Adaptive Behavior-an action is taken; the world responds to the action; and the individual infers something about the world and then adapts his behavior so as to secure desirable responses.

    -all involve the response of a human or an animal
    -teaching, persuading, influencing= all ask communicator to adapt

Increment-the amount that has been learned as a result of the trial

ex. Pr(t+1)=Pr(t)+increment

Alfred example:Pr(t+1)=Pr(t)+0.2←increment

1.  
   1. Alternative behaviors available for the individual (left/right)
   2. State of individual described by probabilities that add up to 1
   3. Alternate responses of the World
   4. State of world described by probabilities of various responses to each behavior (right = 70% cheese)
   5. Set of possible events is combination of possible behaviors crossed by possible world responses (left/no reward, left/ reward)
   6. Specify adaptation equations for each possible event (event 1 = left/no reward)

1.  
   1. Going Left w/ Reward
      1. PL(t+1) = PL(t) + a[1-PL(t)]
   2. Going Left w/ No Reward
      1. PL(t+1) =  [PL(t) - bPL(t)]
   3. Going Right w/ Reward
      1. PR(t+1) = PR(t) + a[1-PR(t)]
   4. Going Right w/ No Reward
      1. PR(t+1) =   PR(t) - bPR(t)

1.  
   1. A = the rate associated a with reward
   2. B = learning rate associated a with non reward

•  
  • **Takes values between 0-1.
  •     -The ‘b’ is multiplied to PL(t) because the non reward reduces the likelihood of doing PL again
  •     -We whittle down PL by b
  •     - ‘a’ and “b” have no relationship

how bad you want or need something. EX in book - water is reward.

1.  
   1.  
      1. size of reward, big often better, need for reward
      2. text assumes that motivation doesn't change during trials

the ability of the subject to draw inferences and modify behavior on the basis of experience.

1.  
   1.  
      1. the closer a and b are to 1.0, the greater the ability
      2. text model can’t disentangle motivation and learning ability

a subjects prior experiences.

1.  
   1.  
      1. can have postitive or negative effects

one alternative is always rewarded and the other never is. The probability of a left response (if left is always rewarded and right never is) approaches closer and closer to one. The approach is steady but occurs at a decreasing rate.

1.  
   1.  
      1. This is what Alfred had

Does not recognize patterns, but assumes that events are defined in some terms. Such as alternate patterns, or choices can be chosen in a patterned response.

1.  
   1.  
      1. Reward on L, R, L, R, L, R

1.  
   1. if we live in a world that rewards NONcreativity, then we will inherently learn to NOT be creative over time
   2. However, creative people may still exist
      1. possibly due to....
         1. sloppy learning environments
         2. chance
         3. slow learning individuals
      2. But!
         1. it is hard to model intelligence
            1. some people may reward new ideas with lots of praise
            2. if we give large rewards to good ideas that happen to be creative, we get more creative people

•  
  •  
    •  
      •  
        •     -the graph shows how one “learns” to be traditional

1.  
   1. the expected proportion of those who will end up going left all of the time is equal to the proportion of those who went left on the first trial so Alfred will end up at PL=.50 in a benign world
   2. in a malevolent world an individual will move toward an equal probability of doing each alternative, oscillating around .5. The proportion of people going left at the end of a long period living in a malevolent world will be .5 regardless of the proportion that goes left at the first trial.
   3. Alfred ends up asymptotic at either R or L, depending on his first random choice.

1.  
   1. skinner placed food in a chamber and had the food be delivered in a set interval.
   2. whatever behavior the pigeons were doing at the time of the food delivery, they attributed to the foods arrival
   3. thus, those behaviors were repeated
   4. this is a patterned reward system because the pellets were dropped at random

1.  
   1. Hair Color
      1. Blonde and Brunette
         1. 4 events
            1. blonde, flattered
            2. blonde, not flattered
            3. brunette, flattered
            4. brunette, not flattered
         2. Reinforcement Assumption 1
            1. blondes have more fun
            2. you can choose your hair color every day
            3. only E1 and E4 can happen
            4. over time, if you choose to be blonde, you will choose blonde more and more based on the previous days experience
            5. this is not superstitious learning - it is a simple adaption model (people are adapting rationally to some social prejudices)
         3. Reinforcement Assumption 2
            1. pretty people and not so pretty people
            2. pretty people are flattered, not so pretty people aren’t
            3. hair color does NOT matter, but pretty people attribute the flattering to hair color - pretty people act as though the reinforcement occurred because of their hair color
            4. for them, E1 and E3 occur: Benign World
            5. for not so pretty people, only E2 and E4 occur: Malevolent World
            6. We find that pretty people will not change their hair color
            7. Not so pretty people will change Hair color Often
   2. Personality Development
      1. baby waking up in the morning (initial probability of each is ⅓)
         1. He cries
         2. He wets his diapers
         3. He gurgles
      2. Assume for each event, something happens
         1. If he cries - Parent goes to the baby, picks him up, cuddles
         2. If he wets his diaper - parent goes to the baby, changes diaper, picks him up, cuddles
         3. If he gurgles - parent goes to the baby, picks him up, cuddles him
         4. No matter what, Parent picks him up, cuddles
      3. Each baby will be superstitious in their learning
         1. ⅓ will always cry
         2. ⅓ will always wet his diapers
         3. ⅓ will always gurgle

•  
  •  
    •  
      •         iv. This model suggests that despite the fact that the environment is identical and             despite a learning process that is oriented toward seeking rewards, radically             different personalities will be generated

1.  
   1. Successful Executives
      1. Suppose executives have very different administrative styles and they all believe they effect the success of the business
         1. we assume that their behavior is dependent on the economy at the time and that their behaviors have little effect
         2. based on the state of the economy, executives may or may not believe that their current model is successful
         3. Example
            1. hiring from within, during a good economy, leads executives to believe this is the right approach
            2. this is superstitious learning

reward probabilities are not fixed; you and the world are both learning to adapt. Example: hunting & Mating

mating
people are more concerned with doing the “same” thing as other
they are less concerned with doing any “specific” thing
hunting
as a person is looking for the reward, the environment is learning to place the reward so the person cannot find it
Example:
Competitors trying to out do the competition
person is trying to kill a bird and the bird is trying not to be killed
Commuters try to schedule travel when other commuters are not

1.  
   1. Everett M. Rogers, Diffusion of Innovations, 4th ed.  Free Press, 1995

Why do some assumptions may not apply to diffusion in real world situations?

Ex: Simplications made for mathematical reasons

-Social (if u have friends)

- if you listen to the source of diffusion

-demographics: age, sex, status, social norms

-technology

-trustworthiness

Project Revere Example

Object:

Basic Method:

1.  
   1. object is to test a means of disseminating information.
   2. basic method - bomb the US communities with leaflets and follow up with phone inquiries.

What is the rate of Diffusion?

What is the Pattern of Diffusion?

(N at start) + (N/T)

(N/T) = change in N over change in T = Rate of Diffusion

1.  
   1.  
      1.  
         1. everyone who hears it believes it
         2. there is a constant rate of information diffusion

1.  
   1.  
      1.  
         1. You can end up with more people that know the information than people in the building
         2. Various things limit the diffusion process

1.  
   1.  
      1. Taking the growth limit, (number of people in the organization) into account
         1. N = Growth Limit; n = number of people that have the info at the start of the period; t = time period; a = diffusion rate
         2. n/t = (a)(n)(N-n)
         3. Really only n is changing thru the diffusion process