Recent Uses of Machine Learning

Two Examples of two machine learning technology:

1. The Harm Assessment Risk Tool (HART) used by Durham policing aids custodial decision making. Individuals are assessed on having either a low, moderate or high risk of committing further crimes within a two-year period, based on algorithms exploring combinations of variables including offending history, known as predictor values. Those identified as a moderate risk and likely to commit a non-serious offence are considered for referral to a rehabilitation programme called Checkpoint. Although, the technology only has an overall 63% accuracy rate, the technology focuses on the avoidance of false negatives (individuals predicted as low risk of re-offending but go on to commit a new serious offence and in doing so is 98% accurate. However, issues arise regarding bias in decisions within a system that particularly focuses on predictor values concerning socio-demographic characteristics and geological area (University of Cambridge, 2019).

Further Reading click here

2. Machine learning algorithms in Live Facial Recognition (LFR) technology are being adopted by London police in UK trials. By detecting faces from live camera feeds and identifying individuals wanted by authorities by comparing that data to digital images held on the LFR offenders watch list database, it is hoped that LFR will work towards improved safety and crime prevention. However, concerns arise regarding privacy issues and low accuracy rates. The face off report, independently produced by the Big Brother privacy group, suggested that less than 2% of all facial matches were true matches (Fogden, 2018).

Share this: