What biases a first-round interview, and how to fix it

Ask a hiring manager what swayed a first-round call and you will rarely hear “their accent.” You will hear “strong communicator,” “great energy,” “we just clicked.” Those phrases often cover for something older and less flattering: a reaction to how a person sounds, looks, and carries themselves that lands before the content of their answer ever does.

Most first-round bias comes from signals that feel informative and tell you little about the job. The harder you try to ignore them by willpower, the more they leak in through the side door. The reliable fix is to change what the room measures.

The signals that sway us

Accent and name. In a controlled study of interview judgments, an applicant’s accent and name shifted how favorably they were rated: an applicant with an ethnic-sounding name speaking with an accent was viewed less positively than the same profile without those cues, and those judgments fed straight into the hiring decision (Purkiss et al., 2006). Accent and name were doing work that only the answer should be doing.

Confident delivery. A candidate who speaks in clean, certain sentences reads as capable. Fluency is a separate skill, trained by privilege, practice, and a first language that happens to match the interviewer’s. The nervous candidate who gives the better answer loses to the smooth one who gives a worse one, and everyone in the room calls it a gut feeling.

Shared background and rapport. The interview that “flows” is usually the interview with someone like you: same schools, same hometown, same references. Rapport makes the conversation easy, and interviewers read that ease as signal. Mostly it reflects similarity, the exact thing a fair process is supposed to see past.

The résumé halo. A familiar logo at the top of the page tilts everything that follows. Once you have decided someone from a brand-name company must be good, you hear their answers as confirmation. The halo is a prior you forgot you were holding, and it says nothing about the competency in front of you.

Why unstructured interviews amplify all of it

When every conversation is different (different questions, different order, different follow-ups), there is no shared standard, so each interviewer fills the gap with a private one. Private standards are where accent, rapport, and polish do most of their work. The looser the structure, the more room the noise has to move.

The research is unusually consistent on this. In Schmidt and Hunter’s landmark synthesis of 85 years of selection research, structured interviews predicted job performance far better than unstructured ones, with a validity of .51 versus .38 (Schmidt & Hunter, 1998). More recently, after Sackett and colleagues re-examined decades of that research and corrected a systematic overcorrection in how the numbers had been adjusted, structured interviews came out ranked at the very top of the methods we use to predict performance (Sackett et al., 2022). With the same job and the same candidates, what changes the result is whether the interview holds its shape.

We wrote a practical guide to running that structure across hundreds of candidates without turning hiring into a bottleneck: how to run structured interviews at scale.

What actually predicts performance

The signal is the content of the answer, scored against behavior you can actually observe. Look for a concrete decision and the constraint that drove it, and for a tradeoff the candidate named on their own, including what they gave up. Look for the failure case: what went wrong, and what they did about it. And look for ownership, meaning what they did, distinct from what the team did around them.

Two evaluators can read these moments in the same transcript and land in the same place, which is what makes scores comparable. Impressions of an accent vary from one evaluator to the next. Anchor every rating to one of these moments and the vibe has nowhere to hide. We laid out the full discipline, including what to cite and what to deliberately leave out, in evidence-based scorecards.

Why process design beats bias training

The instinct, when you notice bias, is to train it out of people. Sit everyone down, name the biases, ask them to try harder. It is a popular fix because it asks nothing of the process: no new questions, no rubric, no change to how the decision actually gets made.

The trouble is that the signals we have been describing are implicit. They act before deliberate judgment gets a vote, which is precisely why asking people to deliberate harder does so little. Nobody can will themselves deaf to an accent, so change the instrument: ask every candidate the same questions, score them against the same versioned rubric, and require that every rating cite a specific moment in the transcript. Fairness then becomes a property of the process, independent of how hard any one interviewer tries.

One line should be non-negotiable: no biometrics. How a candidate looks, their tone of voice, their facial expressions and their “energy” should stay out of the score entirely. Judge the content of what someone says, never how they look or sound. A system that reads faces or voices rebuilds the exact bias you were trying to design out.

Where Hure fits

Hure is built on that approach. Every candidate for a role gets the same structured, 8–10 minute voice-to-voice interview in their own language, covering the same versioned rubric the same way. The result is a scorecard where each rating cites the moment in the transcript that earned it, so you can audit the evidence behind every rating.

Hure’s interviewer also cannot do the thing this whole post is about. It never sees a face and never judges an accent; it scores the content of what a candidate said, and nothing else. That limit is the core of the design. A human still makes the final call. The structure makes sure that when they decide, they are reading what the candidate actually said.