Crime Analytics and the Role of Dynamic Simulation Models


Nick Malleson

Associate Professor

School of Geography, University of Leeds, UK

nickmalleson.co.uk

These slides: http://surf.leeds.ac.uk/presentations.html

Abstract

The 'data deluge', coupled with related 'smart cities' initiatives, have led to a proliferation of information about how people use urban areas. These data are potentially extremely valuable for a number of domains, including crime science. This talk will argue that, because it focuses explicitly on the actions and behaviours of individuals, agent-based simulation is the most appropriate tool for the study of many criminal phenomena. The talk will present two example applications of the methodology to the study of burglary, before outlining wider applications of agent-based modelling to the study of urban dynamics more broadly.

Overview

Dynamic Simulation - Agent Based Modelling

Explanatory Agent-Based Crime Models

Predictive Agent-Based Crime Models

Predictive Policing (?) & Ethical Implications

Modelling Broader Urban Dynamics

Quantifying Ambient Populations

Dynamic Data Assimilation

Why Model Crime?

Exploring theory ('explanatory' models)

Simulation as a virtual laboratory.

Linking theory with crime patterns to test it.

Making predictions ('predictive' models)

Forecasting the impacts of social / environmental change.

Exploring aspects of current data patterns.

Why is it Difficult?

Extremely complex system:

Attributes of the environment (e.g. individual houses, pubs, etc.).

Personal characteristics of the potential offender and/or victim.

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.

Diagram of regression

Introduction to ABM

Aggregate v.s. Individual

'Traditional' modelling methods work at an aggregate level, from the top-down

E.g. Regression, spatial interaction modelling, location-allocation, etc.

Aggregate models work very well in some situations

Homogeneous individuals

Interactions not important

Very large systems (e.g. pressure-volume gas relationship)

Diagram of regression

Introduction to ABM

Aggregate v.s. Individual

But they miss some important things:

Low-level dynamics, i.e. “smoothing out” (Batty, 2005)

Interactions and emergence

Unsuitable for modelling complex systems

Diagram of the sims

Introduction to ABM

Systems are driven by individuals

(cars, people, ants, trees, whatever)

Bottom-up modelling

An alternative approach to modelling

Rather than controlling from the top, try to represent the individuals

Account for system behaviour directly

Autonomous, interacting agents

Represent individuals or groups

Situated in a virtual environment

A termite mound.
Attribution: JBrew (CC BY-SA 2.0).

Emergence

One of the main attractions for ABM

"The whole is greater than the sum of its parts." (Aristotle?)

Simple rules → complex outcomes

E.g. who plans the air-conditioning in termite mounds?

Possible to prove the with simple computer programs

Conways 'Game of Life'

Emergence

Why is it important?

Key message: Complex structures can emerge from simple rules

Emergence is hard to anticipate, and cannot be deduced from solely analysis of individuals’ behaviour

You could not work out what a termite mound would look like by dissecting a termite.

Emergence is a characteristic of complex systems

Individual-level modelling is focused on understanding how macro-level patterns emerge from micro-level through the process of simulation.

How is this relevant to crime?

Better Representations of Theory?

Environmental Criminology theories emphasise importance of

Individual behaviour (offenders, victims, guardians)

Individual geographical awareness

Environmental backcloth

Better Representations of Space?

Example of GIS data

Lots of research points to importance of micro-level environment

Brantinghams' environmental backcloth

Crime at places research (e.g. Eck and Weisburd, 1995; Weisburd and Amram, 2014; Andresen et al., 2016)

Agent-Based Modelling - Appeal

Modelling complexity, non-linearity, emergence

Natural description of a system

Bridge between verbal theories and mathematical models

Produces a history of the evolution of the system

Agent-Based Modelling - Difficulties

Will I play with the truck, or the duck?

(actually he played with his trains...)

Tendency towards minimal behavioural complexity

Stochasticity

Computationally expensive (not amenable to optimisation)

Complicated agent decisions, lots of decisions, multiple model runs

Modelling "soft" human factors

Need detailed, high-resolution, individual-level data

Individual-level data

Crating an 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.

Modelling agent behaviours

Many behaviours are hard / impossible to model

Choose those that are the most important. Cannot include everything!

Some can be very simple - e.g. threshold-based rules (Kennedy, 2012)

IF hunger IS ABOVE hunger_threshold THEN search_for_food
OTHERWISE do_something_else

These are the most common (Birks et al. 2012, 2013; Dray et al. 2008; Groff 2007a,b; Hayslett-McCall, 2008)

More advanced cognitive frameworks exist

Beliefs, Desires, Intentions (Bratman et al., 1988)

PECS (Schmidt, 2000).

Who else is doing this?

Small, but growing, literature

(apologies to the many who are missing!)

Birks et al. 2012, 2013;

Groff 2007a,b; Hayslett-McCall, 2008

Liu et al. (2005)

Me! (Malleson et. al ...)

A special issue of the Journal of Experimental Criminology entitled "Simulated Experiments in Criminology and Criminal Justice" (Groff and Mazerolle, 2008b)

ABM Explanatory Example (Birks 2012)

Birks (2012) model environment

Explanatory: exploring theory

Randomly generated abstract environments

Navigation nodes (proxy for transport network)

Potential targets (houses)

ABM Explanatory Example (Birks 2012)

One type of agent: potential offenders

Behaviour is controlled by theoretical 'switches'

Example crime theories

Rational choice perspective (decision to offend)

Routine activity theory (how they move and encounter targets)

Geometric theory of crime (how they learn about their environment)

ABM Explanatory Example (Birks 2012)

Validation against stylized facts:

Spatial crime concentration (Nearest Neighbour Index)

Repeat victimisation (Gini coefficient)

Journey to crime curve (journey to crime curve)

Results:

All theories increase accuracy of the model

Rational choice had a lower influence than the others

Simple / abstract model:

Not directly applicable to practice

But simplicity allows authors to concentrate on theoretical mechanisms

Realistic backcloth might over-complicate model (Elffers and van Baal, 2008)

ABM Predictive Example

Predictive: exploring the real world

ABM to explore the impacts of real-world policies

Urban regeneration in Leeds

Example of MasterMap Topographic data
Input layers into the virtual environment
PECS motives and actions
PECS motives and actions

ABM Predictive Example

Awareness space test

Did it work?

Aggregate results

Results of model calibration

Did it work?

Halton Moor

Malleson, N., A. Heppenstall, L. See, A. Evans (2013) Using an agent-based crime simulation to predict the effects of urban regeneration on individual household burglary risk. Environment and Planning B: Planning and Design 40 405-426. [DOI: 10.1068/b38057]

 

 

 

 

 

 

 

 

 

 

 

 

Did it work?

Journey to Crime

Results of model calibration

ABM Predictive Example

Scenario Results

Malleson, N., A. Heppenstall, L. See, A. Evans (2013) Using an agent-based crime simulation to predict the effects of urban regeneration on individual household burglary risk. Environment and Planning B: Planning and Design 40 405-426. [DOI: 10.1068/b38057]

Predictive Intelligent Policing

This is not predictive policing

It could be, in the future, maybe, but not yet

This is a useful tool for exploring the crime system.

It might lead to a better understanding of:

How different behavioural characteristics (offender, victim, or guardian) influence crime outcomes

How the physical infrastructure can be organised to discourage crime

How patrol routes might be most effective

Etc.

Guardian article
Arthur, C. (2010) Why Minority Report was spot on. The Guardian Wednesday 16 June 2010

Ethics

This is not minority report!

"if an agent-based model is good enough, can it tell me where someone will commit their next crime?"

We can't (and wouldn't want to!) predict when/where/who will commit a crime.

Academics have a role to set the boundaries on what is ethically acceptable

Dynamic simulation models have great potential, we need to make the case that they can be used responsibly

Particularly relevant in the 'big data' / 'smart cities' era (e.g. informed consent)

Data protection / privacy legislation might not be sufficient (often there are clauses for research with crime data)

Other Uses: Modelling Everyone

Quantifying the Ambient population

Daily flows of people have a significant impact on crime rates

But usually have to fall back on residential population for want of data

“The general patterns of movement towards and away from activity nodes such as work or school locations, major shopping areas, entertainment districts or bedroom suburbs provide a very general image of where crimes will concentrate” (Kinney et al., 2008)

We need better models of the ambient population!

Malleson, N. and M. Andresen (2015) The impact of using social media data in crime rate calculations: shifting hot spots and changing spatial patterns. Cartography and Geographic Information Science 42(2)
Hotspot disappearing from the city centre
Malleson, N. and M. Andresen (2015) The impact of using social media data in crime rate calculations: shifting hot spots and changing spatial patterns. Cartography and Geographic Information Science 42(2)
Hotspot disappearing from the city centre
Malleson, N., and Andresen, M.A. (2016) Exploring the impact of ambient population measures on London crime hotspots. Journal of Criminal Justice 46 pp 52-63
The process of calibration
Typical model development process

Agent-Based Models of the Ambient Population

Agent-based modelling could do better?

Could combine diverse, up-to-date data with a model to create the most accurate estimate of the real world

But:

Urban systems are complex

They will diverge rapidly from model forecasts

ABM development typically uses 'waterfall' approach

We lack tools to calibrate models dynamically

Dynamic Data Assimilation

Used in meteorology and hydrology to constrain models closer to reality.

Try to improve estimates of the true system state by combining:

Noisy, real-world observations

Model estimates of the system state

Should be more accurate than data / observations in isolation.

Diagram of dynamic data assimilation and an ABM

Early prototype:
Ensemble Kalman Filter (EnKF)

Maths

Broad literature, but generally tied to mathematical models (e.g. differential equations and linear functions)

Working with a mathematician to do the hard work!

In all its glory: Ward et al., (2016)

Advantages

Similar to Kalman Filter (best in class)

But better for nonlinear systems

Ensemble Kalman Filter - Basic Process

1. Forecast.

Run an ensemble of models (ABMs) forward in time.

Calculate ensemble mean and variance

2. Analysis.

New 'real' data are available

Integrate these data with the model forecasts to create estimate of model parameter(s)

Impact of new observations depends on their accuracy

3. Repeat

Ensemble Kalman Filter (EnKF)

Outline

Ward, J., A. Evans, N. Malleson (2016) Dynamic calibration of agent-based models using data assimilation. Royal Society Open Science. 3:150703. (open access). [DOI: 10.1098/rsos.150703]

Simulating Urban Flows (surf)

Big Data, ABM, Smart Cities, Ambient Populations

A city-wide dynamic ABM, constantly re-calibrated to streaming data

New insights into urban mobility patterns and footfall estimates.

surf.leeds.ac.uk/

Future Work: DUST

Data Assimilation for Agent Based Models: Applications to Civil Emergencies

Recently funded by the ERC (Starting Grant), €1.5M over 5 years

Starts on 1st January 2018

For more information:

https://erc.europa.eu/news/erc-2017-starting-grants-highlighted-projects

Shameless Advertising

The Leeds Institute for Data Analytics

http://lida.leeds.ac.uk/

New, ~ £10M investment in a multidisciplinary centre for data analytics

Big role for agent-based modelling and crime science

E.g. forthcoming seminar: Difficulties with reproducible data analytics: Pitfalls in predictive policing

Summary

Dynamic Simulation - Agent Based Modelling

Explanatory Agent-Based Crime Models

Predictive Agent-Based Crime Models

Predictive Policing (?) & Ethical Implications

Modelling Broader Urban Dynamics

Quantifying ambient Populations

Dynamic Data Assimilation

References

Andresen, M. A., Linning, S. J., and Malleson, N. (2016). Crime at Places and Spatial Concentrations: Exploring the Spatial Stability of Property Crime in Vancouver BC, 2003–2013. Journal of Quantitative Criminology, pages 1–21.

Birks, D., Townsley, M., and Stewart, A. (2012). Generative Explanations of Crime: Using Simulation to Test Criminological Theory. Criminology, 50(1):221–254

Birks, D., Townsley, M., and Stewart, A. (2013). Emergent Regularities of Interpersonal Victimization: An Agent-Based Investigation. Journal of Research in Crime and Delinquency.

Bratman, M. E., Israel, D. J., and Pollack, M. E. (1988). Plans and resource-bounded practical reasoning. Computational Intelligence, 4(3):349–355.

Dray, A., Mazerolle, L., Perez, P., and Ritter, A. (2008). Policing Australiaâ’s heroin drought: using an agent-based model to simulate alternative outcomes. Journal of Experimental Criminology, 4(3):267–287.

Eck, J. and Weisburd, D. (1995). Crime places in crime theory. In Eck, J. and Weisburd, D., editors, Crime and Place, pages 1–33. Criminal Justice Press.

Elffers, H., and P. van Baal. 2008. “Realistic Spatial Backcloth Is Not That Important in Agent Based Simulation Research: An Illustration from Simulating Perceptual Deterrence.” In Artificial Crime Analysis Systems: Using Computer Simulations and Geographic Information Systems, 19–34. Hershey, PA: Information Science Reference.

Groff, E. (2007a). Situating Simulation to Model Human Spatio Temporal Interactions: An Example Using Crime Events. Transactions in GIS, 11(4):507–530.

Groff, E. R. (2007b). Simulation for Theory Testing and Experimentation: An Example Using Routine Activity Theory and Street Robbery. Journal of Quantitative Criminology, 23(2):75–103.

Hayslett-McCall, K., Qui, F., Curtin, K. M., Chastain, B., Schubert, J., and Carver, V. (2008). The Simulation of the journey to residential burglary. In Liu, L. and Eck, J., editors, Artificial Crime Analysis Systems: Using Computer Simulations and Geographic Information Systems. Information Science Reference, Hershey, PA.

Kennedy, W. G. (2012). Modelling Human Behaviour in Agent-Based Models. In Heppenstall, A. J., Crooks, A. T., See, L. M., and Batty, M., editors, Agent-Based Models of Geographical Systems, pages 167–179. Springer Netherlands.

Liu, L., Wang, X., Eck, J., and Liang, J. (2005). Simulating crime events and crime patterns in RA/CA model. In Wang, F., editor, Geographic Information Systems and Crime Analysis, pages 197–213. Idea Publishing, Reading, PA.

Malleson, N. and A. Evans (2013) Agent-Based Models to Predict Crime at Places. In G. Bruinsma and D. Weisburd (Eds) Encyclopedia of Criminology and Criminal Justice pp 41-48 . Springer.

Malleson, N. “Agent-Based Modelling of Burglary.” School of Geography, University of Leeds, UK, 2010.

Malleson, Nick, Alison Heppenstall, and Linda See. “Crime Reduction through Simulation: An Agent-Based Model of Burglary.” Computers, Environment and Urban Systems 34, no. 3 (2010): 236–50.

Malleson, Nick, Linda See, Andrew Evans, and Alison Heppenstall. “Optimising an Agent-Based Model to Explore the Behaviour of Simulated Burglars.” In Theories and Simulations of Complex Social Systems, edited by Vahid Dabbaghian and Vijay Kumar Mago, 179–204. Intelligent Systems Reference Library 52. Springer Berlin Heidelberg, 2014.

Ward, J., A. Evans, N. Malleson (2016) Dynamic calibration of agent-based models using data assimilation. Royal Society Open Science. 3:150703. (open access). [DOI: 10.1098/rsos.150703]

Weisburd, D. and Amram, S. (2014). The law of concentrations of crime at place: the case of Tel Aviv-Jaffa. Police Practice and Research, 15(2):101–114.

Crime Analytics and the Role of Dynamic Simulation Models


Nick Malleson

School of Geography & Leeds Institute for Data Analytics,

University of Leeds

nickmalleson.co.uk

These slides: http://surf.leeds.ac.uk/presentations.html