Experts are better problem solvers than novices because they have more domain-specific knowledge stored in their long-term memory. Studies show that chess experts can rapidly recognize patterns on a chess board because they have memorized around 50,000 chess patterns. Their extensive knowledge of a domain is organized in a way that allows them to quickly recognize situations and determine appropriate moves or solutions. Novices perform better on analogy problems when they are given a similar surface problem first that provides relevant context, showing the importance of superficial features for non-experts. Both surface and structural similarities between problems influence how well novices can solve by analogy.
1. 1 Problem Solving by Analogy
Analogies provide a way of applying (mapping) known
information (source) to novel problem solving domains (target):
2 Solve This
An evil dictator ruled a small country from a fortress, which was
located in center of country where many roads converge. The roads
are all mined in such a way that small bands of the dictators thugs
can pass freely, but any large attacking force would get blown up.
A general who wants to free the country has a large force that he
knows could take the fortress if they attacked all at once. Can he do
it?
3 Solve This
You are a doctor with cancer patient & need to kill tumor in
patient’s stomach. It is impossible to operate, but there is a ray
beam that can be shot at tumor to kill it. Enough rays need to pass
through tumor at same time to kill it, but a beam of this intensity
will also kill healthy tissue. Can you save the patient?
4 Analogy Performance
Percent solving cancer problem in different conditions:
w/ naive subjects: 10%
After fortress problem, if not told relevant (and in context of
other problems): 20%; if told relevant: 100%.
After similar surgery problem: 88%
After similar ICBM problem: 58%
Conclusion: surface features are critical: people are superficial!.
2. 5 Structural Similarity
New version of cancer-like story: you don’t have a ray beam
machine that is powerful enough (no mention of damaging tissue).
The abstract structure is now different (limitation is not about too
much=bad, its about not having enough), and people do worse.
So both surface and structural similarity matter:
Holyoak & Koh (1987):
Surface Similarity
High Low Mean
Structural Similarity High 88 56 72
Low 40 13 27
Mean 64 35
6 Expertise
In what way is an expert a better problem solver than a novice?
Better knowledge of specific facts?
Better general strategies/heuristics?
Better..?
7 Chess Experts
Gobet & Simon & Others showed that chess experts
have 50,000 chess patterns stored in memory.
can rapidly encode a chess board (chunking).
don’t search for moves very much (Kasporov without much
time is not that much worse that with time).
Conclusion: expertise is mostly about having more
domain-specific knowledge.
It’s also about how this knowledge is organized.
8 Other Experts
Steve Falloon: Could memorize around 100 digits, by elaborate
chunking strategies and associations. But this didn’t generalize to
other kinds of memory.
Physics experts: reason differently than novices? Some evidence
says yes, other says no.
Implications?? Many people say “study physics or math in college
cuz then you’ll be able to do anything..” maybe not!.
3. 9 So You Want to be an Expert?
One word: practice.
The simple truth of life and everything (in 7 easy steps):
1. Getting good at something means learning new
domain-specific knowledge.
2. (there are no “10 easy steps to becoming an expert..”)
3. Acquiring this knowledge just takes time-on-task.
4. You gotta be motivated to spend the time.
5. You gotta be successful to be motivated.
6. You gotta be good at something to be motivated.
7. Goto 1.
10 What does the Brain Say?
Frontal Parietal
Occipital
motorsoma
vision
planning
control
action
space
auditory
language
episodes
objectsTemporal
Problem solving tasks like tower of hanoi activate prefrontal cortex.
Damage to PFC impairs flexibility, control, abstraction, etc.
Neural network models are very knowledge-dependent: you need
to train the hell out of them to make them smart.