Fuzzy Cognitive Maps for Algorithmic Decision Making in eMental Health

Abstract

Background: eMental health is on the rise to meet growing global mental healthcare needs and so is the need for applications with automatic decision-making capabilities to support clinical practitioners and to personalize eMental health interventions to their users. However, due to unresolved challenges and clinicians’ reluctance to adopt artificial intelligence in mental healthcare, the technology’s current state in mental health is still in the phase of proof-of-concepts rather than providing decision making algorithms that are ready to create clinical impact. Fuzzy cognitive maps (FCMs), a soft computing technique that leverages expert knowledge by combining cognitive maps with fuzzy logic, can be an alternative approach to equipping eMental health with automatic decision-making capabilities. Method: This article provides an introduction to FCMs for developing decision algorithms for eMental health. It demonstrates step by step how FCMs are developed, based on an example from mental health, while discussing challenges and future directions for research. Results: FCMs are a viable option at the disposal of developers and researchers in the field of eMental health that come with their own set of advantages and challenges: (1) they allow the construction of initial mental health decision models without depending on the availability of large amounts of historical client data, (2) they are explainable, and (3) they involve clinicians in the construction process, inherently promoting professionals’ acceptance of the resulting algorithm. Regarding challenges, (4) there is a pronounced trade-off between the simplicity of FCMs and the accurate representation of mental health constructs since their approach offers limited ways for modelling beyond direct causality and for including temporal aspects of mental health problems, and (5) FCM decision models are not usually on par with data-driven black box AI models in terms of classification accuracy. Conclusion: FCMs are an interesting and promising alternative modelling approach for developing decision-making algorithms in eMental health. Their inherent expert-basis, explainability and independence of large clinical data sets to arrive at initial usable decision models makes them especially attractive for the mental health domain.