Having thought the matter through I will therefore be contacting DARPA with my proposal that this can be designed out of future robots by the simple expedient of giving them deeply malign personalities that alienate and enrage their operators. Within 5 years human interfaces will advance to the point where Mark 1 RILF (Robot I'd Like to Frag) can be implemented relatively simply by giving it a preselected canned personality from a suitable list:
- Game Show Host
- TV Evangelist
- Irish Chat Show Host
- White Supremacist
- Life Coach
Within 10 years and using feedback from people who have worked with the MK 1 RILFs we can foresee MK 2, which instead of having a canned personality would study its operators behavior and identify weaknesses it could exploit using custom personality traits and behavior, much in the manner of a school bully. It would also attempt to look just enough like a human to trigger an uncanny valley response from its owners. In order to be truly unbearable the MK2 would also have non-verbal behaviors in case its owners obtained suitable ear protection. Depending on the identified weakness of the target/owner such behaviors could include biting, hiding small objects, lying, 'accidentally' waking its owner and replacing live ammunition with blanks. Eventualy with the MK3 we could even look forward to robots that steal your phone and send offensive messages to your spouse.