top of page

What is Modelling?

There are different ways of leveraging data and knowledge collected by an organisation. There are roughly three ways statistics can be used to leverage data:
​
  1. Explanatory statistical modelling: The way it is usually applied by criminologists explanatory modelling refers to  “the application of statistical models to data for testing causal hypotheses about theoretical constructs”. For example, using statistics to discover which recorded behaviours are correlated with multiple violent crimes. This does not involve predicting outcomes for new sets of behaviours, but rather finding out which features are associated, and to which degree, with outcomes. 

  2. Prioritisation through retrieval or ranking: Using a database search or a search engine to pick out data entries that conform to a set of specified factors to rank or prioritise results. For example, ranking suspects in the area by cumulative Cambridge Crime Harm Index (CCHI), or finding nominals who have been suspected of three or more rapes that have yet to be charged and prioritising their cases for interventions and investigation. This also does not involve prediction or creation of any new features or information. 

  3. Prediction or automatisation through machine learning: Using data to train a model to embody patterns in the data and produce new information, such as labels, or text, or predictions of likely future events. Special care needs to be taken when using predictive modelling in decision making, record augmentation, especially when concerning individuals, communities, or geographies. 

​

It can be useful to examine using explanatory modelling and retrieval as potential solutions, as they directly leverage knowledge within the contained data in a more transparent and explainable way, before attempting modelling. The techniques for explanatory and predictive modelling are not mutually exclusive, and familiarity and insights gathered from examining patterns in data can inform predictive modelling and vice versa. Our ethical framework is mainly concerned with predictive modelling using ML for the cases which have direct impact on individuals or communities. 

  • What is ML? IBM Intro

  • Making Friends with Machine Learning by Cassie Kozyrkov

  • Lifecycle of an AI project by Cassie Kozyrkov

‎pre-modelling-flowchart.‎001.jpeg
bottom of page