Systems Thinking

Cities as complex socio-technical systems

Session Roadmap

• Part I: The Landscape of Urban Complexity
  • The “Brain” of the Smart City & 4 Heuristics
• Part II: The Logics of Systems
  • Deep Dive into Feedback Loops
  • Workshop I: Mapping Your Wicked Problem
• Part III: Decision Support Tools
  • Workshop II: Stakeholders and Data

The Landscape of Urban Complexity

Part I

The City = Complex Systems

• A city is decomposable into semi-predictable parts.
• It is a dynamic process of organized complexity.
• The Smart City Paradox:
  • We have new technical capabilities (Big Data, sensors, AI).
  • But they rely on modelling techniques devised decades ago to solve social problems.

Complexity alive

What the Smart City is NOT

• It is not a crystal ball for perfect prediction.
• It is not a digital twin that is the reality.
• The Map is not the Territory:
  • Models are simplified representations.
  • In a complex system, “Total Control” is an illusion.
  • We model to understand potentiality, not to dictate outcomes.

The Science of Cities

“…cities must now be looked at as constellations of interactions, communications, relations, flows, and networks rather … than locations […] location is, in effect, a synthesis of interactions.”

— Michael Batty, 2013

Organized Complexity

Jane Jacobs

Cities are not:

• Problems of Simplicity
• Problems of Disorganized Complexity (millions of random variables).

They are problems of Organized Complexity:

• A huge number of factors which are interrelated into an organic whole.

The Interconnection of Parts

“The city is not a work of art… the city is a process.” – J.Jacobs

• It is a Bottom-Up system.
• Small, local interactions \(\rightarrow\) Large, global patterns.

Complexity Science

The Reductionist Shortcut:

• A mental habit of breaking the city into isolated “parts” to make it easier to manage.

The Complexity Reality:

• Science not of the “atoms” (buildings, roads, people), but of the connections between them.

Cross-Scale Interactions

• Single street-level decisions ripples up to the metropolitan scale.

Systems

Complex

• Adaptive relationships.
• Emergent outcomes.
• More than the sum of its parts.

Complicated

• Fixed relationships.
• Predictable outcomes.
• The sum of its parts.

The City is not a Tree

“It is for this reason - because the mind’s first function is to reduce the ambiguity and overlap in a confusing situation … it is endowed with a basic intolerance for ambiguity - that structures like the city, which do require overlapping sets within them, are nevertheless persistently conceived as trees.” – Christooher Alexander

Systems Thinking 101

Systems Thinking

Seeing the Web of interactions

Linear Thinking

A \(\rightarrow\) B (Cause and Effect)

FOCUS ON THE RELATIONSHIP BETWEEN PARTS

NOT JUST PARTS THEMSELVES.

Modeling as a Heuristic

A model is not “The Truth.”

It is a Heuristic: a mental shortcut or tool for discovery.

• “All models are wrong, but some are useful” (George Box).

The Logics of Systems

Part II

Smart City’s brain

The Back-end

The heuristics, logic, and models.

The Front-end

Dashboards, visualizations, and stakeholder apps.

Logic must precede display.

The Dimensions

of most urban logic modelling

SPATIAL

RELATIONAL

TEMPORAL

SYSTEMIC

1. The Spatial Dimension

• The Logic of “Where”: Focuses on geometry, distance, and density.
• Key Question: How does the physical arrangement of assets (buildings, parks, roads) determine urban performance?
• Core Concept: Social Physics. Just like gravity, “mass” (high-density jobs or housing) exerts a pull on the surrounding area.

2. The Relational Dimension

• The Logic of “Connection”: Focuses on topology and network hierarchy.
• Key Question: How is the city “wired”? Is it about how far you are in meters, or how many turns you are from the main network?
• Core Concept: Graph Theory. The city as nodes and edges.

3. The Temporal Dimension

• The Logic of “When”: Focuses on change, evolution, and path dependency.
• Key Question: How did the city get here, and how fast is it changing?
• Core Concept: Dynamics. Recognizing that an urban intervention today might take 10 years to show its true impact (delays).

4. The Systemic Dimension

• The Logic of “Why”: Focuses on causality and interdependencies.
• Key Question: What are the unintended consequences? If we improve transport, does it inadvertently raise rents and displace the people it was meant to help?
• Core Concept: Feedback Loops. Circular causality where an output of a system becomes an input for the next cycle.

The Models

only some…

Land-Use Transport	Network Models
System Dynamics	Agent-based Models

Land-Use Transport

LUTI

Primary Dimensions:

Spatial (High) + Relational (Medium)

The Anatomy

• Land-Use data
• Transport

The Engine

Uses gravity equations to find a state of “Equilibrium” between where people live, where they work, and how they travel.

Land-Use Transport

eg. Predicting the impact of a new Metro line on residential density in a suburban area.

Pros

Excellent for long-term strategic planning and infrastructure “stress testing.”

Cons

Highly aggregate (treats people as masses) and mostly static (ignores the “chaos” of daily life).

What it lacks

Temporal and Systemic feedback. It assumes a city is a puzzle that can be “solved” into a perfect balance.

Network Models

NM eg. Space Syntax

Primary Dimensions:

Relational (High) + Spatial (Medium)

The Anatomy

• Nodes
• Edges

The Engine

Topological Analysis. It measures “centralities” to determine expected movement hierarchies.

Network Models

eg. Redesigning a town center to increase footfall for local shops by improving topological connectivity.

Pros

Very high correlation with real-world human movement; requires minimal data (just the map).

Cons

Can be “blind” to land use. It might predict high movement on a street that is topologically central but has no reason for people to go there.

What it lacks

Systemic “Attractors.” It prioritizes the shape of the container over the content inside it.

System Dynamics

SD eg. The Forrester Model

Primary Dimensions:

Temporal (High) + Relational (High)

The Anatomy

• Stocks (Accumulations)
• Flows (Rates of change)
• Feedback Loops

The Engine

Differential Equations. It simulates how quantities (stocks) change over time based on rates of change (e.g., construction rate).

System Dynamics

eg. Modeling how a city-wide “Air Quality Zone” affects car ownership and public health over two decades.

Pros

Powerful for identifying “Policy Resistance” and understanding long-term, non-linear consequences.

Cons

“Spatially Blind”—it treats the entire city (or zone) as a single bucket, ignoring internal geography.

What it lacks

Spatial Granularity. It assumes a city is a “well-mixed” vat of variables rather than a physical place with unique neighborhoods.

Agent-Based Models

ABM & Celullar Automata (CA)

Primary Dimensions:

Spatial (High) + Systemic (High) + Temporal (Medium).

The Anatomy

• Agents: (People, vehicles, firms) with unique attributes.
• Rules: Simple “If/Then” behaviors.
• Environment: The grid or network they inhabit.

The Engine

Bottom-up Micro-simulation. Simple rules from the agents give rise to “emergent” states from thousands of these local interactions.

Agent-Based Models

_eg. Simulating pedestrian evacuation patterns in a high-density transit hub to identify “bottlenecks” that traditional flow models miss.

Pros

Captures “Emergent Phenomena” and diversity; allows for the modeling of irrational or “human” behaviors.

Cons

Computationally expensive; very difficult to “validate” because small changes in rules can lead to wildly different outcomes.

What it lacks

Predictive Certainty. Because it is stochastic (probabilistic), it gives you a range of “possible futures” rather than a single definitive answer.

One model strength is the other model’s weakness

	LUTI	NM	SD	ABM	CA
Spatial	\(\uparrow\)	~		~	\(\uparrow\)
Relational	~	\(\uparrow\)
Systemic			~	~	\(\uparrow\)
Temporal			\(\uparrow\)	\(\uparrow\)	~

\(\uparrow\) = high \(\~\) = medium

Important

Most modern models have figured out a way to incorporate (almost) all dimensions to different degrees. Yet own their heuristic limit their integration efficacy.

Workshop

Mapping your wicked problem

The “Wicked” Problem

• Urban challenges are not puzzles (with one right answer).
• They are Wicked Problems:
  • They are interconnected: solving one creates another.
  • They have no “stopping rule” or perfect solution.
  • They are better or worse, not right or wrong.

Beyond the Mental Model

flowchart LR

subgraph G1["Conceptual"]
    A["Verbal<br/>description"]
    B["Causal-Loop<br/>Diagram"]
end

C["Model<br/>Formulation"]
D["Model<br/>Behaviour"]
E((("Policy<br/>Implications")))

A ==> B
B --> C
C --> D
D --> E
D -..-> B 
E -.-> B
E -.-> A

classDef yellow fill:#FFDC00,color:#2b2b2b,stroke-width:0px,text-align:center;
classDef gray fill:#999999,color:#ffffff,stroke-width:0px,text-align:center;
style G1 fill:#2b2b2b,color:#ffffff,stroke:#eee,stroke-width:2px

class A,B yellow
class C,D,E gray

Setting the Boundary

A model is only as good as what you leave out.

• Endogenous: Factors the system controls.
• Exogenous: Factors outside our influence (the “Environment”).

The Challenge: If you draw the boundary too small, you miss the “Wicked” side-effects.

Feedback Loops

Enable the discovery of :

• emergence
• path dependency
• tiping points
• equilibrium / oscilation

Feedback Loops

loopy

System Archetypes

• Balancing & Limits
• Unintended Feedback
• Resource & Power Loops

• Limits to Growth: A tech-hub grows fast until housing prices drive out the talent, hitting a plateau.
• Eroding Goals: Reducing environmental targets because the current infrastructure “cannot handle” the cost of meeting them.
• Growth and Underinvestment: Demand for public transit grows, but the city delays purchasing new trains, leading to a collapse in service quality.

• Fixes that Fail: Widening a road to fix traffic leads to “induced demand,” creating more traffic than before.
• Shifting the Burden: Using temporary “band-aid” repairs on aging water pipes instead of replacing the fundamental infrastructure.
• Escalation: Two neighboring districts compete to offer the lowest taxes to attract firms, eventually bankrupting their own public services.

• Tragedy of the Commons: Unregulated groundwater extraction by individual buildings leading to the sinking of the entire city’s foundation.
• Success to the Successful: High-income neighborhoods receive the best schools, which increases property values and further concentrates educational funding there.
• Accidental Adversaries: A transit agency and a housing department both try to optimize their own KPIs, but end up blocking each other’s long-term goals.

Workshop: Mapping the Logic

Tool: ncase.me/loopy

The Mission

Translate your “Wicked Problem” from a vague concept into a functional Causal Loop Diagram.

1. Set the Frame

• Define Boundaries: What is in your system? What is exogenous “noise”?
• Temporal Scale: Is this a 24-hour traffic cycle or a 20-year gentrification process?

2. Build the Logic

• Extract Nodes: Identify the 5–8 core variables (Social & Technical).
• Connect & Loop: Draw the arrows. Is the relationship (+) or (-)?
• Find the Archetype: Can you spot a “Fix that Fail” or “Limits to Growth”?

Workshop: What we are doing

Focus on the Logic, not the math.

🎯 The Focus

• Relationships: How A affects B.
• Feedback: Identifying reinforcing and balancing loops.
• Communication: Making the “invisible” logic visible.

🛑 Not Today

• Stocks: We aren’t measuring exact “buckets” of volume.
• Flow Rates: We aren’t calculating precise liters/people per second.
• Validation: We are mapping theories, not yet proving them.