Nonnative Accent Bias Hits the Leadership Half of an Engineer’s Job

Nonnative accent bias lands on the part of an engineer’s job that decides whether they are seen as a leader: one-to-ones, interviews, and trust across teams. Dmitrii Sidorenko describes that split from the inside in a recent article, and the research shows evaluators make the same mistake he warns newcomers against.

Sidorenko separates technical credibility from cultural credibility

Sidorenko moved from Finland to Los Angeles in 2019 to become lead backend developer on a mobile title at an EA studio, after co-founding and running a game studio in Russia as CTO. He writes that his technical standing transferred on day one, while for about six months he could not reliably follow the idioms, accents, and inside jokes of the people around him. The work that filled his calendar, one-to-ones, interviews, and cross-team negotiation, ran almost entirely in that weaker language.

His central warning is that a comprehension gap is easy to misread as a competence gap, especially for someone used to being in control. He closed it by treating language as a work item: logging every phrase he missed, training his ear on regional accents, and asking what jokes meant. He sums up the lesson as “Comprehension comes before control.”

Evaluators show nonnative accent bias where leadership is judged

Sidorenko describes misreading his own comprehension gap. A 2013 study in the Journal of Applied Psychology suggests the people judging a newcomer can make the same misreading, and that they make it specifically about leadership. Laura Huang, Marcia Frideger, and Jone L. Pearce, in Political skill: Explaining the effects of nonnative accent on managerial hiring and entrepreneurial investment decisions, had nonnative and native English speakers give the same scripted answers. The nonnative speakers were less likely to be recommended for a middle-management job.

The mechanism is the useful part. According to the abstract, perceived political skill fully explained the gap, while race, communication skill, and collaborative skill did not. A second study, using start-up pitches from high-tech venture competitions, found nonnative speakers less likely to receive funding, again explained by perceived political skill. The authors argue that political skill is ambiguous and important enough to let this bias appear meritocratic.

Put that next to Sidorenko’s two accounts. The answers in the study were identical, so the gap came from how raters judged the speakers’ political skill, the quality that governs one-to-ones, hiring panels, and cross-team negotiation. That is the cultural account he says starts from zero in a new country, recorded by evaluators as a leadership deficit.

The risk is concentrated in the first months, before the gap closes

Sidorenko’s account also shows that the gap is temporary when the person works at it. He describes six months of difficulty followed by recovered confidence and close friendships on the team. That timeline matters, because many of the formal judgments about a new leader happen inside it: probation reviews, the first calibration cycle, and decisions about who runs the next hiring loop.

The study has limits that leaders should respect. It measured hiring recommendations and funding decisions made from short scripted material, and it did not follow anyone through months on a team. It tells us how first impressions form, and it cannot tell us how large the effect is in a review written after half a year of working together. The safe reading is that early leadership judgments of nonnative speakers deserve extra scrutiny, because that is where the evidence for the bias is strongest and where Sidorenko’s gap is widest.

AI meeting transcripts help listeners and can mislead reviewers

From an AI governance view, many teams now reach for meeting transcription and AI summaries to help everyone follow fast conversations. For someone in Sidorenko’s position, a transcript is a direct aid: an idiom heard once can be read back and looked up the same day, which is close to the running list he kept by hand.

The same tools need care when their output becomes a record about the speaker. Allison Koenecke and colleagues, in Racial disparities in automated speech recognition (PNAS, 2020), tested systems from Amazon, Apple, Google, IBM, and Microsoft on interviews with 42 white and 73 Black speakers. The average word error rate was 0.35 for Black speakers and 0.19 for white speakers. That study concerns US dialects, and it does not measure nonnative accents, so it cannot tell us the error rate for an engineer raised in Russia. It does show that recognition accuracy can differ sharply by speaker group. A summary built on a weaker transcript of one person’s words can make that person sound less clear than they were, and that is the last thing a leadership review should rely on without a check.

Engineering leaders should keep the two accounts separate in reviews

Sidorenko’s fix was personal: he turned comprehension into a work item. Leaders who manage or evaluate someone in his position can apply the same principle to the process around them. My proposals are three practices, drawn from the evidence above and not from any company’s policy.

First, when a review or calibration note raises a leadership concern about a nonnative speaker, ask the evaluator to name the specific behavior and its effect, such as a missed commitment or a decision a team did not understand. A vague sense that someone lacks presence is the kind of judgment the 2013 study found most open to bias. Second, avoid settling leadership judgments during a new hire’s first months in a new culture, and say so to the person, so that a slow start reads as onboarding. Third, turn the private phrase list into a shared team glossary of idioms, acronyms, and internal references. Sidorenko reports that colleagues thought better of him for asking; a glossary makes asking cheaper for the next person.

There is a counter-case. A team could respond by banning idioms and requiring plain speech. Sidorenko’s own ending argues against that: the colleague he found hardest to follow spoke the way he himself speaks in Russian, and that vivid style was part of why rooms listened. The goal is to give newcomers the vocabulary and the time to learn it, and to stop early evaluations from recording the gap as a lack of skill.

Why This Matters for Engineering Leadership

Engineering organizations hire across borders for technical depth, then promote on qualities judged through language and culture. Sidorenko shows how a capable leader can doubt themselves in that gap, and Huang, Frideger, and Pearce show that evaluators can record the same gap as weak political skill. Leaders control the review templates, the timing of judgments, and the records that feed them, including AI transcripts. Adjusting those three things costs little, and it protects the leadership pipeline from a bias that looks like merit.

FAQ