The Police in their use of AI intelligence tools are worried the software may seem to increase prejudice as its algorithms drill down on data to target certain groups.
The Iconoclast in me wonders if this is where PC PC’s (politically correct police constables) clash with number crunching data analytics. There are many ways to be statistically correct but a statistician once told me the most popular; determine the outcome you want then make the stats fit the outcome. Skewed stats of more recent times has sometimes seen complete ‘about turns’ in policing policy.
An example of stats skewing……a tea plantation company boasts on its packets of tea that its employees are paid twice the national average. We’ll applaud that and we think straight away of those hard working pickers in the field being handsomely rewarded. However, if you drill down in to the figures we find that the median worker is paid the national average but the upper management are grossly overpaid. Obviously that skews the average pay across the employees. So if you want to make your company look good, quote average and not median salaries.
AI training programs hopefully ditch the predetermination of my statistics friend’s answer and come up with some useful and relevant intelligence to assist the police in their task. I can almost guarantee they will not be politically correct. Could it be this that ruffles the feathers of the politically correct brigade sufficient to want to question the outcome of the AI intelligence tool?
I have listened to the calls of the parents of victims of stabbings in London to increase stop and search. Politicians argue that it disproportionately affects certain groups and leads to tension in the various communities so should not be used, or at least used proportionally and intelligence lead (whose?). Maybe AI is the answer? If it can more accurately determine the person to stop, rather than the police or the denying politician, perhaps it would be seen as the honest broker in the process? Lets hope so.