Risk assessment

Written By Jozef

Last updated 26 days ago

JobMojito developed series of assessments that can trigger warning or issue based on confidence and severity of the issue. Risks are shown on the top of the individual result. By clicking the “Risk assessment” recruiters can add/edit/remove risks as needed from individual results.

Conversation based risks

Risk	What it means	Score impact
Stall Tactics	The candidate is frequently stalling the interview by asking irrelevant questions.	None
Asking for Repetition	The candidate is frequently asking the AI to repeat the question.	None
Asking for Help	The candidate is asking for help from the AI to answer the question.	0.25 possible risk, 1.0 likely issue, per occurrence
Too many failed attempts	May be due to connectivity issues or attempts to game the recruiter.	0.15 likely issue
Multiple profiles with the same name	Multiple profiles have submitted an interview with the same candidate name.	None

Connection based risks

Risk	What it means	Score impact
Anonymous connection	Unable to detect the user's IP address, or they are using Tor services.	None
VPN or Proxy connection	The connection is coming from a VPN or Proxy service.	None

Proctoring based risks

Captured live during the session. The count is how many times the condition occurred, and the duration is the total time it lasted.

Risk	What it means	Score impact
Left the interview tab	The candidate switched away from or hid the interview tab during the session.	None
Camera turned off	The candidate's camera was turned off or covered during the session.	None

Video based risks

Video assessments must be separately enabled. The system provides answer-specific screenshots and segments raw video for API access.

Risk	What it means	Score impact
Reading answers	The candidate could be reading their answers, determined based on eye tracking. Can also identify AI agents running on a separate device.	0.25 likely issue, per occurrence
Candidate not visible	In the recorded video, the candidate's face is not visible (over 30% of the interview duration).	0.25 likely issue, per occurrence
Multiple people detected	Multiple people have been detected in the recorded video.	None
KYC image and camera image mismatch	There is a mismatch between the biometrically verified KYC image and the camera image. Requires the separate biometric package.	0.25 possible risk, 2.0 likely issue, per occurrence

Audio based risks

Audio assessments require separate enablement with guided navigation to flagged answers.

Risk	What it means	Score impact
Multiple voices	Multiple voices have been detected in the recorded audio.	0.25 likely issue, per occurrence
Voiceprint mismatch	The voiceprint differs from the actual voice in the recorded audio, confirming whether the same person completed the interview. Requires the separate biometric package.	0.25 likely issue, per occurrence

Reporting a risk that was not identified

Use this when you can see something in the recording or transcript that the system did not flag, for example the candidate is clearly reading their answers but "Reading answers" was not raised.

Open the interview result.
Click Risk assessment.
- If risks were already flagged, the button sits in the coloured risk bar at the top of the result.
- If nothing was flagged, the button is in the toolbar at the top of the result, next to Quality feedback.
In the dialog, click Edit risks at the bottom.
Find the risk you want to raise. Risks are grouped into three sections: text and connection risks, Analyzed audio, and Analyzed video. The audio and video sections only appear when audio or video recording was enabled for that interview.
Switch the risk on with the toggle.
Set the severity and, where offered, the detail:
- Possible risk (amber) or Likely issue (red).
- Occurrences: how many times it happened. The score reduction scales with this number.
- The resulting Score reduction is shown next to the risk.
Click Save.

The risk now appears in the risk bar like any AI-detected risk, and the adjusted score at the left of the bar is recalculated.

Behind the scenes the risk is stored as a manual addition, which flags the session as a missed detection (a false negative). During the next round of AI training the case is reviewed so that the same pattern is caught automatically in future.

Tip: if the issue is specific to one answer rather than the whole session, open that answer and use Quality feedback there. It includes Candidate cheating risk not raised, which reports the miss against that exact answer.

Removing a risk that did not happen

Use this when a risk was raised but the recording shows it did not actually occur, for example "Multiple people detected" triggered by a reflection or a poster on the wall.

Open the interview result and review the evidence first. In the Risk assessment dialog you can play the flagged answers under Analyzed audio and Analyzed video before deciding.
In the risk bar at the top of the result, find the risk you want to remove.
Click the × button on the right side of that risk.
Confirm when asked ("Remove the ... risk?").

The risk disappears from the bar and the adjusted score is recalculated immediately. If it was the only risk on the result, the risk bar disappears altogether.

The removal is stored as a false positive against that session and is included in the next round of AI training, so the same false alarm is less likely to be raised again.

Individual answers work the same way: risks listed on an answer have their own × button.

Changed your mind? Open Risk assessment, click Edit risks, switch the risk back on and click Save. Restoring a risk you removed also clears the false-positive flag, so it will not be sent for training.