Decision support tools

Approaching urban challenges with indicators, dashboards, and decision support

Smart City’s brain

The Back-end

The heuristics, logic, and models.

The Front-end

Dashboards, visualizations, and stakeholder apps.

GOAL

To reduce the “Cognitive Load”

without oversimplifying the “Wicked” reality.

• Remind them of the “Brain” analogy. A brain is useless if it can’t move the body.
• Decision Support Tools are the nervous system. They take the data from our models and present it so a Mayor, a Planner, or a Citizen can say “If we do X, then Y happens.”
• Crucial point: A good DST doesn’t hide complexity; it manages it.

History & Evolution

From the Cockpit to the City’s Policy

Origins: Business Intelligence

The Corporate “Cockpit”

KPI Culture

“If you can’t measure it, you can’t manage it” - Peter Drucker

Performance Monitoring: Dashboards were designed for CEOs to track sales, inventory, and efficiency in real-time.

The city as a machine that can be “tuned” for efficiency.

• Start with the roots. Dashboards didn’t start in cities; they started in factories and boardrooms.
• The goal was efficiency and profit maximization.
• Ask: “Does a city work like a factory?” (Hint: No).

The Shift to Urban Planning

Importing the Logic

The Promise

Tech giants sold the idea of the “City Operating System.”

• Real-time feedback: Moving from 5-year census data to millisecond sensor readings.

The Reality

• Complexity: Cities are not closed systems, unlike factories.
• Wicked Problems: Moving a variable in a city (e.g., traffic) affects everything else (pollution, retail, housing).

eg. mycelium network - Transition: When we brought these tools to cities, we brought the “efficiency” mindset with them. - Challenge: Cities are open, complex systems, not closed production lines. - This creates tension between “optimizing flow” and “creating place.”

The Shift to Urban Planning

Monitoring

“What is happening now?”

• Traffic counters
• Air quality sensors
• Energy grids

Analyzing

“Why is it happening?”

• Correlating data
• Finding patterns
• Identifying bottlenecks

Simulating

“What if we change X?”

• Digital Twins
• Impact assessment

Decision Support!

• This is the key evolution. We refrain from just looking at the speedometer.
• We are now building “Flight Simulators” where we can crash the plane (city) safely in digital space before making changes in physical space.
• This is the definition of a Decision Support Tool.

Indicators & Proxies

The Map is not the Territory

The Trap of Abstraction

The “Index” Problem

• Liveability Index: 85/100
• Happiness Score: 7.2
• Walkability Score: 94

Why it fails

Mathematically sound, practically useless.

• If the score drops to 82, do we fix potholes or add schools?

Abstract indexes hide the levers of change.

• Explain the difference between an Indicator (raw, valid signal) and an Index (aggregated, abstract score).
• Ideally, we want indicators that point to specific physical actions.
• An “Index” is good for ranking cities in a magazine, but bad for running a city.

The Trap of Abstraction

“The map is not the territory.”

— Alfred Korzybski

• The Origin: Alfred Korzybski was a scientist and philosopher who wanted to remind us that our description of reality is never the same as reality itself.
• Defining the Map: In our context, the “Map” is the Digital Twin, the GIS layers, the census data, and the AI’s predictive model. It is an abstraction.
• The Value of Omission: A map is only useful because it leaves things out. If a map of Oslo was as detailed as Oslo itself, it would be a 1:1 scale and completely useless.
• The Danger for Planners: The “Trap” occurs when we start making decisions based only on the map’s logic, forgetting the territory’s complexity.
• Example: A transport model (The Map) says a new highway will reduce commute times. But the territory (The People) responds with “induced demand,” filling the road immediately. The map failed to capture human behavior.
• AI and “Hallucinations”: When an AI gives a confident but wrong answer, it is “believing its own map.” It is calculating the probability of words or pixels, not the physics of the street.
• The Planner’s Duty: Our job is not just to build better maps; it is to be the person who constantly walks the territory to see what the map missed.
• Question for the class: “If the Digital Twin shows a neighborhood is ‘optimally served’ by transit, but the residents feel isolated, which one is ‘true’?”

The Unintended Consequences

Optimizing for the wrong thing

The Metric

“Average Vehicle Speed” or “Intersection Throughput”

Optimizing for flow efficiency.

The Result

Stroads & Highways

• High speed = Low interaction.
• A street with “perfect flow” has no life (commerce, lingering, playing).

Goodhart’s Law

“When a measure becomes a target, it ceases to be a good measure.”

• Example of Traffic vs Place.
• If you optimize for one variable (speed), you inadvertently kill another (place quality).
• This is why choosing what to measure is a political act, not just technical.

Good vs. Bad Indicators

Feature	Good Indicator	Bad Indicator
Actionable	“Potholes per km” (We can fill them)	“Road Quality Index” (Vague)
Relevant	Measures what matters to stakeholders	Measures what is easy to count
Comparable	Standardized units (kg, m, $)	Arbitrary scales (1-5 stars)
Transparent	Clear calculation method	“Black Box” algorithm

• Go through the criteria.
• Emphasize “Proxy”: Sometimes we can’t measure the thing (e.g., Safety), so we measure a proxy (e.g., Streetlight lux levels).
• Always ask: Is the proxy actually correlated?

Dashboard’s 101

Anatomy of a Dashboard

Indicators

The “Vital Signs”

• Real-time feeds
• Historical trends
• Key Performance Indicators (KPIs)

Simulation

The “What-If” Engine

• Slider controls
• Scenario toggles
• Impact forecasting

Narrative

The Context

• Annotations
• Storytelling flow
• “So what?” (Insights)

• Refer to your Data Viz notes: “Identifying patterns, trends, and outliers.”
• In an urban context, a dashboard shouldn’t just be a “report card” of the past. It should be a “flight simulator” for the future.

The “Misleading” Interface

The Aesthetic Trap

“If it looks high-tech, it must be true.”

• Shiny 3D maps and “Minority Report” interfaces can mask poor data quality.
• Cognitive Bias: We trust beautiful things more than ugly truths.

The Black Box

• False Precision: Displaying “Safety: 87.42%” implies a level of certainty that doesn’t exist.
• Hidden Algorithms: Who decided the weightings?
• Truncated Realities: What is not lying on the dashboard? (e.g., happiness, social capital).

• Connect this to your Viz Pitfalls: “Misleading graphs, improper scaling.”
• In Smart Cities, there is a risk of “Technological Solutionism.”
• As curators of these tools, the students must ensure the visualization doesn’t lie about the model’s uncertainty.

Different Lenses for Different Roles

High-Level Ops

The Pilot

- Goal: Real-time management. - Timeframe: Seconds/Minutes. - Needs: Alerts, live status, red/green indicators.

Strategic Planning

The Architect

- Goal: Long-term policy. - Timeframe: Years/Decades. - Needs: Trends, aggregate data, scenarios.

Public Participation

The Citizen

- Goal: Transparency & Input. - Timeframe: Daily life. - Needs: Relatability, localized data, feedback channels.

• Different users need different “Front-ends” for the SAME underlying reality.
• A technician needs a “Technical” viz. A citizen needs a “Human-centric” viz.
• Workshop Preview: We will look at how to tailor your tool for a specific stakeholder.

Data Viz Best Practices

Principles for effective dashboards

Gestalt Principles

Only some

Principle	Description
Proximity	Elements that are close to each other are perceived as a group.
Similarity	Items that share visual characteristics are seen as related.
Continuity	The eye follows lines and curves smoothly without abrupt changes.
Closure	Incomplete shapes are perceived as complete wholes.
Connectedness	Items connected by lines or other visual cues are seen as related.
Colour	Utilises hues and shades to convey meaning and hierarchy.
White Space	Strategic use of empty space to enhance readability and focus.

• Quick recap of Gestalt (from the Data Viz full lecture).
• Use these principles to group content on your dashboard.
• Don’t just scatter charts; use proximity and similarity to create logical sections.

Choosing the Right Chart

NYTimes Visual Vocabulary

What NOT to do

“Mistakes, we’ve drawn a few.”

The Don’ts

• Truncated Axes: Starting a bar chart at 50% to exaggerate a 1% difference.
• Spaghetti Charts: Too many lines making trends unreadable.
• Correlated Scales: Forcing two unrelated variables onto dual axes to imply causation.

The Wrong Type

• Using a Line Chart for categorical data (implies continuity where there is none).
• Using a Pie Chart for 20 categories.

• Honesty in Data Viz is crucial for Decision Support.
• If you truncate an axis, you might convince a Mayor to spend millions on a problem that is actually a minor fluctuation.

Narrative in the Data

Workshop

The Green Mobility Dashboard