How AI Can Help Tackle Collective Decision-Making

Collective decision-making is hardly a perfect science. Broken processes, data overload, information asymmetry, and other inequities only compound the challenges that come from large, disparate factions with different goals trying to work together. And the tools that often help with decision-making—data analysis, scenario planning, decision trees, and so on—can falter in the face of the scale and complexity of the biggest problems that groups and leaders face.

This is where AI can help, and is helping. With its ability to analyze vast troves of data about the status quo, understand group preferences, run sophisticated simulations to evaluate hundreds of future possibilities against those preferences, and facilitate consensus-building among participants, AI can be a powerful tool for all leaders facing complex decisions, especially those that must be made collaboratively.

One field already taking advantage of AI’s collective decision-making support is city planning. Three years ago, we started working with the United States Conference of Mayors to understand how AI is helping cities solve their most pressing challenges. Along the way, we studied the story of the German city of Hamburg, which has addressed a housing crisis exacerbated by an influx of refugees.

In 2016, Hamburg partnered with MIT Media Lab, the creator of an AI platform called CityScope. The platform allows urban planners to collect and digest the needs and preferences of swaths of residents, simulate hundreds of building scenarios, identify hidden opportunities, and find common ground among conflicting factions. By demonstrating how CityScope is working in Hamburg, we hope to show leaders across governments, nonprofits, universities, and corporations how they can harness data and AI to democratize and improve their decision-making processes and outcomes.

The Crisis in Hamburg

In 2016, Germany decided to welcome 1 million refugees from the Middle East, and Hamburg was tasked with finding housing for an anticipated 80,000 families in a city of under 2 million people. At the time, the city had been stuck in unproductive conversations about zoning laws for decades, struggling to build enough houses for its own residents.

Three challenges tend to come to the fore in situations of collective decision-making, and Hamburg was no exception:

Processes and incentives are broken.

The traditional process to get things done (in this case, to get new housing built) involves dozens of steps and institutions, each with its own procedural logic and internal culture. A single technical impropriety can delay a project by months, if not years. Stakeholders have no incentive to come together. In Hamburg, the city struggled to get anything built because each project required the approval of a dozen bureaucracies, often sclerotic and opaque.

Information is increasingly abundant and isn’t equally distributed.

Every decision (in this case, about specific developments or zoning laws) involves vast amounts of information across domains—from resident preferences and technical documents to traffic and usage metrics. Furthermore, processes are often expressed in long, detailed, technical documents that the average person cannot be expected to understand. Those with more resources have the time, money, and expertise to get the information they need to form an educated opinion while other community members do not.

Other inequities.

Dominant players possess a wide array of tools to block transformations. In this case, property owners can stop urban developments with historic preservation regulations, minimum lot size requirements, height restrictions, etc. In Hamburg, the only people who participated in decision-making about housing and zoning were wealthier, older homeowners. Successive mayors had launched a few outreach campaigns to get other parts of the community engaged in zoning debates, but none had gained traction.

How CityScope Helped

AI can help to solve these challenges and improve the way that groups make decisions together. Ariel Noyman, one of the key engineers behind CityScope, told us that his team designed the platform to fulfill four key functions:

• Insight: Building a dynamic model of social, economic, and environmental conditions through comprehensive data collection, an environmental scan, transaction analytics, and sentiment analysis; visualization with feedback.

• Prediction: Identifying needs and simulating the impact of alternative interventions by evaluating thousands of “what-if” scenarios.

• Transformation: Iterating possible interventions into validated paths of action.

• Consensus: Engaging stakeholders in a shared, facilitated decision-making process to reach a unified vision of the future.

Here’s how these functions played out in the process of working with residents and planners in Hamburg to move the needle forward.

The first step involved gathering as much data as possible. CityScope drew on data about housing and zoning laws, but also economic development, purchasing patterns, city-wide events and amenities, transportation and infrastructure, employment opportunities, demographic diversity, environmental impact, safety, and more. It also administered surveys to residents to gather their preferences.

To deliver the first function, insight, the platform then correlated the core dimensions of housing—density and diversity of people—with performance indicators such as energy use, safety, resident preferences, and so on. With that data, the platform analyzed the relationship between housing and quality of life in the status quo.

Then the platform went on to make predictions about current trends and hypothetical transformations, identifying constraints that might be valuable to alter along the way. CityScope highlighted the systematic underuse of commercial properties, for instance. It also highlighted the areas where public services were most likely to be strained and those with the most capacity to welcome new residents.

Once the analysis was done, CityScope displayed the results in a simple, easy-to-understand diagram to help citizens and other participants to easily see how potential changes would affect key performance indicators that reflect the common priorities of the city’s residents.

The labels on the vertical bars indicate that the community cared about environmental impact, energy performance, infrastructure, innovation, and overall livability (each of these indicators aggregated hundreds of metrics). The height of the fill of each bar indicates the city’s current performance on each priority (higher is better), and the horizontal lines on each bar indicate the performance for each of these priorities for a given future scenario. The bars with red fill indicate areas that would change for the better; those that are green show the city has already met the targets. Through this chart, CityScope helps residents understand their situation (insight again), extrapolate current trends (prediction again), evaluate possible alternatives (transformation), and find common ground (consensus).

In Figure 2, these metrics are aggregated by street into a 3D map of the city. Positive changes are in green, negative ones in red, and alternatives can be modeled at the scale of the individual street, neighborhood, or the city as a whole. In Hamburg, these representations allowed users to understand trade-offs and find ways to overcome them. For example, the visualization made starkly clear the differences between the preferences of affluent communities, where proposals that threatened lower-density zoning received the worst ratings, and the city as a whole, where building more houses in underdeveloped areas to welcome refugees scored well.

CityScope then helped the residents find the best way to accommodate these conflicting preferences. By evaluating competing proposals, it demonstrated that wealthier neighborhoods could benefit from more houses provided that a new metro line were also built in the process, thereby paving the way to consensus.

The vizualizations also allow CityScope to yet again collect people’s preferences, this time on the trade-offs and competing proposals. In Hamburg, the team administered surveys and organized workshops across the city with an augmented reality (AR) version of the platform (see Figure 3).

The AR interface allowed participants to collaborate with CityScope to see the implications of their choices. Hamburg residents would come into the room and rearrange the LEGO-like bricks representing residential units, office buildings, parks, and other urban amenities in a specific zone, redesigning the city one brick at a time (Figure 3). When participants made these changes, the digital projection updated in real time to show how the proposed changes would affect quality of life. The platform also connected these changes with the zoning laws that would make them possible, bridging the gap between the LEGO game and policymaking.

The interface also allowed participants to collaborate with each other in workshops across the city. That way, CityScope became a platform of direct community engagement, where technical and non-technical people, with different levels of understanding, gathered around the table to understand the impact that their common decisions would have on their city (the consensus function again). In Hamburg, 5,000 residents participated in CityScope workshops in 2016, considerably more than at any conventional town hall. What’s more, by targeting diverse communities across the city, the CityScope team has managed to attract a representative sample of the population, rather than just the older, wealthier citizens who have historically participated in the city’s urban planning.

Making Decision-Making More Democratic

Through its use of AI, CityScope addresses the key challenges we identified in city planning, and in group decision-making more generally:

First, they circumvent slow bureaucratic processes. By aggregating all the relevant data into a dynamic model, CityScope analyzes trade-offs better than city officials ever could on their own because it integrates all perspectives and tests thousands of alternatives. The result is a considerably streamlined process, and also one that takes more perspectives into account more accurately.

Second, they solve the problem of information overload and asymmetry. By intaking and processing vast troves of data and then providing clear visuals and methods for interacting with and sharing them, CityScope removes the informational barriers that favor those with more resources over those who lack money, time, or expertise, giving anyone the opportunity to understand and propose changes.

Third, CityScope enables the full community to find a path to consensus, not just the elite. Residents may not agree on this or that housing project, but they can find common ground around shared priorities. By shifting the focus of deliberation from specific projects or laws to broader priorities for the city, CityScope reframes the discussion towards the bigger picture. Larson calls the platform a “consensus machine” for a reason.

Eighteen months after the partnership with CityScope began, Hamburg had not just housed thousands of refugees: It had strategically distributed them across the city to maximize social cohesion, economic opportunity, and community resilience. Since then, the city has been integrating CityScope into its decision-making processes more broadly, for transportation, energy use, and environmental regulation. When Russia invaded Ukraine in 2022, Hamburg had the tools to welcome tens of thousands of refugees in a fraction of the usual time And the United Nations now funds a project exporting CityScope to other cities that face an unexpected influx of refugees.

Beyond CityScope

Humans are not especially good at processing immense amounts of information and translating it into policy. They struggle to understand complexity and, left to their own devices, seldom find common ground on contentious issues. CityScope shows that AI can help.

However, by themselves, platforms like CityScope cannot solve contentious problems. Most group decisions remain inescapably prone to conflict and while AI can help us understand and navigate tradeoffs, it cannot make tradeoffs disappear. No algorithm, however sophisticated, can replace a culture of healthy disagreement and mutual respect. In Hamburg, it was the residents and their leaders who made this conversation productive, not just the platform itself.

Further, these tools only matter if they are integrated into the right ecosystem. In Hamburg, the city could not take advantage of CityScope without also reforming bureaucratic hurdles that prevented certain spaces from being repurposed or built upon. CityScope helped identify and prioritize those reforms, but without them, the remainder of its functions would have been futile. While AI can streamline processes and help us make better decisions together, only with the right leadership can these tools translate into collective action.

Nevertheless, tools which combine sophisticated simulations with direct engagement can change how we make decisions in all kinds of institutions—cities, but also corporations, universities, or non-profits. Platforms like AnyLogic, FlexSim, and Visual Components are already developing similar tools for corporations, a trend that is likely to accelerate in the years to come. Far from a substitute for human decision-making, these platforms will offer a powerful complement to it: a way to harness data at the service of common goals.

Copyright 2025 Harvard Business School Publishing Corporation. Distributed by The New York Times Syndicate.

Explore AAPL Membership benefits.