Behavioural Analytics

Can organised crime be prevented by predicting people’s characteristics, networks, and intentions from the way they have previously behaved? This new programme takes advantage of technological advancements and data from ‘social signals’ to assess the threat of actors and make inferences about their likely actions. This programme is led by Paul Taylor, Stacey Conchie and David Ellis at Lancaster University.

Work in this programme is focused on understanding what can be inferred about groups from streams of behavioural data. Our efforts are unashamedly theory driven and much of the programme’s focus is on deriving more efficient measures through a solid understanding of what aspects of behaviour matter (i.e., correspond) to the inference in question. This approach seeks to dramatically decrease the complexity of the data model rather than find more efficient ways to handle the complexity. Our work includes examining digital traces of behaviour that are produced through a person’s use of technology and the internet.

Research Focus

Specifically, the early focus of this programme conducts original research exploring:

  • How criminal group activity to be understood in real time.
  • How behavioural markers can reveal characteristics about organised crime groups.
  • How a group’s behaviour can predict how successful they’ll be in fulfilling their criminal intent.
  • How language markers in digital communication can predict the effectiveness of criminal groups.

 Research Findings

 So far, this programme has generated insights such as:

  • Providing groups with mis-information when undergoing a group task reduces their language co-ordination.
  • The age and gender of a person can be predicted from their smartphone behaviours such as screen time and application use.

A synthesis of papers which use language style matching to understand conversation patterns has also been created and published here:

Topics of future investigation

 Future directions of the programme aim to assess behavioural analytics in order to:

  • Understand the relative value of social sensor data for remote assessment.
  • Explore ‘atypical’ or ‘suspicious’ patterns of behaviour that may indicate an imminent attack.
  • Assess behaviour longitudinally and identify behavioural anomalies which may be a precursor of criminal behaviour.

Principle Investigator

Paul Taylor


Lancaster University