PSI - Issue 77

Hugo Mesquita Vasconcelos et al. / Procedia Structural Integrity 77 (2026) 601–610 Hugo Mesquita Vasconcelos/ Structural Integrity Procedia 00 (2026) 000–000

602

2

Nomenclature AI

Artificial Intelligence

AIS Automatic Identification System MHC Model of Hierarchical Complexity OHC Order of Hierarchical Complexity ONC Ocean Networks Canada

1. Introduction In an increasingly AI-dominated world, where increasingly complex tasks are solved by ever more powerful machines, the focus is shifting from pure accuracy to efficiency. While AI models can achieve remarkable accuracy, the energy consumption behind these achievements is becoming a growing societal concern, as stated by International Energy Agency (2025). According to OECD (2025), Portugal’s Exclusive Economic Zone, the twentieth largest in the world, underscores the importance of efficient, low-power systems that can be deployed at sea for active long-term security monitoring. To address this, it is proposed the use of the Model of Hierarchical Complexity (MHC), introduced by Commons (2007), a neo-Piagetian framework that structures learning into discrete stages or orders of complexity. The MHC has been applied in various domains, as presented by Commons et al. (2014), showing that hierarchical learning can accelerate progress and improve efficiency in learning. In this work, it is tested whether MHC can enhance the efficiency of AI training for maritime acoustic signal recognition. For that, it was first necessary to define the specific Order of Hierarchical Complexity for the domain. Three approaches were compared: a traditional unstructured AI model training, a two-stage MHC-guided model, and a full three-stage MHC-guided model. By keeping a common AI backbone and constant hyperparameters, it was possible to evaluate the speed of training and the achieved results. The hypothesis is that MHC-driven training can yield more efficient learning, achieving comparable results faster and with less energy consumption. The Model of Hierarchical Complexity (MHC) was introduced by Commons et al. (1998) in the 90s, and, as described in Commons (2007), it is a “framework for scoring reasoning stages in any domain as well as in any cross cultural setting.”. It is called a neo-Piagetian model as it uses the idea of stages developed by Piaget and builds upon it. The MHC is a quantitative behavioural developmental theory that posits that task sequences create hierarchies that get more complex instead of explaining behaviour variation over age due to mental structures or schema formation. This explains the developmental changes observed since less complicated tasks must be completed, practised, and mastered before learning more complex tasks, Commons and Chen (2014). As Michael Commons says, “Tasks are defined as sequences of contingencies, each presenting stimuli and requiring behaviours that must occur in some non-arbitrary fashion”, Commons (2007). The MHC defines 16 orders of hierarchical complexity (OHC), and due to the previously explained premises, it follows a tree-like structure. It is developed to approach personal development throughout a lifetime. Humans, their biological systems, non-human animals, and machines, including computers, are examples of entities that arrange information. Its vast applicability comes from a direct mathematical approach to defining tasks, and tasks may contain any information. As a result, it can be used in any situation because it just uses quantitative concepts and makes it possible to perform a standard quantitative study of hierarchical complexity in every situation, Commons (2007). The MHC has been applied to vast areas and shown to account for performance in a variety of different domains, as demonstrated by Duran et al. (2018) or Commons et al. (2014). Perhaps one of the more straightforward examples is mathematics and the complexity of the distributive law versus simply adding or multiplying. The task of a+(b+c) is 2. Methodology 2.1. Model of Hierarchical Complexity