Selection

IHow They Made Us Feel

The candidate leaves the room, and we begin describing how they made us feel.

We were excited. We were unconvinced. The conversation flowed, or it dragged. We could imagine working with her. Something about him gave us pause. These are impressions of an encounter, and some of them may tell us something useful. But listen to what happens when we turn them into an assessment.

I felt confident talking to her becomes she has potential. I had to work harder to understand him becomes he lacks presence. The evaluator disappears from the sentence. How we felt about the candidate becomes something we claim to have established about their ability.

That change is easy to miss because the feeling is real. Nobody has to invent their enthusiasm. Nobody has to pretend the conversation was uncomfortable. We can report our experience honestly and still misunderstand what it tells us about someone else.

I work in AI. I sit in the rooms where hiring gets decided. I know the pleasure of meeting someone whose competence feels immediately recognizable. What troubles me is how easily that pleasure can become its own justification. We liked the candidate, and the language of evaluation gives us increasingly respectable ways to say so.

By the time we call it a bet, we may already have decided whose uncertainty interests us and whose uncertainty makes us nervous. One candidate's missing experience leaves room for growth. Another's leaves a hole in the application. Both require imagination. Only one receives it.

My concern begins there: with the permission to become someone we cannot yet demonstrate that we are. A hiring panel controls a small piece of that permission. It can give a person work, income, instruction, colleagues, and a credential that makes the next opportunity easier to obtain. Those are conditions in which ability can develop and become visible.

When we allocate them partly according to how people make us feel, we do more than reward the people we already understand. We help produce the evidence that will make our understanding look correct.

IIThe Interview Beneath the Interview

Here is the interview process I recognize. The calendar invitation tells you the format. It rarely tells you what would make the person on the other side comfortable saying yes. Each stage asks for a different kind of reassurance.

Testing whether someone can explain a decision or respond to criticism is legitimate evaluation; treating the reassurance we feel afterward as sufficient evidence of their ability is the mistake.

The résumé has to make you worth meeting. A record of competence matters, but a stranger has to find something in it that makes them curious. You can be good at the work and still give someone very little reason to keep reading. This is an unpleasant distinction to discover after spending years getting good at the work.

The recruiter needs to believe you want this particular job. You may be exploring several companies. You may need money. You may not yet know enough about the job to want it. The company is still deciding whether it wants you, but you can be expected to sound as though you have already made up your mind. Your enthusiasm may be entirely sincere. The person across from you can only judge its performance. You have to sound sincere on command.

The hiring leader is trying to imagine working with you. Can I rely on your judgment? Will you make difficult work more manageable? If I advocate for you, can I explain why? Your experience gives them something to check that picture against. It helps them believe the working relationship they are beginning to imagine could actually happen.

The case study or live session lets the panel try out that picture. They can examine your reasoning, your preparation, and how you respond when challenged. They also find out what it feels like to disagree with you, wait for you, or need something from you. A technically strong answer can still leave people uncomfortable. Did you dismiss a reasonable objection? Or did you disagree in a way the group isn't used to? The panel has to work out which discomforts matter to the job and which belong to the people watching.

This map is useful if you are preparing for an interview. It also reveals how much extra work the candidate is doing. You have to demonstrate an ability and help the evaluator feel comfortable recognizing it. Some people have been practicing the second job for years without knowing it had a name. It is easier when the people assessing you resemble the people you already know how to speak to.

And luck gets a vote at every stage. Which interviewer you get, who else applies, what the team currently needs, whether the role survives long enough for an offer: any of these can change without a change in your ability. Understanding the social rules helps you prepare. It does not put you in charge.

IIIThe Feeling Has a History

Lauren Rivera's research on elite investment banks, law firms, and consultancies gives this concern a concrete basis. In her study of cultural matching, evaluators favored applicants who shared familiar interests, experiences, and ways of presenting themselves. Similarity could matter more than anticipated productivity. Her later paper, Go with Your Gut, examined the emotional process directly: interviewers used their excitement and enthusiasm toward applicants to evaluate them and make hiring decisions.

These findings concern particular elite firms. They do not establish that every interview works this way, or that an evaluator's enthusiasm is always misplaced.

One banker in Rivera's study described the inference with unusual clarity: “You just hit it off with them. And you feel like they can hit it off with anybody.” The connection with one interviewer had become evidence of a general ability. Anybody is doing a considerable amount of work in that sentence.

Even a written rule has to pass through someone's expectations. A group develops habits: what authority sounds like, what counts as an interesting question, how disagreement should feel. The applicant arrives after those habits have become normal. They can be invited to join while being asked, in a dozen small ways, to reassure everyone that the room will still feel the same.

A group can change its membership while protecting what feels normal. Calling the process fair does not tell us whether we have changed that habit.

Conversation feels easy when little needs explaining. A familiar ambition makes sense without much context. A shared interest gives both people something to contribute. An unfamiliar life can require more questions before its achievements become intelligible. An interviewer may experience the first encounter as evidence of connection and the second as evidence that connection is missing, without noticing the unequal work required to reach it.

That is where taste acquires institutional power. The evaluator's own background supplies a reference point, and the candidate gets assessed by their distance from it. The interviewer is allowed to be a particular person with particular interests. The applicant is expected to demonstrate a general quality called fit.

Being easy for me to recognize is a considerable advantage when I am the person deciding whether you are promising.

Class enters through that advantage without needing to announce itself. Access to travel, leisure, professional adults, and practice speaking to people in authority can give someone experiences an evaluator recognizes and a style the evaluator trusts. The candidate need not be pretending. They may be entirely sincere. The inequality can lie in how much of one person's ordinary life already satisfies another person's expectations.

I was the kid who believed the grades were the game. I assumed effort was legible and that legibility would be rewarded. There is a particular humiliation in discovering that you prepared for the stated test while someone else arrived already fluent in the encounter around it.

It would be comforting to make this an argument about stupid interviewers and charming frauds. That would also let the more serious problem escape. A candidate can be capable, hardworking, and likable, and still benefit from an evaluator's willingness to interpret everything generously. Their advantage does not require their accomplishments to be false.

Imagine two candidates who both have something important to learn. The panel describes one as inexperienced but impressive, the other as insufficiently prepared. There may be excellent reasons for the distinction. The question is whether the panel can identify them independently of which candidate it enjoyed meeting. If it cannot, its confidence has exceeded its evidence.

The benefit of the doubt has material value. It determines who can arrive unfinished and still be welcomed.

IVWhat the Offer Makes Possible

A person who receives the offer can acquire more than the experience listed in the job description. They can learn what good work looks like from people who already do it. They can make mistakes that a manager helps them repair. They can accumulate references, relationships, and achievements with an institution's name behind them.

Those conditions can develop the abilities the interview tried to assess. That is part of what a good job is for.

If the candidate succeeds, the panel may have recognized an ability that was already there. It may also have supplied conditions the person needed to develop it.

We do not get to observe both lives. We do get to observe the successful hire. That makes it tempting to treat the person they became as proof of the person we knew them to be.

The rejected applicant may flourish elsewhere. A rejection is neither a permanent economic sentence nor a reliable judgment of human worth. But the chance to try elsewhere is not distributed equally. Someone with savings, housing, and people willing to introduce them can afford a longer search. Someone who needs an income immediately may have to take work that leaves little time to prepare for another route. The same decision can cost two people very different amounts.

A rejection does not have to ruin a life for the pattern to matter. Give some people more chances to develop, then point to their development as proof that they deserved those chances, and an unequal process begins to justify itself. Someone escaping the pattern does not make the pattern fair.

The claim gets bigger when we stop talking about a particular job and start talking about the kinds of people who deserve to succeed. She did excellent work here is something we can check. We only hire the best asks us to accept a ranking of people the institution never employed and may barely have understood.

That hierarchy is flattering to those of us inside it. It allows our work to count as evidence of our ability while the help that made the work possible recedes from view. We remember the late nights. It takes more effort to remember that someone let us spend those nights on work that would count.

Some of the people living with these consequences are friends of mine. I am lucky to be where I am. I cannot honestly treat my position as proof that I worked harder than they did or deserved more. I wish knowing that gave me more power to help them than it does.

I am sorry for how little I can change for them. That is an uncomfortable thing to admit from a position I am glad to have. I am still responsible for the decisions I can influence: what I count as evidence, whom I am willing to teach, and whose uncertainty I am prepared to take seriously.

VWho Gets to Be Inexperienced

In AI, this gets especially uncomfortable. I help bring technology to market that can change both how people do their work and how many people an employer needs to do it. The industry is also hiring the people who will influence what gets built, whose problems deserve attention, and which forms of work remain worth paying humans to learn.

The evidence already resists a simple story. In a study of customer-support agents, AI assistance increased issues resolved per hour by about 15 percent on average, with larger gains among less experienced and lower-skilled workers. It also provided evidence of learning. A technology capable of sharing expertise this way could make a beginner easier to employ and support.

There are also reasons to worry about access to the job itself. In the payroll data examined by Stanford's Canaries in the Coal Mine?, the researchers report a widening relative employment gap between 22–25-year-olds in occupations with greater and lesser exposure to AI, driven primarily by reduced hiring. The researchers explicitly describe the findings as descriptive evidence rather than a causal estimate of AI's effect. They find no evidence of widespread displacement across the economy.

Those results leave important questions open. A tool can help a beginner learn while an employer uses it to reduce the number of beginners it hires. Another employer could use the same gain to expand its team. Whether new capability becomes new opportunity depends partly on decisions about how work is organized.

An entry-level role gives someone permission to contribute before they can contribute as much as an experienced colleague. If companies remove that period of supported inexperience, they have to explain where future experience is supposed to come from. A promise that new occupations will eventually exist does not tell a person how they will qualify for one.

This returns us to the debrief. There is a risk that as some tasks become easier to automate, employers will place greater weight on broad judgments about adaptability, taste, or potential. Those qualities can matter. But unless an employer specifies how it will recognize them in the work, the words leave considerable room for an evaluator's own preferences.

If fewer people receive a chance to learn, the distribution of that room for judgment matters more. The candidate whose gaps inspire curiosity gets a future colleague willing to teach. The candidate whose gaps provoke discomfort gets a reason to be rejected. We should be able to explain the difference in terms that survive the admission that we liked one of them more.

The companies changing how people enter work have a responsibility to examine whom they let in. We cannot take credit for expanding human capability and then refuse to explain who gets the chance to develop it inside our own institutions.

VIThe Evaluator Belongs in the Sentence

I want a hiring process in which feeling has to answer for the authority it is given. Asking everyone to mean well leaves the hard part untouched. We can change what interviewers are asked to notice, record, and justify. Research comparing structured and unstructured interviews gives us a reason to do so: structure can improve the consistency and predictive value of evaluations.

Feeling at ease with someone can tell us something useful. Perhaps they listened carefully, handled a disagreement well, or made a difficult idea clear. Say what they did. Describe the situation in which they did it. Explain why the job needs it. There is more to work with in that account than in the pleasant feeling it left behind.

This is more uncomfortable than agreeing that everyone has biases. The evaluator has to stay in the sentence. I found this answer difficult to follow gives us something to examine: the answer, the question, my familiarity with the subject, and the demands of the role. He is not a clear thinker closes the case before we have done that work.

The same discipline applies to praise. If I call someone high-potential, I should be able to say what I saw, what they still need to learn, and why I believe they can learn it. If another candidate has comparable gaps and I regard those gaps as disqualifying, I owe an explanation. My enthusiasm for the first candidate does not supply one.

Start before the interview. Decide which parts of the work a candidate needs to demonstrate, and, where possible, give candidates comparable, bounded chances to demonstrate them. Pay for substantial work. Afterward, have each interviewer write down what happened before the panel starts comparing impressions. If someone says fit, ask which behavior they mean and why the job needs it. None of this gets rid of judgment. It gives the rest of the room something to check.

And an institution that says it hires for potential should be able to show what it does with people who need to develop. Who receives training? Who gets feedback? Whose early mistakes remain evidence of inexperience, and whose become confirmation that they never belonged? The generosity promised in the interview has to exist after the offer.

I have three children. A great deal of what I am teaching them is how to present themselves: how to explain an idea, ask for help, make their effort intelligible to someone who can give them a chance. I want them to have those abilities. I want them to meet adults who see something worth encouraging before it is fully formed.

What troubles me is how close that preparation comes to teaching them that their security will depend on becoming the kind of person a stranger feels good about choosing.

I cannot resolve that by withholding the preparation. I can teach them that another person's difficulty winning approval is not evidence of lesser worth. I can refuse to tell them that being chosen makes the choice just. And when I am the stranger deciding, I can ask what I have actually learned about the candidate and what I have learned about my own preferences.

The feeling belongs to the evaluator. The consequences reach the candidate.