Case Study: CityScope Predict — The City Council Decides

The Scenario

It is a Thursday evening in Millhaven, a city of 340,000 in a politically diverse swing state. The city council chambers are full. Local news cameras are rolling. The council is holding its final public hearing before voting on whether to approve a one-year pilot of CityScope Predict, a predictive policing system that would use historical crime data, environmental factors, and time-series analysis to generate weekly "risk maps" directing police patrols.

You have followed CityScope Predict since Chapter 1. You have analyzed how it makes predictions (Chapter 7), evaluated its fairness properties (Chapter 9), examined the privacy implications of its data sources (Chapter 12), and considered governance frameworks for its oversight (Chapter 13). Now the decision point has arrived.

Tonight, five people will testify before the council. Each represents a different perspective on the same question: Should Millhaven adopt CityScope Predict?

The Testimonies

Testimony 1: Chief Maria Rodriguez, Millhaven Police Department

"Council members, I have been in law enforcement for twenty-three years. I have watched my officers make patrol decisions based on gut feelings, anecdotal experience, and — I will be honest — sometimes based on assumptions that are not fair to every community we serve.

CityScope Predict offers us a chance to replace those subjective judgments with data. The system does not know a suspect's race. It does not have a bad day. It does not get tired at the end of a double shift and start cutting corners. It analyzes patterns — time, location, type of incident — and tells us where to direct our limited resources.

In the three cities where CityScope Predict has been deployed, reported property crimes dropped by an average of 11% in the first year. Response times improved by 8%. I have spoken with chiefs in those cities, and they are satisfied.

Are there concerns? Of course. Every tool has limitations. That is why we are proposing a one-year pilot with quarterly reviews. If the data shows the system is not working, or is producing unacceptable disparities, we stop. But I think we owe it to our residents — including residents in high-crime neighborhoods who are asking us for more help, not less — to try."

Testimony 2: Councilmember Aisha Thompson, District 4

"I want to be clear: I am not anti-technology. I use AI in my daily life. I am not opposed to data-driven policing in principle. What I am opposed to is deploying a system that has not been independently validated for our city, using our data, with our demographics.

I represent District 4. It is the most racially diverse district in Millhaven. It has also been the most heavily policed district for as long as I have lived here — and I have lived here my entire life. My constituents have been subject to stop-and-frisk at rates three to four times the city average. They have watched their children get pulled over for 'matching a description.' They have sat through traffic stops that had nothing to do with traffic.

Now we are being asked to feed all of that history — all of that biased enforcement — into a machine and let it tell our police where to go next. The chief says the algorithm does not see race. Maybe not. But it sees zip codes. It sees prior arrests. It sees the data produced by twenty years of over-policing my district. And it will send the police right back to the same corners.

I have read the research. Predictive policing systems in Los Angeles, Chicago, and New Orleans produced documented racial disparities. Two of those cities have discontinued their programs. Why are we eager to repeat their mistakes?

I am asking the council to vote no — or at the very least, to require an independent racial equity audit before deployment, not after."

Testimony 3: Dr. Marcus Chen, Data Scientist and University Professor

"I am neither an advocate nor an opponent of CityScope Predict. I am here to help the council understand what the system can and cannot do.

First: the accuracy question. The vendor reports 73% accuracy in predicting 'elevated crime risk' at the census-block level on a weekly basis. That sounds good. But accuracy depends on what you are measuring. If a patrol is directed to a neighborhood and makes an arrest there, did the algorithm succeed? Or did the arrest happen because the officers were there in greater numbers? This is the feedback loop problem, and it is real.

Second: the fairness question. I have reviewed the vendor's technical documentation. The system does not use race as an input. But it uses zip code, prior incident density, and socioeconomic indicators — all of which are correlated with race in Millhaven. I ran a preliminary analysis using publicly available crime data and census demographics. The system's predicted 'high risk' zones overlap heavily with the city's historically redlined neighborhoods. This is not surprising — it is exactly what the published research predicts.

Third: the validation question. CityScope Predict was validated in three other cities, none of which share Millhaven's demographics, policing history, or crime patterns. Validation is not transferable. A system that works well in one context can fail in another.

My recommendations: If the council moves forward, require independent validation using Millhaven-specific data before deployment. Require monthly fairness reports disaggregated by neighborhood demographics. And establish a clear, measurable threshold — defined in advance — for what level of racial disparity would trigger suspension of the program."

Testimony 4: Sarah Okonkwo, Director, Millhaven Community Justice Alliance

"I am here tonight because the people who will be most affected by this decision are the least likely to be in this room. They are working evening shifts. They are taking care of children. They do not have the time or the resources to attend a city council meeting on a Thursday night.

I have spent the past month talking to residents in the neighborhoods where CityScope Predict would be most active. Here is what they told me:

They want to feel safe. They want police to respond quickly when they call. They want drug activity off their corners and guns out of their children's schools. They do not want to be abandoned.

But they also want to be treated with dignity. They want to walk to the store without being stopped. They want their teenagers to drive home without being pulled over. They want to know that the police are in their neighborhood to protect them, not to profile them.

CityScope Predict cannot distinguish between these two things. It can tell you where to send officers. It cannot tell officers how to behave when they get there. And in communities where trust between residents and police has been damaged by decades of aggressive enforcement, sending more officers — even with the best algorithmic justification — can make things worse, not better.

I am not asking the council to reject technology. I am asking the council to invest in trust first. Fund community policing. Fund violence interrupters. Fund the diversion programs that have been on your agenda for two years with no funding. Build the relationship, and then — maybe — the data tools can help strengthen it."

Testimony 5: James Hartley, CEO, CityScope Analytics

"Thank you for the opportunity to present. I want to address some of the concerns directly.

Our system does not replace officers' judgment. It is a decision-support tool. It provides information. What officers do with that information is determined by department policy and training — neither of which we control.

On the question of bias: we take it seriously. We have invested significantly in fairness research. Our system uses regularization techniques to reduce the influence of proxy variables. We publish annual transparency reports. And we have offered to work with an independent auditor selected by the city — at our expense — to conduct a pre-deployment equity analysis.

On the question of feedback loops: we agree this is a real concern. Our system includes a 'novelty score' that flags when predictions are driven primarily by enforcement-generated data rather than independent incident reports. It is not a perfect solution, but it is a meaningful safeguard.

On the question of past failures in other cities: context matters. Los Angeles and Chicago deployed earlier versions of predictive policing technology with less sophisticated fairness tools and less robust oversight. We have learned from those experiences. Our product is not the same as PredPol circa 2013.

I am not asking you to trust us blindly. I am asking you to evaluate our system on its merits, with appropriate oversight, in a controlled pilot. If it does not meet the standards you set, shut it down. You have that power."

The Decision

The Millhaven city council must now vote. The options are:

Option A: Approve the pilot as proposed — Deploy CityScope Predict for one year with quarterly reviews as Chief Rodriguez recommends.

Option B: Approve with conditions — Deploy the pilot only after specific prerequisites are met (independent validation, community oversight board, equity audit, defined suspension criteria).

Option C: Reject the proposal — Do not deploy CityScope Predict. Invest resources in alternatives (community policing, diversion programs, violence interrupter programs).

Option D: Defer — Table the vote for six months. Commission an independent study. Hold additional community forums in affected neighborhoods.

Analysis Framework

Use the following framework to structure your analysis:

1. Stakeholder Mapping

For each of the five speakers, identify: - Their primary concern - What evidence they rely on - What they want the council to do - Whose interests they represent - What they might be overlooking

2. Evidence Evaluation

Assess the claims made by each speaker: - Which claims are supported by data or research cited in this chapter? - Which claims are assertions without cited evidence? - Where do the speakers agree? Where do they disagree about facts vs. values?

3. Constitutional Analysis

Apply the constitutional concepts from Section 17.3: - Does CityScope Predict raise due process concerns? If so, for whom? - Could its deployment produce disparate impact? How would you measure this? - What procedural safeguards would a court likely require?

4. Accountability Mapping

Using the accountability framework from Section 17.4: - If CityScope Predict directs officers to a neighborhood and an innocent person is stopped, who is accountable? - If the system produces documented racial disparities after six months, who has the authority and responsibility to respond? - Are there accountability gaps in the current proposal? How could they be closed?

5. Your Recommendation

Take a position. Which option (A, B, C, or D) would you vote for, and why? Your recommendation should: - State your position clearly - Address the strongest counterargument to your position - Include at least three specific, measurable conditions (if you choose Option B) - Explain how your recommendation serves the interests of the most affected communities

Discussion Questions

1. Perspective shift: Which testimony did you find most persuasive? Least persuasive? Now consider: would your answers change if you were a resident of a heavily policed neighborhood? A police officer? A crime victim?

2. Missing voices: Who was not represented in the testimonies? Whose perspective is missing? How might their input change the analysis?

3. Framing effects: Notice that the question is framed as "Should the city adopt this system?" Could the question be reframed in a way that opens up more possibilities? For example, "What does public safety look like in Millhaven, and what role — if any — should predictive technology play in achieving it?"

4. Transferability: Could this same scenario play out around a different AI system — say, an AI hiring tool, a content moderation algorithm, or a healthcare triage system? What elements of the debate would be the same, and what would be different?

5. The timeline question: Councilmember Thompson wants a racial equity audit before deployment. The chief wants to start the pilot and evaluate along the way. Is there a principled basis for choosing one timing over the other, or is this purely a political judgment?

Mini-Project Options

Option A: Council simulation. In a group of five, assign each person one of the testimony roles. Each person prepares a two-minute statement and a one-minute rebuttal. After all statements, the group votes and discusses what drove their decision.

Option B: Policy brief. Write a 500-word policy brief to the Millhaven city council recommending Option B (approve with conditions). Specify at least five concrete conditions, explain the rationale for each, and describe how compliance would be verified.

Option C: Community engagement plan. Design a community engagement process for the six months before a decision is made. Specify how you would reach residents in the most affected neighborhoods, what information you would provide, what questions you would ask, and how their input would influence the final decision.

Option D: Comparative analysis. Research what two real cities did when faced with a similar decision about predictive policing. Compare their processes, outcomes, and lessons learned. What would you recommend that Millhaven learn from their experiences?

Note on Sources

This case study is a Tier 3 illustrative composite. The scenario, characters, and city of Millhaven are fictional. However, the arguments, data points, and dynamics are drawn from documented real-world experiences in cities that have evaluated or deployed predictive policing systems, including Los Angeles, Chicago, New Orleans, Santa Cruz, and Oakland. The testimonies reflect arguments that have been made by real stakeholders in real proceedings.