An institute evaluates a tool to deduce emotional states from students. I propose not to transform an automatic interpretation into a certain description of the experience and to subject the use to specific regulatory and professional verification.
In the example, observable behavior may have different explanations. To know a difficulty you need appropriate listening.
The educational proposal is to create channels of dialogue and support, not to classify people through opaque indicators. Technology does not replace the context nor the responsibility of those who intervene.