the point where the script stops being useful.

This is what worries me about the promise of an automated interview. Artificial intelligence can compose 10 excellent questions about almost any public figure. It can transcribe a recording, summarize an hour of conversation, identify recurring themes, and suggest a polished article structure. Some of those functions are welcome improvements to a demanding craft. But the value of an interview often lies in the moment when the interviewer abandons the question she expected to ask.

The distinction reaches beyond journalism. Managers interview prospective employees. Historians collect testimony. Physicians ask patients to describe symptoms. Teachers try to discover what a student understands. In each setting, the visible structure resembles a sequence of questions and answers. The work underneath involves attention, interpretation, and judgment about what deserves a second look.

A transcript can be accurate down to the syllable and still conceal the real story of how an answer arrived. It doesn’t necessarily explain why one question made a person suddenly careful, why another produced relief, or why a seemingly routine detail caused the questioner to change direction. Even when audio remains available, a summary that removes hesitation and repetition can turn a difficult conversation into a confident account that neither participant would recognize.

There’s nothing virtuous about listening to long recordings for the fifth time. A tool that reduces the time required to locate a quote can let a reporter spend more time checking the quote's meaning. A searchable archive can make forgotten voices easier to find. The problem starts when the person using the tool believes that a clean summary has completed the act of understanding.

Studs Terkel's Working, published in 1974, remains a useful challenge to that belief. Terkel collected the voices of people whose daily labor seldom became the subject of cultural admiration: people in fields, factories, offices, and service jobs, alongside professionals whose titles sounded grander. His book created an expansive portrait of American work through the accounts of those who performed it. The original publication and its scope are documented by its publisher.

The accomplishment was more than putting a microphone in front of people. Terkel treated lives that might’ve been summarized as occupations as worthy of sustained attention. A person who appears in an economic table as a job category becomes, through a conversation, a speaker with frustrations, habits, contradictions, pleasures, and a memory. A summary of the nation's work could capture many of the same subjects. It’d struggle to reproduce the experience of hearing people speak for themselves.

An interview also has a moral dimension. An interviewer exercises considerable power. She chooses the opening question and the excerpt that becomes a headline. She can press for evidence, soften a misunderstanding, conceal a contradiction, or give an answer a misleading emphasis. A machine can assist with the mechanics of comparing a statement against the recording. The person publishing the piece remains accountable for what the reader is led to believe.

Picture a different interview, this time with a city official discussing a new AI system for public services. The official says it will save staff time. A pre-written follow-up might ask how many hours will be saved. An attentive reporter might ask whose hours, and where the savings will go. Will residents spend less time waiting for an answer? Will employees have more time to correct the mistakes that automated decisions sometimes create? Has anyone measured the results after correcting errors? Those questions turn an attractive talking point into a claim that can be tested.

The reporter didn’t need to memorize an encyclopedic inventory of AI facts. She needed enough command of the subject to hear the gap between a benefit described and a benefit demonstrated. The gap is where a serious interview earns its value.

Experiments show why the lure of assistance is powerful. In a study published in Science, researchers Shakked Noy and Whitney Zhang found that an AI writing assistant reduced the average time taken to complete designated professional writing tasks by 40 percent and raised evaluated output quality by 18 percent. The experiment involved 453 college-educated professionals, and its claims concern those writing tasks rather than the full practice of interviewing. That boundary is worth maintaining. The research is available here.

If a reporter can turn recorded notes into an initial outline in minutes, there’s no reason to mourn the lost hours of copying and pasting. More time could be spent reading background material, checking a public record, or returning to the source with a sharper question. A newsroom could also respond to the savings by demanding twice as many interviews. Then the work becomes faster while the interviews become thinner. Productivity measures don’t answer which outcome occurred.

Psychologists Evan Risko and Sam Gilbert describe cognitive offloading as the use of actions and external resources to reduce mental demands. Their review examines how people decide to shift a task outside their own minds and how those decisions are influenced by their assessment of their own abilities. Those assessments can be imperfect. The review doesn’t show that delegating a task necessarily weakens the person who delegates it. It helps explain why people must keep examining what their tools are doing for them and what remains theirs to do.

An experienced interviewer practices a particular form of this judgment. She can offload the transcription without offloading the capacity to recognize a contradiction. She can ask for background research without surrendering her curiosity. She can use a system to suggest possible questions and still refuse the question that sounds impressive but misses the human being in front of her.

The difference becomes apparent when the answer’s uncomfortable. An interviewee may speak in a way that suggests self-protection rather than candor. The questioner needs to decide whether she has enough evidence to challenge a statement, whether the topic is relevant to the story, and whether a confrontation will produce illumination or merely a performance of aggression. Such decisions require a grasp of context. They also require a willingness to be surprised and, occasionally, to admit that the interviewer was wrong.

Good questions are often born of a specific incompleteness. Someone says a local theater struggled after a neighborhood changed. How did it change? Who stopped coming? What did the audience once do together? A musician says an album represents a return to roots. Which roots, and what had been lost? A publisher says people want shorter books. Which readers, and where is the evidence? These questions sound ordinary when written down. Their timing is the craft.

That timing also makes interviews an exercise in respect. There’s a difference between offering someone a space to speak and processing someone into usable quotations. The latter treats each answer as material to be extracted. The former acknowledges that a person may reveal something whose importance neither participant could’ve predicted beforehand. Respect can include rigorous questioning. It doesn’t require treating every claim as true.

I don’t want to preserve transcription as a rite of suffering. Give reporters the strongest available tools for organization, research, comparison, and checking. When a claim needs verification, use a system to locate the relevant document and then inspect the document. When a recording is long, search for the passage instead of listening repeatedly to unrelated sections.

That human work begins with a small discipline: identify the point at which the prepared framework no longer explains what you’re hearing. It may be a discrepancy in a number, a change in tense, or a sudden shift from concrete experience to polished slogans. It may simply be a detail that doesn’t fit the story the interviewer arrived expecting to tell.

A culture can produce endless statements without becoming better at listening. We’re already surrounded by articulate people, prepared messages, and confident interpretations. What we need from interviewers is the courage to notice when a rehearsed answer fails, the humility to let another person's experience complicate a thesis, and the skill to ask one more question without reducing the speaker to a data point.

The recorder can run for an hour. The transcript can be immaculate. The article can be ready before lunch. The measure of the interview will still be whether anyone in the room was willing to hear something that changed the next question.

—Gleb Tsipursky, PhD, behavioral scientist, CEO of Disaster Avoidance Experts, and author of The Psychology of AI Adoption at Work: From Resistance to Results (Georgetown University Press, 2026).

Discussion

Register or Login to leave a comment