Violation vs Evidence

Let's start with definitions:

Violation - An event that AutoProctor detects which is anomalous. For example, if there is no face being detected on the camera feed. Or if a test taker switches to a different tab

Evidence - Evidence of the violation. For example, a photo of the camera feed when no face was detected. Or, a screenshot of the tab the test taker visited when they switched the tab

The Trust Score is calculated on the basis of the type, frequency and duration of violations. The evidence are show for the test administrator to review what happened. For example, maybe the test taker lost power and the room went dark. So, the violation would say that no face was detected. The evidence will show a photo with hardly any light.

Why do some violations have missing evidence?

When a test taker starts the test, AutoProctor assigns a fixed number of URLs where it stores the evidence. For example, let's say this number is 30. So, whenever a violation occurs, the evidence is uploaded onto one of the 30 URLs. If a 31st violation occurs, AutoProctor randomly decides if it needs to overwrite one of these 30 URLs with the evidence of the new violation, or not store the evidence for this violation.

If the former occurs, then the earlier violation doesn't have evidence anymore. If the latter occurs, the newer violation won't have evidence.

Every test has a fixed number of URLs where it can store the evidence. Once all the evidence URLs are used up, AutoProctor only stores the violation, without the supporting evidence.

The Trust Score calculation depends only on the violations, not on the evidence.

Updated on: 09/21/2023