A Boeing engineer once explained to me that aeroplanes do not really exist. These sleek monsters that carry us to the farthest reaches of the planet were, to him, no more than a million parts flying in close formation, not a single one of which was capable of flying by itself. What turns a million parts into an aeroplane is the organization and interconnection of those parts. It would take an unusually smart AI to look at such a pile of parts and understand that their purpose is to fly. It would be even less likely that an AI could determine from such a pile of parts how to make a “better” aeroplane.

Cities are not designed and engineered like aeroplanes. While individual city blocks, parks, transportation, and so forth are designed and engineered, the city as a whole emerges as patterns of behaviour that persist despite the evolution of the natural- and built-environments. Generations of inhabitants organise and interconnect the city’s many parts so that it evolves into, not merely a built-environment on top of a natural-environment, but a living entity that we might easily call the “soul” of the city. Just as the aeroplane’s ability to fly emerges from the organisation and interconnection of a million parts, so the city’s ability to sustain its population emerges from how that population connects, organises, exploits, and further develops its millions of services and affordances.

In recent years our New England village has made a practice of welcoming Syrian refugee families . In June 2016 a family that was bombed out of Homs arrived and quickly became self-sufficient, but still carries a model of where and how they used to live. In September 2017 I attended a talk given by Stephanie Saidaña, an American journalist living in Jerusalem who is collecting stories from families that have been resettled in European countries. She spoke of their efforts to sustain their family traditions by seeking out traditional ingredients, as if to pull that supply chain back into existence far from home and to resurrect the souls of their former cities. In Oslo in November 2017, I heard a relief worker speak about the persistence of place in Syrian cities that have been bombed. Even when buildings are reduced to rubble for the residents the soul of the city remains. It still structures and orients their daily lives.

So, if cities have souls, how can observe them?  How can we come to understand what a city does?  How can we make it “better?”

 

A Modest Proposal

Even as smart city engineers, we too easily pass over thinking of the city as a complex of systems and narrow our view to some simpler, though still complicated, system that we can study in isolation. But cities do not function simply as a collection of parts. They function through the ways in which their citizens and other actors organize and connect these parts into personal and civic systems through which they can live their lives. Actors both exploit and contribute to the soul of the city. If we understood how they do this, we might then apply our skills and technologies to making cities  “better,” although, as Jacob Bronowski liked to point out, “first you have to tell me what ‘better’ means.”

 

To Begin at the Beginning

My mental model of a city is rooted in the decisions made by the various kinds of actors within the natural-, the built-, and the social-environments. These decisions are manifest as transactions that are made between actors in these environments. That is, for each transaction there are two participants, one providing a good or service and one consuming it, although each transaction ripples away into the environments. In the social-environment these actors include individuals as well as collections of individuals such as governments, enterprises, civic organisations, and informal groups. In the social-environment some of these transactions involve financial exchange, but generally they represent competition for some good or service – a subway carriage seat or a liter of water. For simplicity, I use financial terms for all transactions, but these terms should be interpreted in context. In aggregate these transactions form patterns or systems at various scales. For simplicity, I assume that all systems are in their linear regimes, although we eventually need to deal with tipping points.

Such transactions are motivated by the short- and long-term goals of each actor and are intended by the actor to maximise its satisfaction against some set of needs and desires. I do not assume, however, rational behaviour on the part of any actor. Instantaneously the cost of each transaction depends on the availability of and demand for the good. Some goods, a newspaper, for example, are relatively freely available, while others, such as a subway carriage seat, have, instantaneously,  a fixed capacity. Some can be easily replaced, a liter of water, while others, a hectare of land, can only be consumed once in the short-term.  By “satisfaction” I mean the ability of an actor to progress towards satisfying its needs and desires in the pursuit of short- or long-term goals. For individuals, we may take Maslow’s hierarchy of needs as an example framework. An actor’s decision to commit to a transaction depends on its awareness and understanding of offered choices. An actor that chooses to buy coffee a local Starbucks may not be aware of a new café opening. On the other hand, news of a café opening three kilometers away is unlikely to reach or motivate this same actor. A key aspect of any actor’s behaviour is this ability to gather, to assess, and to transmit intelligence relevant to its decisions at multiple spatial and temporal scales.

In addition to these motivations, decisions are also constrained in many ways:  the actor’s ability or willingness to pay, the need to comply with  laws, regulations, policies, beliefs, and the actor’s beliefs, habits or tastes.

In complex systems, a critical consideration concerns scales. In addition to spatial and temporal scales, we also need to include the scale of the transactions as well as the timescales of the goals to which the transaction relates. These may extend from life decisions such as getting married to minor decisions such as where to cross the street.

A complication for urban systems is that of the city’s boundary. Many such boundaries are defined, including ecological and environmental factors, various forms of government, flows of resources and traffic, and local, regional, national, and global trade. In practical terms it will be necessary to represent some of these as externalities.

A final, major complication concerns initial conditions. Launching such a model requires knowledge of the vast number of starting points from which it will evolve. In 2008 IBM, in collaboration with the City of Portland, Oregon, developed a Systems Dynamics model of the city, looking only at the inter-dependences among a dozen or so municipal services. At its peak, this model required the specification of several thousand initial conditions. By simplifying the model, this was eventually reduced to around one thousand for which data were available.

The model I propose here is far more complex than the Portland case and requires far more initial conditions, eventually millions. If we are modeling a real city, then much of the information about the natural-environment may be available from the city’s or the country’s governments or universities. Information about the built-environment should be available from the municipal government, supplemented with information collected via satellites or by private organisations such as Google. Information about the social-environment may be available from city records, tax records, smart city or IoT Big Data, and credit card transactions. Assembling and preparing such information is one of the major challenges. Even then, it will be spatially and temporally incomplete and out of date.

 

Creating a Model

With the above as prologue, my proposal is to create a model of a city in which social actors compete with others at various spatial, temporal, and transactional scales to increase their satisfaction by committing to choices among offerings within the natural-, built-, and social-environments.  The soul of the city emerges from their collective drive towards satisfaction.

My thought here is to begin with simplified natural-, built-, and social-environments for which observed information is available and, through exploration, to constrain the model’s behaviour to match the real-world behaviour. We thus begin at roughly the complexity of the Portland model mentioned above (although I am not thinking of a Systems Dynamics model) and use an AI to tune the transactions  to match the available observations by iterating over distributions of the actors’ characteristics and constraints. When this process reaches equilibrium, we move to the next higher level of complexity, that is larger numbers of actors, higher spatial and temporal resolutions, more complex interactions, and, if possible, more detailed observations and repeat the exercise until we reach the desired granularity of transactions.

So what does this look like when it is all dressed up?  It looks like an immense multi-player online game, in which of social actors compete to maximise their individual satisfactions. The games are the exploration over a sequence of time steps of each actor’s attempt to find and to commit to transactions that it believes will advance its overall satisfaction. The granularity of the time steps may range from, say, one minute up to several years with appropriately different ranges of choices.

 

Outcomes

The ensemble of these satisfactions for all actors within the boundary of a city and within a specific period might represent what Geoff West has called the “Hamiltonian of the city”, where “satisfaction” stands in place of energy. An economist might consider how closely the city’s actors are to a Nash Equilibrium. However, my goal is not specifically to maximise individual or aggregate satisfaction, although this may be a by-product. From the perspective of smart cities, the questions that interest me concern our ability to improve the satisfaction of some or all of the actors of the city.  This is what “better” means.

With this model we might explore many questions:

• How effective is the built-environment at supporting the needs of the social-environment?
• How easily can actors discover choices?
• How much is the natural-environment stressed by the built- and social-environments?
• How wasteful is the social-environment in its resource consumption?
• How close is the city to a balance of trade with the external environments?
• How do the latter influence the long-term competitiveness of the city?
• How close is the city to achieving a Nash Equilibrium?
• What effects would certain policies have on achieving such an equilibrium?
• What levels of information-based skills are required for a social-actor to be competitive within a given city?

 

What is the Purpose of a Smart City?

The primary effect of deploying smart city technologies would seem to be to increase the volume of information that is captured about the states of the three environments, that is made available to some or all actors, and that can be assessed by these actors and used by them to recognise the choices that are available to them to improve their individual satisfactions.

As smart city engineers over the past fifteen years or so, we have shown great creativity in applying technologies to making discrete systems, such as transportation, “better.” But so far, we have not tackled the bigger question of how these discrete improvements make the city as a whole “better.” While some of these discrete improvements, for example shared and autonomous transportation systems, have great promise in reducing congestion and improving the ease of use of the city’s built-environment, in other cases the actual impact has been marginal and sometimes transient.

Perhaps the greatest achievement of smart cities to date has been to stimulate the diffusion of IoT and Big Data systems. Given appropriate and privacy-protecting access, these may become the foundation for the emerging information-environment on which AI can now build an understanding of the soul of a city.