Should autonomous weapons need a reason to kill?

Purves et al. argue against deploying automated weapons because they fail to act for the right reason. Given that soldiers do not necessarily act in an ideal way, I argue that it is morally preferable to deploy autonomous weapons that are incapable of acting for the wrong reason over combatants that are likely (although not guaranteed) to act for the right reason (i.e. regular troops). Preference for regular troops based solely on reasons for acting is justified only in the case of an ideal conception of these troops. This claim is, however, vulnerable to an additional argument favouring the deployment of autonomous weapons as a means of preserving human life. The possibility that the behaviour of autonomous weapons is a direct expression of the reasons given by their human commanders for their deployment is also considered, but rejected as a suitable reason to favour deploying regular or even ideal troops over autonomous weapons, for reasons discussed.