Three points for today

1
Three points for today

• We distinguish between two types of knowledge
  Procedural and Declarative.
• 2. Knowledge comes in several unit sizes, from
  small (concept) to big (schema), with units at
  one level connecting to form next level.
• 3. How these units are processed, especially how
  connections among units are formed, influences
  learning.

2
1. Types of Knowledge
3
Two Types of Knowledge

• Declarative Knowledge
• Knowledge that you know you have and that you
  can report (declare)
• Procedural Knowledge
• Knowledge of how to do things (e.g., tie your
  shoes, do arithmetic).

4
2. Knowledge Units (a) Sizes
5
Declarative Knowledge Units

• Concept
• Smallest unit of knowledge (not of meaning)
• A concept is a mental representation of a
  category of things in the world (e.g. DOG)
• Concept allows you to decide whether a stimulus
  is a member of the category
• Issue nature of the representation prototype?
  Set of exemplars? Feature list?

6
Declarative Knowledge Units

• Proposition
• Smallest unit of meaning that has a truth value
• A proposition asserts some quality or behaviour
  of some entity
• Basically, Subject, Verb, Object or Quality
• e.g., The dog barked The dog is brown The dog
  wore sneakers

7
Declarative Knowledge Units

• Schema
• Stored knowledge structure that influences
  perception and comprehension
• Capture important information about people,
  situations and events
• What usually happens? What is usually present?
  When does an event usually occur?
• Acquired slowly difficult to modify

8
Procedural Knowledge Units

• Production
• If CONDITION holds, then perform ACTION
• Anderson argues that all behaviour can be
  modeled as sequences of productions
• A sequence of productions can become automatic.
  This is proceduralization.

9
Procedural Knowledge Units

• Script
• due primarily to work of Roger Schank in
  artificial intelligence.
• a script is like a schema for a process
• detailed, because computer programs wont run
  unless you specify everything necessary
• more recent versions allow scripts to be created
  as needed from stored components (episodes,
  actors, settings)

10
2. Knowledge Units (b) Connections
11
Network Models of Knowledge

• Two basic types of models
• Local Representation
• nodes in the network represent concepts
• Distributed Representation
• nodes dont represent anything concepts are
  represented in patterns of activation

12
Local Representation Models

• Quillian Collins (1969) TLC
• Teachable Language Comprehender
• Hierarchical organization
• Nodes are empty. They are placeholders in the
  network.
• All links the same length.
• ISA links and property links

13
Local Representation Models

• Collins Loftus (1975)
• less hierarchical version of TLC
• structure of network reflects persons
  experience rather than objective scientific
  information
• explained typicality effects
• introduced very important concept of spreading
  activation for retrieval of information
• explained priming effects

14
Distibuted Representation Models

• Parallel Distributed Processing (PDP)
• neural network or connectionist models
• models have units (neurons) and weighted
  connections (axons/synapses)
• concepts are represented as patterns of
  activation across many units
• each unit participates in many patterns no unit
  represents any one concept
• knowledge is stored in weights on connections

15
Parallel Distributed Processing Models

• Models start out not knowing anything weights
  on connections are random.
• Weights are adjusted during learning so input
  pattern becomes more likely to cause activation
  of appropriate output pattern
• One set of weights works for all concepts
• PDP models are very good at handling problems in
  which multiple constraints have to be satisfied
  at the same time

16
What are the last two letters in this stimulus?
17
3. Type of Processing
18
Types of Processing

• In early 70s, cognitive psychologists began to
  be
• less interested in structural questions (e.g.,
  what is the capacity of STM?)
• more interested in process questions (e.g., what
  is the best way to encode information for later
  retrieval?).

19
Types of Processing

• Rehearsal
• Maintenance rehearsal
• Simple repetition of stimulus
• Elaborative rehearsal
• Drawing connections between stimulus and what
  you already know

20
Types of Processing

• Levels of Processing (Craik Lockhart)
• processing types vary on a depth dimension
• semantic processing is deep form processing
  (e.g., colour, shape) is shallow
• deeper processing facilitates retrieval
• Bransford what matters is match between codes
  generated at encoding and type of retrieval cues

21
Problems with levels of processing theory

• Baddeley has argued that LoP theory
• is circular
• is an empirical failure under some conditions
  (doesnt work with recognition)
• Still,it is a useful heuristic How you encode
  matters

22
How you encode matters

• Deeper processing requires elaboration
• Elaboration builds connections between new
  information and old
• Elaboration makes new information more
  distinctive
• But how you retrieve also matters

23
How you retrieve matters, too

• Tulving and encoding specificity
• Recognition vs. Recall
• Bartlett and reconstruction
• Relearning

24
Tulvings encoding specificity idea

• Remembering is best when conditions at retrieval
  match conditions at learning
• paired-associates
• type of code generated
• mental and physical state
• context (e.g., Smith, 1986 Godden Baddeley,
  1975)

25
Recognition vs. Recall

• Recall
• retrieve learned materials with no further cues
• Recognition
• identify learned materials when presented or
  distinguish learned from unlearned

26
Reconstruction

• originally suggested by Bartlett (1932)
• remembering involves computing what must have
  happened, on basis of
• Some encoded material
• Some knowledge of the world at concept,
  proposition, and script levels.

27
Relearning

• First observed by Ebbinghaus (1885)
• When you re-learn some material, you acquire it
  faster than when you learned it the first time
• Reduction in effort or time required savings
• Holds even over very long intervals (years)
• Very sensitive measure of memory

28
Implications for instruction
Encourage deeper processing Encourage
elaboration Encourage use of mnemonics and other
strategies Reconstruction is going to happen
dont try to resist, try to guide it
29
Implications for instruction

• Make information retrieval more effective by
• Matching encoding and retrieval conditions
• This includes context and students state
• Providing relevant cues at retrieval
• Using prior knowledge to reconstruct missing
  information

30
Title