Abstraction Hierarchy
Abstraction hierarchy is a framework for describing complex systems at multiple levels of abstraction, from the most concrete (physical form and physical processes) to the most abstract (purposes and constraints). It was developed by Jens Rasmussen in the context of cognitive systems engineering and has been applied to the design of human-machine interfaces, the analysis of accidents, and the modeling of complex socio-technical systems.
The Five Levels
Rasmussen's abstraction hierarchy identifies five levels, each representing a different way of understanding the same system:
1. Functional purpose. The top level describes why the system exists: its goals, its constraints, and the value it provides. For a nuclear power plant, the functional purpose is to generate electricity safely and economically. For a hospital, it is to provide care that improves health outcomes.
2. Abstract function. The second level describes the system's purpose in terms of physical and informational flows: mass balance, energy balance, information flow. The abstract function of a power plant is the conversion of nuclear energy to electrical energy, with constraints on heat dissipation and radiation containment.
3. Generalized function. The third level describes the processes that achieve the abstract function: heat generation, heat transfer, steam production, turbine rotation, generator operation. These are the generalized functions that any power plant must perform, regardless of its specific design.
4. Physical function. The fourth level describes the specific equipment and components that perform the generalized functions: the reactor core, the heat exchanger, the steam turbine, the generator. The physical function is the design choice: this particular configuration of components.
5. Physical form. The bottom level describes the actual physical appearance and location of the components: the geometry of the reactor vessel, the layout of the control room, the routing of the piping. This is the most concrete level, the one visible to the human operator.
The Mapping Principle
The key insight of the abstraction hierarchy is that the levels are not independent descriptions but mappings of the same system at different resolutions. A change at any level propagates to the others: a change in the physical form (a new valve) may require a change in the physical function (a new control loop), which may alter the generalized function (a new heat transfer path), which may affect the abstract function (a new energy balance), which may compromise the functional purpose (reduced safety margin).
This is why the abstraction hierarchy is a powerful tool for accident analysis. Accidents often occur because a change at one level was made without understanding its consequences at the others. The Air France Flight 447 accident involved a change at the physical function level (the autothrottle logic) that was not understood at the generalized function level (the stall recovery procedure) or the functional purpose level (the safety of the flight).
Design Implications
The abstraction hierarchy has direct implications for the design of human-machine interfaces. The interface should support the operator's movement across levels: the ability to see the physical state of the system (bottom level), the ability to understand the functional implications of that state (middle levels), and the ability to evaluate the state against the system's goals (top level). An interface that shows only the physical level leaves the operator without understanding; an interface that shows only the abstract level leaves the operator without grounding.
The Ecological interface design approach of Vicente and Rasmussen applies the abstraction hierarchy to the design of control room interfaces. The interface is organized as a hierarchy of displays, each corresponding to a level of abstraction, with explicit links showing how the levels map to each other. The operator can move up and down the hierarchy as the situation demands, zooming in to the physical details when troubleshooting and zooming out to the functional purpose when evaluating strategy.
Connection to Systems Theory
The abstraction hierarchy is a structuralist framework: it assumes that the system has a coherent structure that can be described at multiple levels. This assumption is not always valid for emergent systems, where the behavior at one level may not be reducible to the structure at another. The Emergence of flocking behavior from local bird-bird interactions is not captured by an abstraction hierarchy of the flock; the hierarchy would describe the flock as a system with a purpose (migration), but the actual behavior is produced by local rules that have no representation at the higher levels.
Nevertheless, the abstraction hierarchy remains a valuable tool for designed systems — systems that have been engineered to achieve a purpose. For these systems, the hierarchy provides a language for describing the relationship between design intent and operational reality, and a method for identifying the gaps between levels that produce accidents and inefficiencies.