Traditional engineering treats systems as a collection of isolated components. An engineer might optimize a mechanical linkage for maximum stiffness, or a computer scientist might tune a machine learning model for minimum loss on a static dataset. However, as technologies become more autonomous and deeply integrated into critical infrastructure, this siloed approach breaks down. An AI model that performs flawlessly in a clean laboratory environment can fail catastrophically when deployed in a chaotic hospital ward, a power grid, or an autonomous vessel.
To prevent these failures, a modern research direction shifts the focus from individual components to the co-optimization of three interconnected pillars: the human agent, the technological medium, and the operational setting. This approach treats complex deployments not as isolated machines, but as integrated socio-technical systems.
The Tripartite Framework
The core mechanism of this design philosophy relies on a tripartite collaborative framework. Instead of treating the environment or the human operator as unpredictable external variables, engineers explicitly model them as active, interdependent subsystems.
As Xuehua Sun et al. demonstrate in Analysis on the Man-Machine-Environment Collaborative Teaching Method for Mining Engineering Major, organizing a complex domain into these three interacting components allows for the optimization of knowledge transfer and operational safety in high-risk settings. By mapping out how the human agent interacts with the technological medium within a specific operational setting, systemic failure modes can be anticipated before deployment.
This integrated perspective is particularly vital as automation enters environments with high spatial complexity and strict regulatory constraints. For example, in the maritime industry, J.Y. Kim et al. note in Trend Research on Maritime Autonomous Surface Ships (MASSs) Based on Shipboard Electronics: Focusing on Text Mining and Network Analysis that developing autonomous ships requires balancing algorithmic navigation with physical sensor limitations and environmental laws. The machine cannot be engineered independently of the legal and physical ecosystem it inhabits.
Managing the System of Systems
When multiple autonomous entities interact within an environment, the engineering challenge scales from a single machine into a "System of Systems" (SoS). In an SoS architecture, the primary goal is to maintain the autonomy of individual components while ensuring their interoperability.
To coordinate these large-scale, interconnected entities, Tomoki Hamagami explains in Preface to the Special Issue on “Systems Innovation by Machine Learning” how machine learning can be utilized to manage decentralized coordination, allowing independent systems to function as a cohesive whole.
[ Human Agent ] <=======> [ Technological Medium ]
^ ^
| |
v v
[ Operational Environment ]
This framework helps bridge the notorious "translation gap" in fields like digital health. Many highly accurate clinical algorithms never reach practical deployment because they are designed in isolation from hospital workflows. To resolve this, Azadeh Assadi et al. propose a structured framework in An integration engineering framework for machine learning in healthcare that treats machine learning models as interdependent components within a larger network of technical, human, and environmental factors. By engineering the interfaces between the model, the clinician, and the institutional environment, the technology becomes resilient to real-world friction.
Modeling Symbiosis via Multivariate Perspectives
If humans, machines, and environments are deeply coupled, traditional single-dimension metrics—such as algorithmic accuracy or mechanical efficiency—are no longer sufficient. Evaluating a system based on a single metric risks optimizing one component at the expense of the broader system's stability.
To address this, Jairo Inga et al. introduce a multivariate framework in Human-machine symbiosis: A multivariate perspective for physically coupled human-machine systems. They decompose human-machine symbiosis into four distinct dimensions:
- Task: The objective goal being pursued.
- Interaction: The physical or informational coupling between user and machine.
- Performance: The efficiency and effectiveness of the joint system.
- Experience: The cognitive and subjective state of the human operator.
Accounting for these dimensions requires a shift in foundational engineering practices. As József Kövecses et al. observe in Preface, disciplines like multibody system dynamics have transitioned from classical rigid-body mechanics toward multidisciplinary modeling that integrates mechatronics, optimization, and vehicle dynamics to handle these complex real-world interactions.
Why the Systems Approach Works
The systems approach works because it explicitly targets emergent properties—behaviors and failure modes that only appear when components interact.
When advanced technologies fail in real-world settings, the root cause is rarely a lack of isolated technical capability. For instance, in an investigation of robotic care services, Helen Dickinson et al. demonstrate in “We’re Still Struggling a Bit to Actually Figure Out What That Means for Government”: An Exploration of the Policy Capacity Required to Oversee Robot Technologies in Australia and New Zealand Care Services that deployment failures typically stem from a lack of understanding regarding the broader societal, ethical, and systemic impacts of the technology, rather than a deficit in technical expertise.
Similarly, in safety-critical network infrastructures like power grids or air traffic control, Gianvito Losapio et al. argue in Human-AI interaction in safety-critical network infrastructures that effective human-AI collaboration requires an interdisciplinary synthesis of computer science, psychological constructs, and industrial practices to ensure transparency and human agency.
To manage these multi-dimensional dependencies systematically, researchers rely on structured organizational frameworks. For example, Rui Zhang demonstrates in Large-scale project system management based on Hall three-dimensional structure how transforming classic systems engineering theories into practical activity matrices allows engineers to integrate time, logic, and professional knowledge across complex technical and administrative boundaries. Ultimately, by moving away from isolated component design and embracing the interdependent relationships between humans, machines, and environments, engineers can build systems that are inherently resilient, predictable, and aligned with human intent.
Key Foundational Work
IEEE Transactions on Cybernetics publication information (2014) — contributed 3 concept elements to this direction.
IEEE Transactions on Cybernetics publication information (2013) — contributed 3 concept elements to this direction.
Messages from the conference chairs (2014) — contributed 1 concept elements to this direction.