http://www.computermilk.net/2009/06/27/emotional-humanoid-robots/ Straight from the teats of technology Mon, 19 Dec 2011 22:21:20 +0000 hourly 1 http://www.computermilk.net/2009/06/27/emotional-humanoid-robots/comment-page-1/#comment-10572 Correy Allen Kowall Mon, 19 Dec 2011 22:21:20 +0000 http://www.computermilk.net/?p=136#comment-10572 In some sense the notion of an emotion offers a type of behavioral closure for problems for which no immediate solution is available. If we simply use a set of identifiable goals that are not to be extended, modified, or extensively reinterpreted by an agent, it will frequently encounter situations which there is no obvious solution to seeking those goals. However if we give an agent a goal and a emotional disposition with respect to that goal it can regulate the amount of effort it extends toward trying to extend the set of things it knows how to do such that it can bridge the gap toward fulfilling an innate imperative. People have long argued the case that emotive capacity is important for interactions between machines and users. I am not sure how I feel about this. However from a strictly behavioral standpoint the argument presented above has a solid basis in the type of environment AI practitioners say requires a Partially observable Markov decision process. This type of environment has information which isn't available at all points and at all times; therefor a given decision to reach some goal cannot be regarded as optimal because not all the information necessary to make that claim is known. In some sense the notion of an emotion offers a type of behavioral closure for problems for which no immediate solution is available. If we simply use a set of identifiable goals that are not to be extended, modified, or extensively reinterpreted by an agent, it will frequently encounter situations which there is no obvious solution to seeking those goals. However if we give an agent a goal and a emotional disposition with respect to that goal it can regulate the amount of effort it extends toward trying to extend the set of things it knows how to do such that it can bridge the gap toward fulfilling an innate imperative. People have long argued the case that emotive capacity is important for interactions between machines and users. I am not sure how I feel about this. However from a strictly behavioral standpoint the argument presented above has a solid basis in the type of environment AI practitioners say requires a Partially observable Markov decision process. This type of environment has information which isn’t available at all points and at all times; therefor a given decision to reach some goal cannot be regarded as optimal because not all the information necessary to make that claim is known.

]]>