CounterCrime - Using Counterfactual Explanations to Explore Crime Reduction Scenarios

Area of Science:

• Data Science
• Urban Planning
• Criminology

Background:

• Crime analysis is complex, influenced by socioeconomic and urban factors.
• Understanding these impacts is crucial for effective public policy and crime mitigation.
• Counterfactual explanations offer a way to explore hypothetical scenarios for crime reduction.

Purpose of the Study:

• To introduce CounterCrime, a visual analytics tool for crime analysis.
• To leverage counterfactual explanations for generating "what-if" crime scenarios.
• To aid decision-makers in understanding and mitigating regional crime rates.

Main Methods:

• Developed CounterCrime, a visual analytics tool with interactive metaphors.
• Organized analysis across city, region group, and regional levels.
• Employed a greedy strategy for variable selection and similarity grouping for counterfactuals.

Main Results:

• The tool facilitates exploration of counterfactual scenarios at multiple granularities.
• Identified key variables influencing crime rates and recommended them for intervention.
• Validated findings using crime data from São Paulo, Brazil, through case studies.

Conclusions:

• CounterCrime offers a novel perspective on crime analysis through "what-if" scenario exploration.
• The tool can anticipate changes to improve regional safety and inform policy.
• Context-specific interventions are necessary, as effective scenarios may vary between regions.