This chapter attempts to establish a level of plausibility for increasing the automation of traffic violation detection and deterrence. Although the application of 'artificial intelligence' (AI) in this area is problematic, the computation of an AI analysis enables various hypotheses to be tested without any commitment to a practical implementation. Although knowledge-based analysis does not necessarily have a direct on-site application, it contains important heuristic guides to prototype development. Most traffic violation occurs because many if not most drivers do not view it as criminal behaviour. Effective enforcement operates by a deterrence through detection principle. One possibility for investigation is that an automatic system, using some aspects of AI, can provide an effective and sensitive intervention tool. Many current enforcement programmes are rather ineffective on a meaningful scale, because of their transiency and uneven coverage. Both semi-automatic and automatic systems have been shown to reduce violation rates significantly. An automatic system's main tasks would be to detect speeding offenses and dangerous driving. The question is not whether automatic policing is feasible, but whether it can be made acceptable. It is therefore a political question. After discussing possible present and future applications of AI, the author concludes that AI will be unable to help traffic policing.
Abstract