The Sciences of the Artificial

Abstract

This excerpt from the first chapter of The Sciences of the Artificial
(1969; 1996) by Herbert A. Simon establishes the epistemological foundations for distinguishing natural sciences from the “sciences of the artificial.” While natural sciences seek hidden patterns to explain how things are, the sciences of the artificial deal with objects synthesised by human beings, characterised by functions, goals, and normative imperatives—addressing how things “ought” to be. Simon introduces the crucial concept of the artefact as an “interface” between an “inner” environment (the substance and organisation of the object itself) and an “outer” environment (the context in which it operates); the artefact’s effectiveness depends on the successful adaptation of these two environments to one another. The text further explores the role of simulation as a source of new knowledge that can reveal the hidden implications of known premises. It defines both computers and the human mind as “physical symbol systems.” According to Simon, intelligence is fundamentally the work of these systems, which can encode information, manipulate structures, and adapt to their environment.
The re-proposal of this classic text within the contemporary context of urban studies and Artificial Intelligence (PlanAIr) is driven by three fundamental reasons. First, Simon provides a critical ontological definition, reminding us that the world we inhabit is predominantly man-made. In this view, the city is the artefact par excellence:
not a natural phenomenon to be passively observed, but a complex, designed system that must answer to human purposes, thus legitimising urban planning as a rigorous science of the artificial. Second, the vision of the artefact as a “meeting point” between inner and outer environments offers a powerful metaphor for Urban AI. Intelligent technologies in the city act as an interface between physical infrastructure and citizens’ social or environmental dynamics, requiring mutual adaptation to function effectively. Finally, as Simon’s work is foundational to symbolic Artificial Intelligence, revisiting it today allows us to grasp the theoretical roots of rule-based and logical AI. This historical perspective is crucial for distinguishing and potentially integrating symbolic approaches with the currently dominant data-driven paradigms, thereby recovering the capacity to reason about goals, meanings, and design imperatives rather than relying solely on raw data processing.
In this article, we examine these ontological issues, discuss existing frameworks that aim to unify fragmented information, and explore the practical implications for urban AI applications. The thesis is that ontologies—structured and formal representations of knowledge—offer a powerful tool to address the challenges outlined above, while serving as a blueprint for defining, categorizing, and interrelating the entities present in urban environments and putting them to work in urban planning.