The taming of the sleuth – problems and potential of autonomous agents in crime investigation and prosecuting

Over the past decade, autonomous agents have become an increasingly mature technology to collect data over the internet. Intended originally for commercial purposes, e.g. comparing prices on internet auction sides on behalf of their owners, submitting offers or forming strategic alliances to benefit from bulk discounts, the same attributes that make them suitable in the commercial sector have recently also attracted the interest of police forces worldwide. Using agent technology for the collection of evidence on the internet is increasingly becoming the new paradigm for the policing cyberspace, and is sold as the last best hope to cope with the overwhelming amount of data produced online. The implications, both for civil liberties and the law of evidence, are considerable. First, it has the potential to change dramatically the numerical relation between law enforcement officers and citizens. Second, traditional evidentiary safeguards, e.g. prohibitions against entrapment or illegal searches, might either not apply in law to autonomous agents, or might be inapplicable due to the technological features of this technology (e.g. the ability to self-destruct when detected) Third, agent technology allows to “outsource” police work to private parties, increasing the problem of evidential safeguards. This paper will present briefly the technological parameters of agent technology in criminal investigations and point out some of the implications for due process and civil liberties. In the second part, we propose an amendment to certain formal language for agents that allow them to “understand” some of the legal safeguards that regulate offline evidence collection. Hohfeld’s schema of rights and duties has already been used to teach autonomous agents basic features of private aw. The focus in the past was however on the right-duty relation, leaving the other categories unused. we will show how the concepts of “liberty” and “privilege” can be formally represented in an agent programming language, and the potential they have for preventing evidence collection by virtual police officers that would be deemed inadmissible when carried out by a real life officer