Assignment - Agent-Based Investigation
The latter portion of this module has introduced the field of agent-based modelling (ABM), and discussed its use as a form of experimentation in crime science: as a means of testing the viability of behavioural theories, or as a method for carrying out 'what if' analysis that would not be practical in the real world. We have introduced the general principles of the approach, and implemented our own ABM of urban crime in Python. In this assignment, you will investigate the behaviour of this ABM and seek to draw insights with respect to real world behaviour.
The ABM we have developed is a simple model of urban crime, in which citizen agents move around an urban area, making decisions about whether to commit acquisitive crime at each location they visit. The area is also populated by police officers, whose presence can prevent crime from happening. The overall mechanism, therefore, models crime events as the result of interactions between offenders, targets and (the absence of) guardians, as proposed within Routine Activity Theory.
The task in this assignment is to use this model to investigate an issue of theoretical or practical interest for urban crime. You should formulate the question in experimental terms, and carry out simulations in such a way as to reflect the varying experimental conditions. You should then analyse the quantitative output, using the techniques introduced in this module and its predecessor; that is, using descriptive statistics, inferential tests (e.g. T-tests) and statistical modelling approaches (e.g. regression). Following this, you should interpret the results and their real-world implications.
The particular question or issue that you choose to examine is up to you. Several aspects of the model that we have built can be manipulated to examine their effects on the output: the numbers of citizens and police officers, the attractiveness of targets, and the radius within which officers have a deterrent effect, for example. Furthermore, you are free to adapt the model code to add extra features to the model or to reflect different behaviours: you might change the movement rules to something more complex than a simple random walk, for example, or implement 'patrols' by police. The assignment is designed so that a solid piece of work can be done using the code provided, but so that there is also scope for a more ambitious approach (which will be rewarded).
Report -
Your report should take the form of a Jupyter Notebook, containing both explanatory text and the code required to perform the simulations and analysis. There is no prescribed structure for the report, though it would be sensible to follow a standard approach: introduction, model, analysis, discussion. The report does not need to take the form of an 'essay', and things like bullet points are acceptable - however, it should be written using formal language and should be readable as a document in its own right. You should use references if and when they are relevant.
There is no need to explain in detail the features of the basic model in the form that it has been provided (though it would be sensible to refer to its main principles) - that may be assumed. However, any additional features that are added should be explained fully, both in the text and via comments in the code.
Criteria -
This assessment has five main components, which are the criteria against which the work will be assessed. These do not correspond to any particular section, and will be considered across the report as a whole.
Design of investigation -
You should outline the issue(s) that you are going to investigate, explaining why they constitute meaningful questions from a theoretical or practical perspective. You should then describe how you are going to investigate this using the model: what you will manipulate, and how you will examine its effect.
Implementation -
You should perform model simulations in order to examine its behaviour in the way that you have specified. You may do this by simply varying the parameters of the model provided, or by adapting the code in order to add further features. Credit will be given here for the sophistication of the approach.
Analysis -
The output of the model will be quantitative in form, and you will need to analyse it in order to understand the results of the model. You should do this using statistical techniques, drawing on those that have been introduced in Probability, Statistics and Modelling I & II. As usual, you should justify the approach taken and credit will be given for both the selection of approaches and their execution.
Interpretation -
You should use the output of the model, and the quantitative analysis of it, to draw conclusions in relation to your overall investigation. You should discuss the strength of your findings and express their implications in real-world terms. Consideration should be given here to the advantages and limitations of the agent-based approach.
Presentation -
The report should be written clearly and attractively, making use of the Jupyter Notebook form. Credit will be given here for clarity of expression and the display of quantitative information.