Software agents that mimic the behaviour of real children are been used to detect paedophile grooming behaviour on the Internet.
Called ChatNannies, the technology is the brainchild of IT consultant Jim Wightman, of Wolverhampton in the UK. The software runs with thousands of sub-programs - dubbed nanniebots - which log onto chatrooms and make conversations with youngsters there.
While engaging in pop-culture inanities, the program analyses the behaviour of other participants in a chat room looking for tell-tale signs of grooming or slip-ups that suggest a user is an adult and not the child he might claim to be.
The software emails suspicious conversations to Wrightman, who screens theses messages prior to passing details of suspect users (IP address etc.) and transcripts over to the police.
We imagine the software generated plenty of false positives, particular when the technology was first used.
Wightman told New Scientist his tip-offs have led to police investigations, though this remains unconfirmed, perhaps unsurprisingly. It's standard police practice (at least in the UK) not to talk publicly about active investigations.
No-one has detected the bots, Wightman added. Given the short duration and limited scope of chat room conversations this is plausible.
ChatNannies uses a neural network program build up knowledge and refine their responses. Each of the bots is programmed to display distinct "personalities". The technology is capable of updating itself with pop culture trivia gleaned from the Internet.
Wightman doesn't want to sell his software or turn his technology into a business. However he would welcome financial support for government-run child protection bodies so that he can expand his system beyond four servers at his work. ®