Interview integrity / Axiom Cortex
A change is a reason to check.
The documented integrity approach requires an opening experience question that provides a usable baseline. Later changes can prompt source review of possible read or assisted answers, including AI-provided responses. They do not prove AI use. Pauses, accents, slower speech, disability, or second-language communication alone are not evidence of deception.
We use science to align tomorrow’s IT talent with the work ahead.
Start with the candidate’s own experience.
The baseline is an open work-experience question asked before harder technical prompts. It gives the candidate room to explain familiar work and creates a source-grounded comparison sample.
Before the technical questions, tell us about your path into engineering, the work you have done, and why this role fits your experience.
If that answer is absent, interrupted, heavily prompted, or not preserved, do not make a baseline-conditioned integrity flag.
Why use an experience question first?
TeamStation AI treats it as a possible low-pressure primer before technical questions. This is a company-reported design observation that needs separate validation. It does not measure calm, personality, or a private mental state.
Check the source and the context.
- Confirm a usable baseline.
- Locate the later source segment.
- Check other explanations.
- Ask a relevant follow-up.
Later changes may prompt review of a possible read or externally assisted answer, including an AI-provided response. The flag is not proof of AI use and must not determine candidate eligibility.
- Preserve the recording, timing, speaker turns, and full answer.
- Check whether the question became harder or less familiar.
- Account for recording problems, interruption, translation, and code-switching.
- Ask the candidate to explain a mechanism, tradeoff, or concrete example from the answer.
Give thinking time its proper context.
Pauses and slower speech can reflect thinking, language choice, or the interview setting. Latin American candidates and other L2 or ESL speakers should not be judged against an assumed native-English pace.
Accent, hesitation, disfluency, translation, disability, polished wording, and slower delivery alone do not establish weak technical reasoning or hidden AI use. Review the substance and what the role actually requires.
Useful follow-up questions
- What changed your first approach?
- Which assumption matters most in that design?
- What did you personally implement?
- What would fail first under a different constraint?
Know what a flag can support.
The public claim is a baseline-conditioned reason for source review. Public accuracy and false-positive results for detecting AI-read answers are not established by this website. A plain or cleaned transcript can’t reconstruct how the candidate delivered an answer. Messy audio can hide context, so check the source info.