“A moderator may govern the room, but an interviewer must earn the soul of the encounter. The moment we replace human attention with synthetic attention, we do not merely change the tool—we change the truth that is able to emerge.”

— Aditya Mohan, Founder, CEO & Philosopher-Scientist, Robometrics® Machines

The distinction must be made with surgical clarity: a moderator stands among three parties and is judged partly by neutrality, tempo, and procedural balance; an interviewer stands opposite a single human being and is judged by something far more intimate—the quality of attention. The difference matters from the very beginning, because interviewing asks something more intimate and more exacting of the human encounter. In moderation, one may reasonably prize distance. In interviewing, distance is often a defect. The interview is not merely a mechanism for extracting answers from a candidate, witness, artist, founder, patient, or citizen. It is a reciprocal scene in which one person says, in effect, I am willing to spend a fragment of my life understanding you, and the other answers, Then I will risk candor. That exchange is not administrative. It is relational. It is not a queue being processed. It is a human test of seriousness.

The Value of the Interviewed Person’s Time

For that reason, the value of the interviewed person’s time does not fall when an AI conducts the interview; it rises, because the interviewee is often forced to do more of the hidden labor himself—over-explaining tone, compensating for reduced feedback, and second-guessing whether a pause, qualification, or moment of ambiguity will be understood as intended. The old managerial fantasy says the opposite: if the interviewer is cheaper, then the encounter itself must also be cheaper. But that confuses employer cost with human value. To speak carefully about one’s work, motives, failures, judgment, and hopes into a synthetic listener is often more effortful, more estranging, and more cognitively burdensome than speaking to a person. There is more friction, not less: friction of uncertainty, friction of dignity, friction of wondering whether one is being understood in the round or merely parsed into features. There is also the experiential cost—the faintly humiliating sense of offering one’s biography to an instrument that cannot truly incur the moral weight of receiving it. So the organization may save money, yes; but what it purchases cheaply is not the interview. It merely discounts the interviewer. The candidate, by contrast, pays a premium in adaptation, self-monitoring, and emotional compression.

How People Respond to an AI Interviewer

Human beings also respond differently when the next prompt is generated by a machine, or even when a human script is merely delivered through a machine voice or animated face. The semantic content may be identical, yet the social meaning is not. A question spoken by a human interviewer carries biography, accountability, and situated intent; the same words spoken by an AI system arrive stripped of much of that anchoring context. People still answer, of course. They may even answer fluently. But fluency is not the same as disclosure, and disclosure is not the same as trust. A human interviewer can signal that a difficult answer has been understood, can soften or sharpen a line of inquiry with tact, can reveal curiosity without menace, can alter cadence to make room for hesitation, and can communicate—before a single clarifying question is asked—whether the conversation is safe, adversarial, ceremonial, bored, generous, hurried, or alive. In practice, the interviewee is never replying to words alone; he is replying to the whole social surface on which those words are laid.

The Science of the Encounter

The science matters, and it must be stated carefully, because findings across research on social presence, eye contact, rapport, and interview performance in mediated settings all point in the same direction: once the channel changes, the human encounter changes with it. There is no honest fixed number by which one may say that speech is only some small fraction of an interview and everything else is the remainder; that popular formula has been stretched far beyond what the underlying research supports. The stronger claim is subtler and truer: interviews are multimodal encounters. Eye contact regulates turn-taking and shared attention. Facial expression and vocal prosody help people infer intention, confidence, warmth, uncertainty, and sincerity. Gesture, posture, latency before response, tiny repairs in timing, and the rhythm of mutual adjustment all contribute to rapport and social presence. Even where the spoken answer remains the same, changes in eye contact, perceived presence, and impression management alter how people perform and how they are evaluated. The interview, then, is not a transcript with decorations attached. It is an organism of coordinated signals. Remove, flatten, or counterfeit too many of those signals and you do not merely digitize the interview; you change its substance.

Why This Differs from Moderation

Here the difference from moderation becomes decisive. In a moderated exchange, the moderator’s ideal is often procedural fairness among others: to keep two or more parties in frame, to distribute time, to prevent domination, to maintain order, and to arbitrate the flow. There, an artificial system may indeed excel at timing, queuing, transcription, summarization, rule enforcement, and even neutrality of turn allocation. That is the point at which this essay reconnects with “AI as a Better Moderator or Arbitrator”: the case for AI is stronger when the task is to manage a multi-party process rather than to meet a single person across a line of consequence. An interview, however, is dyadic. Its essential question is not only what was said? but what passed between these two minds while it was being said? A moderator can succeed by being a well-designed third term. An interviewer cannot be reduced so easily, because the interviewer is not merely governing the channel; the interviewer is half of the relationship. Martin Buber, writing long before machines learned to simulate conversation, put the matter with severe economy: “All real living is meeting.” The interview is precisely such a meeting—or it ought to be. Once the other party is converted from a present intelligence into a synthetic proxy, something central is not streamlined but forfeited.

Recommendation

The recommendation, then, is against AI interviews as a primary evaluative practice, and against them on four grounds at once: value, science, ethics, and results. On value, they cheapen the organization’s side while asking the interviewee to bear greater cognitive and emotional cost. On science, they degrade or distort the interpersonal signals from which trust, nuance, and judgment are partly formed. On ethics, they risk turning a consequential human assessment into an interaction in which responsibility is diffused, appeal is murky, and dignity is subordinated to throughput. On results, they invite a dangerous category mistake: confusing measurable consistency with genuine understanding. AI may be useful around the interview—for scheduling, note-taking, transcription, accessibility support, and structured comparison after the fact. But the act of interviewing itself should remain stubbornly human. A civilization may automate many things with honor. It should hesitate before automating the moment in which one person asks another, with consequence, who he is.