People sometimes treat information as context-free, but no set of data or form of information can really be understood without enough environmental details to interpret its meaning accurately. An example can be found in a story by Matt Wood, who wrote about an algorithm that “predicts crime a week in advance, but reveals bias in police response.” ¹ The article goes on to say: “Data and social scientists from the University of Chicago have developed a new algorithm that forecasts crime by learning patterns in time and geographic locations from public data on violent and property crimes. The model can predict future crimes one week in advance with about 90% accuracy.” What is different about this model compared to prior ones was the move away from purely spatial models. “Communication networks respect areas of similar socio-economic background,” writes James Evans in the same article, not just formal boundaries.
Context is a major reason for why the algorithm performed better with these data than using other models. The model developers did not worry about street boundaries, but focused on “areas of similar socio-economic background”.¹ They split the topography into tiles of about 93 square meters without respect for neighborhood or political boundaries.“ The model can predict future crimes one week in advance with about 90% accuracy.” The model was also tested using data from seven similarly urban cities with similar results.
As data the model used “two broad categories of reported events: violent crimes (homicides, assaults, and batteries) and property crimes (burglaries, thefts, and motor vehicle thefts).” The reason for these categories was their greater likelihood of being reported, which meant that the data would be more consistent and reliable. The goal of the algorithm was not as a tool to encourage police concentration in particular areas, but to give a better tool to let researchers “evaluate police action in new ways”. The algorithm also enables voters and politicians new ways to look at a complex problem.
1: Wood, Matt. 2022. ‘Algorithm Predicts Crime a Week in Advance, but Reveals Bias in Police Response | Biological Sciences Division | The University of Chicago’. 30 June 2022. https://biologicalsciences.uchicago.edu/news/algorithm-predicts-crime-police-bias. Photo by GeoJango Maps on Unsplash