LIN 463 1st Edition Exam 2 Study Guide Social Robots o Features Species general able to recognize member of same species and aware of relationships between species Able to monitor and infer others actions goals and intentions Able to engage other contingently Uniquely human Representational theory of mind o TOM mechanism in brain takes a while to emerge compared to other components of brain o Information in the Theory of Mind Mechanism essentially exploits the machinery of metarepresentation and decoupled primary Triadic attention o Shared joint attention Read about this in chapter 12 of the textbook pages The Brain Recognition and representation of others o Fusiform face area Sensitive to faces visual o Extra striate body area Sensitive to bodies visual o Posterior STS Interpreting body motion re goals o TPJ Representing contents of others mental states such as beliefs The Brain Triadic attention Medial Prefrontal Cortex o Dorsal stream triadic attention shared collaborative goals o Ventral stream more affective sympathy empathy Dunbar s number o Approximate number of social connections a species can have given brain size o For humans it is about 150 Mirror Neurons o Neurons accidentally discovered in primates shown to fire equally when doing an action and when viewing someone else do the same action o Responds to intention actions both by self and others as well as intention o Mirror neuron system as a possible inference about cognition Neural networks o Single Layer Networks Compute only linearly separable computations Contain an input and output layer Perceptrons o Multiple Layer Networks Compute any computable function Think universal Turing machine Key Hidden units Trained by back propagation Modulates how much activity gets passed It learns to produce the correct answer Back Propagation o Generalized delta rule Mismatch between output and input is used to adjust the weights to better guess the right answer each time o Supervised learning correct answer must be known to compute mismatch Split vs unsplit models Reward system o Reward sensitive neurons o Orbitofrontal cortex strongly sensitive Representations of primary and secondary reward o Striatum n accumbens Hedonic and motor Anticipation reward N accumbens o Amygdala Stimulus features o Dorsal anterior cingulate Prediction errors o Dorsolateral PFC weakly sensitive Goal related Actual reward activated medial PFC Dopamine o In striatum and amygdala Temporal Difference TD
View Full Document