Piaget's ideas about assimilation and accommodation are being used today in order to better program machines. In order for robots to complete tasks in a variety of situations and environments, they must be able to understand the task as opposed to a programmed set of actions. Machines must accommodate and assimilate meaning in order to adapt their methods to different situations and tools.
An example is cutting. This task can be performed in a variety of ways with many different tools. Humans are able to understand what it means to cut and are able to perform this task in a variety of situations. Robots need to be programmed step by step in a series of actions that will cut an object if location, orientation, tool and other specific criteria are met (Arbib, 1992). Our robots do not have knowledge but programmed action. Watch the video below from the show Nathan for You for an example of how robots currently perform tasks.
An example is cutting. This task can be performed in a variety of ways with many different tools. Humans are able to understand what it means to cut and are able to perform this task in a variety of situations. Robots need to be programmed step by step in a series of actions that will cut an object if location, orientation, tool and other specific criteria are met (Arbib, 1992). Our robots do not have knowledge but programmed action. Watch the video below from the show Nathan for You for an example of how robots currently perform tasks.
The task of taking off Nathan's pants would only work in the controlled, artificial environment presented. The pants were heavily modified and the robot could only perform a series of programmed motions. It did not understand what pants, buttons or any number of schemata involved that would allow it to perform this action under different circumstances.
Cognitive robot researchers have developed Semantic Event Chains (SECs), a way to observe (use of 3D imaging) and encode information (save to memory) on how to perform a task. (Aksoy, E. et al., 2013) SECs are used to teach robots tasks by demonstration. The robots are able to categorize these procedures or action representations as entirely new tasks (accommodation) or alter existing action representations (assimilation). In testing, cognitive robots have been able to determine similarities and differences from a data set of cutting, chopping, stirring, pushing, hiding, putting, taking and uncovering. Of these eight actions, the cognitive robot chose to assimilate cutting and chopping while the rest remained their own SEC or action representation. (Aksoy, E. et al., 2014)
Each of these tasks involved a choice of tool for the job. In the end the robot decided it was alright to stir with the spoon and the knife while chopping and cutting were attempted with the knife, spatula and cleaver. Think about a daily task in your house that you would train your cognitive robot to complete. What schemata are involved? Alternately, choose a response to see where the cognitive robot may not understand and what ways would you correct this behaviour. Would you train a child in the same way?
Cognitive robot researchers have developed Semantic Event Chains (SECs), a way to observe (use of 3D imaging) and encode information (save to memory) on how to perform a task. (Aksoy, E. et al., 2013) SECs are used to teach robots tasks by demonstration. The robots are able to categorize these procedures or action representations as entirely new tasks (accommodation) or alter existing action representations (assimilation). In testing, cognitive robots have been able to determine similarities and differences from a data set of cutting, chopping, stirring, pushing, hiding, putting, taking and uncovering. Of these eight actions, the cognitive robot chose to assimilate cutting and chopping while the rest remained their own SEC or action representation. (Aksoy, E. et al., 2014)
Each of these tasks involved a choice of tool for the job. In the end the robot decided it was alright to stir with the spoon and the knife while chopping and cutting were attempted with the knife, spatula and cleaver. Think about a daily task in your house that you would train your cognitive robot to complete. What schemata are involved? Alternately, choose a response to see where the cognitive robot may not understand and what ways would you correct this behaviour. Would you train a child in the same way?