Refining the ML/DL Argument for the SensorAble Project

Abstract

Is Machine Learning/Deep Learning (ML/DL) a technological necessity when implementing SensorAble or is it something to be investigated because of its potential? Should ML/DL be implemented because it permits processing large quantities of multimodal data enabling modelling of autistic neurocognitive processes that well relate to distractibility and anxiety? Or would interventional prototyping using old-fashioned Artificial Intelligence (AI), Bayesian theory or a hand-crafted rule be preferable?Following Participant Public Information (PPI), can ML/DL techniques permit greater understanding of how disruptions occur and properly align/prepare the groundwork for an interventional prototype? Would heuristics, data mining, or perhaps some other statistical approach adequately provide evidence proceeding a design? With the constellation of supervisors who have invested in this project, can fundamental science properly situate SensorAble in a broader vision that creates practical tools? It is one thing to understand and model a problem. It’s another to simply design/build. Doing the latter may inform the user, but how does it guarantee that other stress factors, ethical issues and newly created anomalies aren’t inadvertently introduced?