Crime Research in Geography Resource allocation: we helped design the police Basic Command Unit families for the national resource allocation formula in the 1990s. More recently been advising West Yorkshire police. Predictive policing: assessment of potential methodologies project. Crime prediction.
Crime Research in Geography Ongoing collaboration with SaferLeeds [local police/government crime prevention partnership]. Builds on work using microsimulation and gravity modelling to look at offender-to-target burglary flows. Uses Agent-Based Modelling ( ABM ).
Project Modelling burglary in Leeds. Ongoing relationship with Safer Leeds Crime and Disorder Reduction Partnership Provide essential data. Expert knowledge to supplement criminology theory.
Why Model? Exploring theory ( explanatory models) Simulation as a virtual laboratory: Linking theory with crime patterns to test it. Making predictions ( predictive models) Forecasting social / environmental change. Exploring aspects of current data patterns through prediction.
Spatially patterned therefore predictable(?) Why burglary? Spatio-temporally variations key to understanding system. System with history of qualitative theorisation that needs testing. Data good (geocoding, reporting). Largely individually initiated in UK therefore don t need so much data-poor social interaction modelling. Should be possible to run what if tests (specifically, urban regeneration in Leeds). Significant component of fear of crime in UK.
Why difficult? Extremely complex system: Attributes of the individual houses. Personal characteristics of the potential offender. Features of the local community. Physical layout of the neighbourhood. Potential offender s knowledge of the environment. Traditional approaches often work at large scales, struggle to predict local effects Computationally convenient. But cannot capture non-linear, complex systems.
Individual-level Crime Modelling: Agent-Based Models (ABM) Create an urban (or other) environment in a computer model. Stock it with buildings, roads, houses, etc. Create individuals to represent offenders, victims, guardians. Give them backgrounds and drivers. See what happens. Much better understanding of relationship between: Individuals (offenders, victims, and guardians). Their routines. Street-level environment. Perceptions of urban areas. Inherently spatial and dynamic.
Coding an Agent Agent Class Has an update method called each iteration, eg. move(), trade(). Has a position. Has a list of all other agents and can get their position. Can communicate with other agents if necessary. Environment Class Has environmental conditions. Calls the agents to update. Agents might, for example, trade with their nearest neighbours.
Agent-Based Modelling Autonomous, interacting agents Represent individuals or groups Situated in a virtual environment
Commonly Used Platforms Netlogo: http://ccl.northwestern.edu/netlogo/ Repast: http://repast.sourceforge.net/ MASON: http://cs.gmu.edu/~eclab/projects/mason/ Ascape: http://ascape.sourceforge.net/ ABLE: http://www.research.ibm.com/able/ Agent Analyst: http://www.spatial.redlands.edu/agentanalyst/
NetLogo basics Two windows: Interface and Procedures Interface contains graphical elements Procedures are user-defined functions
Better Representations of Theory Environmental Criminology theories emphasise importance of Individual behaviour (offenders, victims guardians) Individual geographical awareness Environmental backcloth Offender Crime Guardian Victim Routine Activity Theory Geometric Theory of Crime Rational Choice Perspective
Examples of Agent-Based Crime Models Abstract Environment Predictive Birks et al. (2012) Residential burglary Simple behaviour Switch on/off theoretical components Model dynamics reflect expected (theoretical) outcomes? Spatially Realistic Environment Explanatory Groff (2007) Street robbery Interactions of victims and offenders Simple behaviour Highlight high-crime intersections Nick & Andy Residential burglary GIS data Advanced (?) behavioural model
Example 1 Burglary (Explanatory) Birks at al. (2012) Randomly generated environments Theoretical switches Compare results to expected outcomes: Spatial crime concentration Repeat victimisation Journey to crime curve Results: All hypotheses are supported Rational choice has lower influence Theory Enabled Disabled Routine activities Rational choice Awareness space Agents assigned a home and routine paths Victim attractiveness (based on risk, reward, effort) Dynamic awareness alters offender decision-making Figure 1. Example Model Environment Random movements Homogenous target attractiveness Uniform environment awareness Taken from Birks et al. (2012)
Example 2 Street Robbery (Predictive) Groff (2007) Street robbery in Seattle Interactions of victims and offenders Simple behaviour Highlight high-crime intersections
Example 3 Burglary (Predictive) Virtual Environment Physical objects: houses, roads, bars, busses etc. Social attributes: communities Virtual victims and guardians Virtual Burglar Agents Use criminology theories/findings to build realistic agent behaviour PECS
Modelling Behaviour PECS Framework Needs Lifestyle, Sleep, Drugs Cognitive map of environment Decision process leads to burglary Drug level Social level Sleep level Personal preference, p Personal preference, p Personal preference, p m = p / s m = p f(t) / s m = p f(t) / s 1. PECS Behaviour -> Decision to Burgle 2. Choose community to search 3. Travel to community and search Time of day, t Time of day, t Intensity of drugs motive Intensity of social motive Intensity of sleep motive Determine Strongest Motive Agents Burglary Decision Process 1. Attractiveness 2. Social difference 3. Previous successes 4. Distance Plan Actions Communities in the Agent s Cognitive Map 4. Choose property to burgle Agent s Thought Process 1. Collective Efficacy (community) 2. Occumpancy levels (community) 3. Accessibility 4. Visibility 5. Security 6. Traffic volume (road) Objects in the Environment
Simulation Video
Interesting Finding Halton Moor Result Halton Moor area significantly under predicted by model Explanation Motivations of burglars in Halton Moor Model failures can help to indicate where we misunderstand the real world
Result Forecasting Burglary after Simulation Test the effects of a large urban regeneration scheme A small number of individual houses were identified as having substantially raised risk Why? Location on main road In the awareness space of offenders Slightly more physically vulnerable Urban Regeneration
How much realism? Abstract Environment Tractable simulations, better able to understand fundamental rules Explore theory unencumbered by geographical complexity (e.g. Elffers & Baal, 2008) Not applicable to the real world? GIS Environment More accurate representation of the real world Forecasts / predictions / scenarios Simulations are no easier to understand than the real system, and less accurate?
Who else is doing crime simulation? Researchers: Elizabeth Groff: street robbery Daniel Birks: burglary Patricia Brantingham et al.: Mastermind (exploring theory) Lin Liu, John Eck, J Liang, Xuguang Wang: cellular automata Books / Journals: Artificial Crime Analysis Systems (Liu and Eck, 2008) Special issue of the Journal of Experimental Criminology (2008):``Simulated Experiments in Criminology and Criminal Justice'
Future Dynamic data We are currently looking at mining twitter feeds for population numbers around the city, and travel routes. More socio-economic data coming online all the time. Utilise this dynamically to dampen errors. Ethical issues Currently anonymize and randomise real offender data. Could we imagine a day when resources were directed to predictions of real people? Up to us to take a lead on what we do and don t find acceptable.
More information General info: http://crimesim.blogspot.com/ Play with a simple tutorial version of the model: https://github.com/nickmalleson/repastcity Papers: http://www.geog.leeds.ac.uk/people/n.malleson http://www.geog.leeds.ac.uk/people/a.evans