Based on the provided sources, Goodhart’s Law—“when a measure becomes a target, it ceases to be a good measure”—can be explained through the lens of observer reflexivity as a dynamic interaction between the Observer (who freezes the world into a model) and the Observed (a complex, anticipatory system that adapts to that model).
In complexity theory, this is not merely a data error, but a fundamental property of biosocial systems where the “observed” possesses its own models of the observer[1],[2].
1. The Model as a Static “Lie”
To measure a system, an observer must first create a formal model. This involves defining an equivalence class (e.g., “downtime,” “risk,” or “profit”) that simplifies the infinite complexity of the material world (the “Other”) into a finite set of proxy variables[3],[4].
• Freezing Complexity: The observer “freezes” the dynamic world into a static structure (a rate-independent rule) to make it tractable[1],[3].
• The “Lie”: This process creates a “simulacrum” or a “lie”—a simplified narrative that asserts, “No, really, it is just that simple”[5],[6]. Goodhart’s Law begins when the observer mistakes this simplified map for the territory[7].
2. The System as a Reflexive “Other”
Unlike mechanical systems (e.g., billiard balls), biosocial systems (e.g., employees, traders, or organisms) possess their own internal models and narratives[8],[9]. They are anticipatory systems; they do not just react to the past, they predict the future based on the observer’s actions[9],[10].
• The Observer-Observation Complex: You cannot separate the observer from the observed. The system is watching the observer watch it[11],[12].
• Mimicry and Deception: Just as a cockroach mimics a ladybug to deceive a predator (the observer), a social system mimics the behavior required by the metric[2],[13]. The system optimizes its appearance to fit the observer’s definition of “good,” effectively “lying” to the model[2].
3. The Collapse of the Measure (The Law in Action)
When a measure becomes a target, the observer has imposed a constraint (a rate-independent rule) on the system[14],[15]. Because the system is complex and adaptive, it evolves a new behavior (a realization) that satisfies the constraint (the metric) while changing its underlying essence (the actual quality or risk profile)[16],[17].
• Breaking the Equivalence: The measure relies on a stable relationship between the proxy variable (the metric) and the underlying reality (the essence). When the system adapts to hit the target, it breaks this relationship. The observed entity no longer belongs to the “equivalence class” the observer originally defined[18],[19].
• Example: If an IT department is penalized for “downtime” (the measure), they may change the definition of downtime or keep systems offline during “maintenance” windows to avoid the penalty. The data (the sign) remains good, but the service (the essence) degrades. The realization (the specific behavior) has evolved to defeat the model[16].
Summary
Through the lens of reflexivity, Goodhart’s Law is the inevitable result of the “Other” (the material system) reasserting itself against the Observer’s definitions. The observer tries to constrain the system with a static model, but the system—being a “becoming” entity with its own agenda—evolves to render that model obsolete[20],[13].
References
[1] Allen - Narratives and transdisciplines for a post-industrial world.pdf [2] Allen - Narratives and transdisciplines for a post-industrial world.pdf [3] Allen 2006 - The Epistemological predicament associated with purposive qunatitative analysis.pdf [4] Allen 2006 - The Epistemological predicament associated with purposive qunatitative analysis.pdf [5] Allen - Narratives and transdisciplines for a post-industrial world.pdf [6] Allen 2006 - The nature of ecological complexity.pdf [7] Allen - The loss of narrative - paradigms models mechanisms and analogy.pdf [8] Allen - Narratives and transdisciplines for a post-industrial world.pdf [9] Allen 2006 - The Epistemological predicament associated with purposive qunatitative analysis.pdf [10] Allen 2006 - The Epistemological predicament associated with purposive qunatitative analysis.pdf [11] Allen - Narratives and transdisciplines for a post-industrial world.pdf [12] Allen - The loss of narrative - paradigms models mechanisms and analogy.pdf [13] Allen 2006 - The nature of ecological complexity.pdf [14] Allen - Confronting Economic Profit with Hierarchy Theory the concept of gain in ecology.pdf [15] Allen 2014 - Holons creaons genon environs in hoerarchy theory where we have gone.pdf [16] Allen 2014 - Holons creaons genon environs in hoerarchy theory where we have gone.pdf [17] Allen 2014 - Holons creaons genon environs in hoerarchy theory where we have gone.pdf [18] Allen 2004 - Type Scale and Adative Narrative - Keeping Models of Salmon Taxicology and Risk Alive to the World.pdf [19] Allen 2004 - Type Scale and Adative Narrative - Keeping Models of Salmon Taxicology and Risk Alive to the World.pdf [20] Allen 2006 - The Epistemological predicament associated with purposive qunatitative analysis.pdf