5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 1 of 7
Chapter 2: In Unfamiliar Territory Without a Guide
¤ What supervised learning situations reveal about the profound difficulties of
unsupervised learning (relevance, importance, ordering, and reliability of information).
¤ Particular characteristics of classroom learning and child-directed speech.
¤ Safe and unsafe learning environments.
¤ Making difficult perceptual learning more tractable.
¤ Heuristics from the golden age of learning theory for simplifying the problem of
learning without supervision (and taking advantage of supervision when available).
Chapter 3: The importance of objects
How essential it is to be able to treat things as the same, and hence the importance
of perceptual learning
Why this seems easy to us despite how difficult it is
To illustrate the challenges and the impressive and powerful way in which they
are handled in the human brain, we tell the story of how Hebb developed his
theory of cell assemblies
Cooperation through competition
How cell assemblies form
How cell assemblies function
A puzzle - why human infants are guaranteed to be incompetent
The slow, conservative nature of perceptual learning
Diffuse activity becomes focused and organized as cell assemblies form
Flexibility-competence tradeoff
5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 2 of 7
Putting the pieces together: Building models
Cognitive maps
Associations
Sequence
Hierarchy
Horizontal and vertical structure
Affective coding
Slow learning, active learning and other challenges
The slow, conservative nature of the learning process, while essential for perceptual
generalization, creates a potentially serious handicap. There are times when a faster
learning rate is not merely desirable, but essential. If a given cave happened to house a
bear that one narrowly escaped from when one innocently wandered by, avoiding that
cave in the future has to be learned immediately, and not with the usual requirement of
multiple repetitions of the experience. Clearly some way to override this conservative
mechanism when appropriate is called for. The continuation of the closed loop activity
of the cell assembly provides a substitute for repetitions in the environment.
Consolidation is what we call this means of achieving the repetition required for learning
in the mind rather than in the world that can be used as an override mechanism. One
need not have certain experiences repeatedly if, when an instance does occur,
consolidation provides the necessary multiple repetitions.
Not sure if we want to include mention of Keppel, but if we do perhaps the following
would be helpful:
Do we get to learn, or is all up to our internal mechanisms?
The consolidation process is a central actor in the way internal supervision takes
place. However some years ago a distinguished student of the learning process
expressed not only his personal opposition to such a mechanism, but what he felt
was the opposition of most experts in this area. The reason he cited: the distaste
for thinking of learning as a process in which an individual plays no active role.
This distaste motivates an attack on a series of studies that seems to provide
evidence for such a built-in mechanism. Yet the authors of this series of studies
themselves have emphasized the role of active learning. What‟s going on here?
THINGS TO RAISE:
1. The criticized studies are sound
2. There must be a built-in mechanism
3. There is no conflict between active learning and a built in mechanism
4. A key source of the misunderstanding is a failure to distinguish levels of
mechanism
5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 3 of 7
Positive Feedback: Utilizing the Benefits and Managing the Risks
The blow-up problem
The dynamic trace
PIWIT
Regional inhibition
How the brain flexibly and automatically shifts from parallel to serial and back as
the circumstances require
Limited capacity and competition serve as a means of maintaining focus despite
the awesomely parallel capability of the perceptual systems.
Limited capacity and the control of clutter
Gibson's challenge - the more one knows about something, the longer it would take to
think about it?
Dealing with new information about old things
How to handle new information that is in some way different from what one had
learned previously. Both updating and refining skills depends upon modifying the
information one has stored rather than merely adding to it.
Catastrophic forgetting
Structural clutter
keeping stored information up to date without being overwhelmed with
information
PIWIT
CLR
Dynamic clutter
turning off persistent activity in a cell assembly
fatigue
Zeigarnik effect
5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 4 of 7
The quest for clarity
What are the key motivational and emotional inclinations that an information-oriented
organism needs?
Inclined to seek opportunities to explore new things and make sense of them without
endangering yourself by getting in over your head
Exploratory inclinations
Curiosity
Boredom
Safety enhancing inclinations
Detest being confused
Clarity seeking inclinations
Figuring something out is pleasurable
How the clarity mechanism provides these inclinations and guides behavior
How the clarity mechanism works to detect the state of the system
Mention automaticity in this summary? Mention that clarity is often guiding us without
our awareness?
Outcome sensitive learning
Through a mixture of arousal, pleasure and pain, clarity and environmental impact, it is
possible to describe a set of heuristics that could make identifying “importance” feasible.
What we are proposing is that the only practical way to approach this problem is to
generate heuristics, rules of thumb. In a world where the future cannot be known and
decisions as to whether to store information or not must be made nonetheless, these best
guesses are remarkably suited to this difficult task.
5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 5 of 7
Control mechanisms
Roughly speaking, control mechanisms achieve one or more of the following goals:
1. Prevent the system from going awry
2. Managing and normalizing thoughts and inputs to keep them within the bounds
that the system can handle
3. Allocation of cognitive resources – essential due to the small brain/big world
challenges
4. Setting brain „operating parameters‟ appropriate to current goals, knowledge,
and environmental conditions
The need for detecting system-level states that require attention and perhaps intervention
How clarity and metacognition detect these states
Something about metacognition
Attention
Clarity failure -- a decline in clarity that is large and/or of rapid onset -- and its
management
Pursuing Our Purposes
¤ The ceaseless scheduling problem of working out how much time and energy to
allott to a myriad of possible purposes, behaviors, and thought processes.
¤ Will-power, temptation, and delay: why can‟t humans simply determine whether
or not delaying is advantageous in a particular situation and, if so, do it without fussing?
Why should this be difficult? What are the properties of the brain that make delaying an
unpleasant (and not infrequently unsuccessful) struggle?
¤ The evolution of will-power in social control.
¤ Will-power as a limited mechanism
¤ Will-power as an enabling technology co-opted for new purposes.
¤ The inhibitory basis of will-power
¤ The historical role of inhibition in psychological theory
¤ Voluntary and involuntary attention
¤ Mental fatigue and the debilitating effects of attentional deficits
¤ The commonalities between mental fatigue, ADHD, and prefrontal cortex injuries.
5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 6 of 7
Pathways in the Mind: Solving Problems
¤ "In the Head" vs "In the World" problem solving.
¤ Variation and selection in problem solving and natural selection.
¤ The power of models -- problem-solving as off-line trial-and-error.
¤ Potential dangers and drawbacks of mental models.
¤ An incremental approach to off-line problem-solving
¤ Cognitive maps and problem spaces
¤ Memory retrieval as problem-solving
¤ The apparent weakness of human problem-space searching capabilities
¤ Perception, framing, analogies, and generic structures
¤ The effects of framing on some real world social problems
Advantages and disadvantages of knowing a whole lot: On Expertise
Why the following isn't true: The more we learn the more there is to think about
– and the longer everything takes?
How being able to operate economically and efficiently plays a big role in
expertise
One tradeoff of this is the inarticulable nature of expertise
Content-specific nature of expertise
Differences between experts and novices
Ads and disads
Tradeoffs
What is lost What is gained
Flexibility Competence
Information Speed, efficiency, compactness, economy
Articulability Speed, efficiency, compactness, economy
5e214f7f-5145-4b6a-9403-0468c1d163e7.doc Page 7 of 7
Keeping on track: Temptations and other hazards
¤ The modern dilemma part 1 -- society imposes many tasks that are not inherently
compelling and for which the human mind is not ideally suited.
¤ The modern dilemma part 2 -- societies offer a vast, readily accessible supply of
potential hazards and temptations that are inherently fascinating (and, in fact, often have
been intentionally designed to be so).
¤ Direct and indirect effects of temptations
¤ A taxonomy of artificial supernormal stimuli: sneaky, simple, and synthetic
¤ The threat of a limitless supply of fascinating, worthy, valuable, intelligent stuff
¤ The dangers of bigger-than-life virtual experiences.
¤ Serial-killers, terrorists, fist-fights, gang-wars, car-chases, heiresses, tycoons,
generals, champion athletes, mansions, mega-yachts, prostitutes, mafia dons, government
conspiracies, flawless skin, perfectly-toned bodies, hurricanes, tornadoes, floods, and
fires
¤ Biases: adaptation levels, misplaced priorities, and unrealistic beliefs
¤ Implicit associations