Diff for "DissertationNotes"

proposal outline

1. Intro/problem statement
   1. Humans speaking a shared language can
      1. understand things they've never heard before (e.g. new voices don't cause catastrophic failure of the language system)
      2. infer missing data (e.g. phoneme restoration, gardenpath sentences)
      3. be influenced by ideological beliefs, even in perception
      4. command metaphors
   2. How does human perception mitigate the effects of sparse data caused by biological and linguistic variation?
   3. Why do paralinguistic social factors influence low-level perception?
   4. Do linguistic perception and social perception share a common processing faculty?
   5. Can an integrated socio-linguistic processor explain metaphoric processes?
2. Explain exemplar models
   1. How does it work
      1. Events stored as memory traces, ie exemplars
      2. Percepts compared to existing exemplars
      3. Percept categorized based on nearest exemplars
   2. What phenomena does exemplar theory explain well
      1. Categorization and speech perception
      2. Recognition and recall
      3. Frequency effects
   3. What does it struggle to explain
      1. Non-frequency effects, esp social effects
      2. Category genesis and generalization
3. Explain REMERGE
   1. What is it
      1. connectionist model
      2. theory of generalization
   2. How does it improve on exemplar models
      1. recurrent similarity computation
      2. generalizes structural properties
   3. What does it lack
      1. Storage of generalizations
4. Explain persona construal
   1. What phenomena necessitates a persona construal model?
      1. Indexical ordering
      2. bricolage
   2. How does freeman and ambady's model work
      1. perceptual cues activate low level input nodes
      2. top-down pressures (task demands, stereotypes, high-level cognitive processes) activate high level input nodes
      3. activations are passed between nodes until a steady state is reached

      4. probability of a response x at time t is proportional to activation of node x at time t

   3. How could it improve previous models?
      1. Generalizations learned by REMERGE can be added as stereotypes or categories in this model
      2. Input and category nodes can be shared, allowing for top-down effects in REMERGE
      3. Maybe allows generalization based on stereotypes, not only perceptual data (i.e., bricolage)
5. Does adding social info as top-down bias predict memory and processing as found by Zion, Sharese, SK, and Meghan?
6. Does connecting a persona construal model in place of parameters (as previously) change the pattern of results. If so, how well does it predict indexical ordering?