!!!Our Brains Always Make Best Guesses!!!
Perception: our personal assumptions about the world (seeing is believing).
• we have expectations of regularities in the world which generate perceptions.
▯ ie. light comes from above or faces poke out
• we have an intuitive feeling that what we are seeing is correct, even though it is
• our brain creates creates experiences which determine our perception.
• perception can be inﬂuenced by a variety of factors including prior knowledge,
memories, and source confusion.
Attention: a person’s immediate conscious perception.
• ensures continual cognitive processing.
• can be selective (divisible and shiftable) or limited (focusing on only one stimuli).
• multitasking is difﬁcult, especially when the tasks require the same part of the
• Some skills we need to consciously think about (controlled processing), whereas
others are easier if we don’t think about them (automatic processing).
Memory: our recollection of past experiences and perceptions.
• memory is a reconstructive process because we often can’t remember the exact
story, so we ﬁll in the missing pieces with inaccurate information.
• memory can be inﬂuenced by language, for particular words can make you
remember an event in a different light.
▯ ie. when describing a car crash, saying smashed rather than hit might make you
▯ think the car crash was more severe
• the more we hear something, the more we will believe it to be true
▯ ie. presidents say things over and over again until you believe what they are
▯ saying is true
Judgement: a process through which we consider evidence and draw a conclusion.
• judgement is based on feelings of familiarity which comes from memories and past
perceptions (prior knowledge).
• inﬂuenced by internal and external environment.
• judgement can be inﬂuenced by words and patterns.
▯ ie. if a fair coin is ﬂipped and you get 5 heads in a row, you believe the next one
▯ will be a head also
People have no idea why they do what they do.
1. example: electrically stimulating the brain to cause someone to do an action. Even
though the action is initiated by a machine, people believe that they did that action by
their own conscious will. 2
2. example: magnetic stimulation of the brain to cause someone to raise a certain ﬁnger.
People show no awareness that the decision was not their own.
• decision-making is a prediction that is based on how plausible or reasonable
• making a decision also largely depends on the reward or punishment that the
choices of a decision present. People favor decisions that are beneﬁcial.
3. example: mirror box - people put arms in a slot and look at a mirror to draw a straight
line. They see a hand, but the hand is not actually theirs. The hand that is not theirs
tried to draw the line and the person tries to follow, feeling as though they have made
the decision to draw the line.
• people are often not consciously aware that the will to do something is not theirs.
• how do we know that any decision that we have made in our lives was actually our
• inﬂuenced by language, emotions, other people, the environment, and in some
Everything about us is susceptible to inﬂuence!
The three problems:
1. The mind-body problem
• how do we explain how the mind and body interact?
2. The homunculus problem
• how do we explain the mind without mentioning the homunculus?
3. The problem of physical characteristics
• how can we scientiﬁcally explain the mind if it is unobservable?
Schematic Cognitive Systems
Information processing theory: is a completely physical theory of the mind that
suggests the brain is the hardware of a computer and the mind is the software.
What kind of computer are we?
How do we think about the decisions we are making?
1. block of wood and golf tee game
• the point of the game is to jump pegs over other pegs in order to remove all pegs.
• we can think of the game is a structured way.
• there are speciﬁc rules, speciﬁc positions, it’s self-contained, it has deﬁnite states,
and it’s checkable.
Problem solving consists of starting in one position with a goal to get to another. There
are a variety of intermediate steps before we can get to the ultimate goal and we learn
through trial-and-error. 3
The four properties of symbolic systems:
1. semantically interpretable representations
2. transformations of these representations (language-like rules)
3. serial processing
4. “sense-think-act” cycle
Problem solving is like a google-search through this system. What are the optimal
moves to get you there? How do I move/ what steps do I take to get to the goal? - solve
the problem as efﬁciently as possible (hobbits and orcs, eleven steps is the most
efﬁcient way to get to the goal).
Psychologist robot - works by manipulating simple grammatical transformations.
She has computer representations of words that she transforms depending on her basic
• tacks on previously said words (if they said “afraid”, she says “why are you afraid”)
• repeats earlier sentences
• watches for keywords
• stalls when confused or when the other rules are not applicable
SHRDLU (Terry Winograd, late 1960s):
Manipulate blocks - the system is aware of where things are and how they are related to
Maintain a representation of the “micro world”/ virtual world of the blocks. Has rules:
• An object is always supported by something
• A block can support another block
• A block can support a pyramid
• You can’t pick up a block if it’s supporting something else
Interprets commands by instantiating goals related to the command words and using
“backtracking” procedure. Sets up a list of goals.
This is a symbolic system because:
• can represent language with words that represent things in the micro-world
(semantically interpretable representations)
• such as the word block which represents the information of the micro-world
• it transforms the representations by moving them and interpreting them
(transformations of representations)
• the strategic way by which it decides to follow a set of rule and instructions (serial
• it takes in information, interprets information, and responds (sense-think-act) 4
MYCIN (Shortliffe, 1970s):
Blood disorder expert. Enter a bunch of information about a patient who has a blood
disease and mycin will request information trying to ﬁnd out what blood disorder the
Mycin is built from a giant data-base of everything we know about blood diseases.
• if-then rules
• backward chaining: starts with a list of possible hypothesis and asks questions
• certainty factors: the infection is proven when the system is highly certain
The problem with Mycin: He doesn’t learn. People learn from their mistakes. Mycin
follows a list of reasoning rules and we can program new facts into him, but he doesn’t
adjust if he makes a mistake.
SOAR VIDEO QUESTIONS
1. What is the problem being solved?
• Railway problem
2. What is being represented?
• Train cars and the physics of train tracks
• Has rules for moving things around
• Multiple possibilities, chooses one, fails or succeeds
3. What are the rules that the system uses?
• switch versus selection
4. How does Soar learn?
• resolving impasses stores a pattern of conditions called chunks, stores in long term
• trial and error
Tri-Level Explanation (David Marr)
Marr suggests that there are three levels to explaining a symbolic system.
• task Description - the big picture (input)
• algorithmic Level Description - the step-by-step to achieve the task (formula)
• physical Implementation - the physical steps a system must take to do the
Understanding the physical helps us to understand the computational. Even though it is
not necessary, it is practical.
How do we do the math in a biologically plausible way?
• the algorithms we use need to run on the hardware that we have got.
• picture of a neuron: the inputs are the dendrites, the processor is the soma/ cell body,
the outputs are the axons. 5
• connectionism is way of doing “the math” using neurologically inspired units.
• a set of nodes that have connections, the connections have certain strengths, and the
layer at the top produces an output.
• difference and similarities to symbolic systems. Think: In what way does it change the
game/ in what way is it similar?
Word Superiority Effect: people are slower to identify a letter when it is attached no a
non-word, compared to when identify a letter attached to a complete word.
• If we are processing words that are combinations of letters, we need to be able to
process the letters prior to the words, so why does it matter to us if the letter is a
part of a non-word or a complete word?
The Interactive Activation Model: