Geographic information science II: Mesogeography

1 The search for grand social laws

Social physics is driven by a belief that laws, theories and models similar to those found in physics can also be discovered for social phenomena (Barnes and Wilson, 2014). Social physics has a long history, with early origins in Thomas Hobbes’ Leviathan and the work of John Graunt on London mortality data in the 17th century (Ball, 2004; Stewart, 1950). In the 19th century, French philosopher Auguste Comte used the term ‘social physics’ originally to describe the new discipline he later called sociology, emphasizing its focus on systemic organization of social life and the scientific method as the means to understand these systematic properties. Also in the 19th century Belgian statistician Adolphe Quetelet used the term ‘social physics’ to describe the application of probability and statistics to analyze regularities in social data (Scott, 2016).

Quetelet believed that studying the individual in isolation was fraught with danger since it is difficult to parse the general and particular at this level. This could only be achieved with a sufficient amount of data. With enough measurements, the same apparent causes would lead to the same apparent effects, and social phenomena would resemble physical phenomena in their regularity. Quetelet introduced the concept of an average man as an ideal type representing the social and physical characteristics of a population. Besides being an ideal, the average man is also an empirical regularity determined through a large number of measurements (Mosselmans, 2005). As an example, Quetelet developed an index describing the relationship between height and weight for a normal body type; this now serves as the basis for today’s widely-used body mass index (Eknoyan, 2008).

Social physics came to fruition in the mid-20th century with the work of George Zipf on power law distributions and John Q. Stewart on population potentials (Barnes and Wilson, 2014; O’Sullivan and Manson, 2015). Social physics found its way into human geography via William Warntz. Along with Stewart, Warntz argued for a ‘macrogeography’ focusing on general principles, contrasting with the descriptive ‘microgeography’ that dominated the discipline in the mid-20th century (Stewart and Warntz, 1958). ¹ Warntz is also known today for his Law of Refraction in transportation; he used the term ‘social physics’ in the title of the famous paper in which this is laid out (Warntz, 1957). There are direct connections from social physics to spatial analysis and GIScience, with pioneers such as Jack Dangermond, Bill Bunge and Michael Goodchild influenced by Warntz during his time at Harvard’s Laboratory for Computer Graphics and later at the University of Western Ontario (Barnes and Wilson, 2014).

Classical mechanics describes physical phenomena based on fundamental dimensions such as position, mass and force. Social physics describes human phenomena based on the fundamental dimensions of people, distance and time (Stewart, 1950). An article in the Princeton Alumni Weekly, describing efforts to build a Department of Social Physics, summarizes the perspective elegantly:

Professor Stewart and his colleagues…treated large aggregates of individuals as though they were composed of social molecules, without attempting to analyze the behavior of each molecule. They then attempted to describe demographic, economic, political, and sociological situations in terms of such physical factors as time, distance, mass, and numbers of people. ( Princeton Alumni Weekly, 1955: 17)

2 Social physics on the skids

Social physics raised uncomfortable questions from its very beginning. The concept of social forces raised questions about free will, and the idea of an average person raised questions with sinister implications about whether we could pro-actively improve the distribution of social traits in a population (Ball, 2004). In the 1970s and 1980s, social physics and its geographic offspring, spatial analysis, came under intellectual attack. Humanistic geographers and other social scientists criticized its cold view of humanity: people as particles within an abstract nowhere rather than intricate beings living in unique places. Marxists reject social physics’ neglect of social conflict and its seeming reification of laissez faire capitalism (Cresswell, 2013). Social physics is also discounted as a form of monism: a belief that there is one set of principles that explain both physical and social worlds (Barnes and Wilson, 2014).

In the next section, I will show how the new social physics addresses the charges of monistic crushing of rich reality from social research. I will not address the Marxist critique, other than to point out that Marx himself sought a scientific theory of society, and there is nothing inherent in the methods discussed below that precludes addressing questions surrounding differential relationships to the means of production, the accumulation of wealth via surplus value from labor and the evolution of social norms to create order (see Barnes, 2009; Conte et al., 2012; Epstein and Axtell, 1996; Saam and Harrer, 1999; Wyly, 2009).

1 More is different

The epigraph at the beginning of this report suggests the major problem with classic social physics: a belief that aggregate outcomes are a straightforward summation of individual behaviors. It turns out that many physical and social systems cannot be described in this manner beyond a first approximation.

In a landmark paper, Anderson (1972) argues that for many phenomena there is broken symmetry between reductionist approaches that search for fundamental laws and constructionist approaches that try to explain phenomena in terms of these fundamental laws: ‘the ability to reduce everything to simple fundamental laws does not imply the ability to start from these laws and reconstruct the universe’ (p. 393). For example, while particle physics can describe well the fundamental principles governing the behavior of a small number of elementary particles, these explanations are inadequate when extrapolated to the behavior of a large and complex collection of particles. As the title of Anderson’s paper states, more is different.

Social physics does not imply there is one set of principles to explain physical and social phenomena. The broken symmetry between reductionist and constructionist approaches creates a natural hierarchy in science, building from the physical to the biological to the social sciences. Each level is grounded in the principles of the previous level. While explanations at each level ultimately should be consistent with principles at lower levels, each level introduces new principles that require fundamental research. Reduction to physical or biological principles cannot explain human phenomena (Anderson, 1972). ²

This does not mean that methods and models based on physical principles are inappropriate for understanding human geography. But often these techniques can be given a firmer foundation and enhanced with principles that emerge at higher levels. Consider the ‘gravity’ or spatial interaction (SI) model. Originally formulated as an analogue to Newton’s Law of Gravitation, Alan Wilson later provided a stronger physical foundation by deriving SI models via entropy maximization. Subsequently, Daniel McFadden provided a behavioral foundation for SI models, and Stewart Fotheringham extended this to spatial choice. The result is a sophisticated family of SI models, including ones that address the spatial context of origins and destinations (Fotheringham, 2017).

2 Spatial context matters

Collective systems consisting of a multitude of interacting components can display aggregate behavior that is qualitatively different from the behavior of elementary components considered in isolation. Interactions among the components generate emergent properties that are unpredictable, even if one completely understands their fundamental laws. These kinds of emergent properties are evident in human systems such as cities and regions characterized by organized complexity and intricate feedback loops leading to surprising behaviors and counterintuitive outcomes from policy interventions (Batty, 2007; Pollock, 2016). Understanding these systems requires a ‘middle path’ between studying the system at the individual and aggregate levels, focusing instead on the emergence of collective behavior from the behavior and interactions of individual agents (Flake, 1998).

It is now trite to argue for a complex systems approach to understanding human and coupled human-physical systems. However, that is not my point. My point is that 21st-century social physics has moved beyond a sterile conceptualization of people and geography to emphasize heterogeneity and spatial context as key drivers of interesting behavior. As Phillip Ball states in his book Critical Mass:

[social physics does not imagine] that people are so many soulless, homogeneous effigies to be shuffled this way and that according to blind mathematics. Instead, what physicists are now trying to do is gain some understanding of how patterns of behavior emerge – and patterns undoubtedly do emerge – from the statistical melee of many individuals doing their own idiosyncratic thing: helping or swindling each other, following the crowd or blazing their own trail. (Ball, 2004: 31)

1 Seeking mesogeographic knowledge

I use the term mesogeography to describe a focus on the role of spatial context in determining the behavior of collective systems such as cities and regions. Mesogeography is particularly concerned with interactions among entities in the system and how aggregate behavior and patterns emerge from these interactions. Besides a middle ground between description-seeking and law-seeking, mesogeography is also a middle path between the reductionist and aggregate approaches.

Mesogeography seeks generalizations, but not universal laws: the casual relations holding in a range of environments (see Glymour, 1983). This is similar to the middle-range theories proposed in the mid-20th century by sociologist Robert Merton: empirically grounded theories that address recognizable social phenomena rather than abstract entities, serving as hypothesis generators, a basis for policy, and as stepping stones to grander theories. It also shares some features with critical realism, in particular, a distinction between necessary structural conditions and contingent local conditions (see Harvey and Reed, 1996; Yeung, 1997).

Mesogeography abandons the quest for universal laws but not evidence-based science and policy. As I discuss in the first report in this series (Miller, 2017b), while the application of evidence-based policy to public health, transportation and social problems is promising, critics argue that policy interventions in the real world are highly context-dependent. Generalization from a single study is not sufficient for building effective policy: required instead are parallel studies of policy interventions in different times and places. Similarly, building mesogeographic knowledge requires not only measuring how processes evolve in a particular place and time but comparing among processes and outcomes across different places and times.

2 GIScience and mesogeography

GIScience is well-positioned in the quest for mesogeographic knowledge, spanning both law-seeking and description-seeking: the former in its algorithms and methods, the latter in its data (Goodchild, 2004). While 20th-century GIScience inherited the sterile geography of 20th-century social physics, this was due to practical limits on computing power and data collection. These limitations are more relaxed in the 21st century. In GIScience these include techniques for measuring and analyzing spatial context, data-driven geography, experimenting in real and simulated worlds, and analytical time geography.