It is undeniable that AI and Machine Learning are becoming the integral tools for many processes and activities across all industries and fields of study. This leaves no exception for criminology, as police forces are starting to adopt it globally, from using it as tools in solving criminal cases to developing it into a crime predicting model. But did we think this through?
AI performance statistics could be over glamorized where its implementation flaws could be neglected. For example, the developed AI model can be so advanced that some model has a 90% accuracy in predicting crime (E&T, 2022). However, the development also revealed that their data’s bias was difficult to account for. For example, in their previous version of the model, when it comes to predicting which individuals are likely to be involved in a violent crime in Chicago, more than half of black male population of the age between 20 and 29 years old were listed on it. This is due to the bias in the arrest data where people from the black community were often wrongly arrested due to the history of racial profiling. In fact, the data regarding negligible crime such as marijuana arrest and traffic stops had to be omitted, and more data regarding homicides and assaults had to be added to remove the bias (Verma, 2022). This suggests that any advancement or development of an AI-based, or a data-driven system in enforcing justice must be closely monitored, where higher quantity of data does not always lead to better performance.
This is not to neglect the potentials and past achievements of AI in criminology. An AI-based system could help law enforcers look up crucial information of an individual or could speed up the analysis of a surveillance video by finding a specific individual with a specific outfit color in a lengthy video full of people (Lee, 2020). In the forensics’ field, an AI based model was developed and managed to identify model of footwear from footwear impressions at a higher range of accuracy than a normal individual (Bennett and Budka, 2021). However, this level of accuracy cannot beat the footwear experts’ accuracy at almost 100%, due to some factors that are difficult for AI to account for, such as how footwear impressions can vary with how old the shoes are. This suggests that although AI could be used in performing basic analysis tasks, the more complex tasks should be left to a person with expertise to solve.
AI technology may not be perfect to run on its own in enforcing justice and solving crime, but it could serve as a great tool for humans to do so with great care. It is not a question whether humans or AI should be used exclusively in criminology, but how both can work together in improving the world.
Sources
Bennett, M. and Budka, M., 2021. We trained AI to recognise footprints, but it won’t replace forensic experts yet. [online] The Conversation. Available at: <https://theconversation.com/we-trained-ai-to-recognise-footprints-but-it-wont-replace-forensic-experts-yet-161686> [Accessed 6 October 2022].
E&T. 2022. AI ‘predicts crime with 90 per cent accuracy’. [online] Available at: <https://eandt.theiet.org/content/articles/2022/07/ai-predicts-crime-with-90-per-cent-accuracy/> [Accessed 6 October 2022].
Lee, J., 2020. How AI technology is helping solving crime. [online] Police1. Available at: <https://www.police1.com/police-products/police-technology/police-software/articles/how-ai-technology-is-helping-solving-crime-7vb577RVrWliW57H/> [Accessed 6 October 2022].
Verma, P., 2022. The never-ending quest to predict crime using AI. [online] The Washington Post. Available at: <https://www.washingtonpost.com/technology/2022/07/15/predictive-policing-algorithms-fail/> [Accessed 6 October 2022].
