Risk levels can only be reactive to data that we have the ability to pass through to our models. Predictions are not often 100% accurate. If assignment grades, attendance, and other information is not available to our models, the models will not be able to use that information to assign risk levels
Often, humans are able to use the contextual information that is not able to be put into the models in order to further improve on the raw model output. As such, risk levels are meant as an initial indicator that provides a starting point to prioritize and find students that may need additional support, to prioritize outreach and resources. Then, human level support can be utilized to better understand the whole student. As is often the case, technology does not fully replace human knowledge, but can be used as a supplementary tool to increase the effectiveness of a strong support staff.