Imagine that you build a robot and programme it with all the same inputs that people have nowadays then weight the inputs for selfish gene vehicles; greatest benefit at lowest cost.

2.1 million, thefts, robberies and crimes against vehicles were reported in 10/11.  221,000 were charged.  The home office reported that about 77% are convicted.  The maximum sentence is 7 years for theft, so we’ll assume 3.5 years as the likely sentence.

So if you’re a thief and you get reported, you have a 10% chance of being charged and a 77% chance of being convicted; that about an 8% conviction rate.  The expected “cost” to the ‘perps’ of these crimes is then about 3 months in prison (3.5 years x 8%).

So you’ve already got a 92% chance of getting away with it, assuming no pesky kids in a mystery van of course.

The 3 months in prison does seem to prevent individuals from thieving and destroying cars in the main.  However in a riot situation, the likelihood of being caught is going to decrease isn’t it, making it almost certain that you’ll get away with it.

So are the benefits worth the cost?  Well if my robot has mates and has been watching TV and reading magazines, it’s going to want a lot of electronics, gadgets and booze as well as some fancy trainers in order to increase his likelihood of attracting a mate to build some baby robots with.

Given that my robot would work on game theory and with selfish programming, at the critical decision point when its Blackberry buzzed to offer the choice, I think that for my Lootatron 500™, riot=1 would be very a very likely output.

Being a conscientious robot owner though, perhaps I’d need to programme it with a few moral routines or additional conditional adverse impacts, to offset the increased attractiveness of looting…

Or to ensure integration with society, I could load it with Vista, to slow it down enough to get caught and report all its activity to Microsoft.