Understanding the analysis
Confidence
Some analysis sections show labels such as Strong Signal, Moderate Signal, or Light Signal. These labels describe the model's output and do not certify that the underlying claim is true.
What the labels mean #
The analysis schema normalizes some confidence values into qualitative signal labels. The workspace displays these labels beside generated interpretation or evidence. A label can help signal that the model presents an item more or less tentatively, but code does not define these strings as calibrated statistical probabilities.
The label applies to generated text in context; it is not a confidence rating assigned by a human reviewer, and it is not a measure of the employee's credibility.
What the labels cannot fix #
A Strong Signal label does not correct a transcription error, prove that a quotation is complete, establish causation, or account for information outside the transcript. Conversely, a Light Signal does not establish that the topic is unimportant.
The model may include no label at all. Empty or missing confidence information should not be interpreted as a hidden score or a neutral conclusion.
Verify each statement #
Open the transcript, check the exact wording and speaker, and review any additional evidence independently. If there is no reliable support, do not rely on the generated interpretation regardless of its label.
Use confidence as a reason to ask better questions, never as a substitute for fact-checking, employee input, or human judgement.
Use this guidance thoughtfully
This page describes the current product workflow; it is not an HR policy or a promise about every deployment. Before acting on information about an employee, verify the source record, consider context the system cannot see, and use your organization's process. If a control behaves differently from this guide, check the current workspace UI and confirm the saved result rather than assuming an action completed.