The application of computational approaches in behavioral and cognitive science has advanced our understanding of the biological mechanisms involved in learning and decision-making. However, these advances have not been effectively translated into clinical applications, partly due to the limitations of small-scale, lab-based experiments in capturing the complexity of the brain and its interaction with the environment. The widespread use of smartphones presents an opportunity to overcome the limitations of these lab-based experiments. This thesis aims to develop and validate smartphone-based methodologies that can provide richer data sets and larger samples for studying human decision-making and facilitating clinical translation. Several case studies were conducted to demonstrate the feasibility of these methodologies. The first step involved moving experimental settings from laboratories to naturalistic settings using lab-in-the-field methodology. Two case studies were conducted: one involv- ing individuals with chronic pain and another involving patients with Parkinson’s Disease. These studies revealed correspondences between altered decision-making performance and clinical variables of interest, showcasing the potential of testing decision-making pheno- types outside the laboratory. To capture momentary phenotypes in different contexts and time points, the Ecological Momentary Assessment (EMA) methodology was employed. A proof-of-concept EMA study assessed momentary tinnitus experience and its impact on decision-making using a mobile app, demonstrating the possibility of dense sampling of human decision-making in the daily contexts. Ethical and legal constraints on large- scale human phenotyping were also addressed. The feasibility of training computational models of decision-making using distributed learning strategies was examined using a "many-labs" data pool. Federated Learning offered an alternative when large-scale private data collection using smartphones is restricted. In summary, this thesis explored smartphone-based methodologies to develop robust neurocognitive models of mental health conditions, implemented through various be- havioral and cognitive studies. These methodologies have the potential to complement traditional lab-based experiments, significantly enhancing our understanding of cognitive mechanisms and expediting clinical translation.