Spreading Activation

• If you activate a node, and let the activation spread to other nodes you have a useful similarity metric.
• A node is similar or related to nodes with which it's connected.
• You can then spread another step farther.
• If you want to see how two nodes are related, start spreading from both and see where they meet.
• There is some evidence that this works in human Psychology with associative memory.
• This was used rather heavily in a flavour of connectionist models known as the interactive activation model.
• It uses a more refined mechanism of spreading. Activation doesn't spread equally to all nodes.
• In the CABot research, we use this in our Maes nets for planning.
• Spreading activation is really central to the way people think.
• Try the 10 question associative memory task (instead of 20 yes/no).