UPI Study › Courses › Introduction to Artificial Intelligence › Free lessons › Lesson 2.1 Free with a free account
Learnable Intelligence: Insights from Perkins' Framework
Introduction to Artificial Intelligence · Chapter 2: Intelligence, Cognition, and Agent DesignSection 2.1 · lesson 9 of 6913 min · free with a free account
Free preview
Create a free account to keep watching
It takes a minute and costs nothing. Your free account unlocks all of Chapter 1 and 3 more lessons of Introduction to Artificial Intelligence.
Lesson complete
Earn the 3 college credits for Introduction to Artificial Intelligence
Enrolling adds a quiz after each lesson, the graded work and the official transcript. Recommended for credit by NCCRS.
Next lesson: Evaluating How We Think: Cognitive Function Analysis →← Machine Learning vs. Artificial Intelligence: Key DifferencesNext lesson: Evaluating How We Think: Cognitive Function Analysis →
In this lesson
Perkins' three kinds of intelligence and what machines can learn.
- Describe Perkins' neural, experiential, and reflective kinds of intelligence
- Explain why effective intelligence can grow even when neural intelligence barely changes
- Identify strategies for teaching reflective thinking, including hugging and bridging for transfer
- Compare fixed programming with machine learning using Perkins' framework as an analogy
Free study notes
Get this lesson's notes by email
The key points and 4 practice questions with answers, in your inbox. One email, free.
One email. Unsubscribe in one click.
Transcript
Everything said in the video, section by section (1,806 words). Read along, or search it.
Opening Question
Same Knowledge, Different Scores
Two students take the same algebra quiz, and both know the material. One rushes, and drops easy points. The other pauses, and checks each answer. Same knowledge, very different scores. So was the second student simply born smarter? And if intelligence isn't fixed at birth, which parts of it can a person, or a machine, actually learn to do better?
Background
Perkins and Outsmarting IQ
Many people grow up believing that intelligence is something you either have, or you don't. David Perkins, a researcher at Harvard's Project Zero, challenged that belief in his 1995 book, Outsmarting IQ. His central idea is hopeful. Being smart isn't one single ability, and some of its most important parts can be learned. That matters for AI too. Engineers face the same choice every day: give a system all its abilities in advance, or let it learn them from experience?
Three Kinds
Engine, Road Map, and Driver
Perkins said everyday intelligence blends three different things. Picture a driver on a road trip. Neural intelligence is the engine: the basic efficiency of the brain's wiring, how quickly and accurately it handles information. He saw it as set mainly by genes and early development. Experiential intelligence is the road map: the knowledge and know-how you collect from experience in one particular area. And reflective intelligence is the driver's judgment: the habit of planning, checking, and questioning your own thinking. All three decide whether the trip goes well.
Experiential Intelligence
Domain-Specific Know-How
Experiential intelligence is easy to spot. A chess master recognizes thousands of board patterns. An experienced nurse senses that a patient is getting worse, before the numbers show it. A good mechanic can name an engine fault just from its sound. But this kind is domain-specific. Expertise in one area doesn't automatically carry over to another. A chess master's patterns won't diagnose an engine.
Reflective Intelligence
Mindware and Metacognition
Reflective intelligence matters most for this lesson. It has three ingredients. Strategies, such as list the alternatives first. Attitudes, such as curiosity and carefulness. And metacognition, which means thinking about your own thinking. Perkins calls these learnable thinking tools mindware: software for the mind, as opposed to the brain's hardware. He also argued that IQ tests mostly capture the first two kinds, mental speed and stored knowledge, and say much less about the third.
How Much Each Changes
Two of Three Keep Climbing
How much can each kind change? Neural intelligence changes little. It's mostly set by biology. Experiential intelligence grows steadily, with practice and exposure. And reflective intelligence can be taught, practiced, and strengthened. This is a sketch of the idea, not measured data. But the shape is the point: two of the three lines keep climbing.
Core Argument
Four Steps to Learnable Intelligence
Perkins' argument that intelligence can be learned runs in four short steps. One. Experiential intelligence grows with experience. Two. Reflective intelligence is made of habits and strategies, and those can be taught. Three. Real performance depends on all three kinds together. A powerful engine with a careless driver still ends up in a ditch. Four. So effective intelligence, how well you actually think and solve problems, can increase, even if neural intelligence barely changes.
Evidence
Chess Memory, Rising Scores, and Training
Three lines of evidence support this view. First, expertise research. Chess masters remember real game positions far better than beginners do. But scatter the pieces at random, and their advantage almost vanishes. Their edge comes from stored patterns, not from a faster brain. Second, rising test scores. In many countries, average IQ scores rose about three points per decade, through the twentieth century. That's too fast for genes to explain, which points to environment and learning. Third, programs that teach reasoning strategies have reported gains in problem solving, though results vary, and the benefits don't always carry over to new areas.
Misconceptions
Three Beliefs Versus Perkins' View
This argument is easy to misread, so set three common beliefs against what Perkins actually says. Belief one: you're born with a fixed amount of intelligence. In fact, only one of the three kinds is mostly fixed. The other two are built over time. Belief two: if I can't do something, I lack the brainpower. More likely, you lack knowledge, or a thinking habit, and both can be developed. Belief three: learnable means anyone can become anything. No. It means growth is possible. Starting points and limits still differ from person to person.
Worked Example
The Bat and Ball Puzzle
Here's why a thinking habit matters. A bat and a ball cost a dollar ten together. The bat costs a dollar more than the ball. How much is the ball? The quick answer is ten cents, and it feels right. But test it. The bat would then cost a dollar ten, and the pair would cost a dollar twenty. Now slow down. Call the ball x. The bat is x plus a dollar, so two x plus a dollar equals a dollar ten. Two x is ten cents, so the ball costs five cents. Neural intelligence didn't change between those two answers. A reflective habit did: pause, and test your first answer.
Teaching Thinking
Five Strategies for Reflective Skill
So how is reflective intelligence taught? Good thinking is rarely picked up by accident. Researchers describe five strategies that work together. Make thinking visible: a teacher solves a problem aloud, wrong turns included. Name the thinking moves, with short names like estimate first, or check the opposite case. Practice with feedback, on the process, not only the answer. Build habits, like spending ten seconds planning before a task. And transfer deliberately: carry a strategy into a new area on purpose, like using a checking habit from math on a science report.
Transfer
Hugging and Bridging
That last strategy needs attention, because transfer doesn't happen on its own. Skills learned in one area tend to stay stuck there. Perkins and Gavriel Salomon described two ways to encourage it. Hugging means practicing in a way that closely resembles the real situation, where the skill will be used. Bridging means consciously connecting an idea to other settings, by asking: where else could this strategy help me?
Habit Cycle
Notice, Name, Practice, Reflect, Transfer
Put together, these moves form a cycle that turns a strategy into a lasting habit. Notice a moment that calls for careful thinking. Name the strategy, with a short, memorable name. Practice it on varied problems. Then reflect. Did it help? What would you change? Finally, transfer it into a different setting, and the cycle begins again.
Growth Mindset
Mistakes as Information
Beliefs matter too. Educators use the term growth mindset for the belief that abilities can develop through effort, good strategies, and feedback. Then a hard problem becomes a chance to build skill, not a verdict on your ability. And mistakes are information. They show which strategy, or which piece of knowledge, needs work. Effort should go into better strategies, not only into working faster or longer. And slowing down on purpose is a skill, especially when a problem is new, or the stakes are high.
Quick Check
An Accountant in a Bike Shop
Quick check. Lena is a careful thinker. She plans, double-checks, and questions her own reasoning. After years as an accountant, she starts work at a bike repair shop. On day one, a customer's wheel squeaks, and Lena has no idea why. Which kind of intelligence is she missing? Pause, and decide. Experiential intelligence. Lena's thinking habits came with her, but know-how is domain-specific. Years of accounting taught her nothing about wheels and bearings. Reflective is the tempting choice, but she plans and checks, so that part is strong. And nothing suggests her brain got slower, so it isn't neural. Her good habits will help her learn quickly. Time in the shop will build the rest.
Applying to AI
Fixed Programming Versus Machine Learning
Perkins wrote about people, but his framework is a useful lens for AI. The key question: are a system's abilities written in once, or built up from experience? Fixed programming means a programmer writes explicit rules. For example: if an email contains the phrase free prize, mark it as spam. The system's knowledge is frozen the moment the code is written. It gets better only when a person edits the rules. Machine learning works the other way. The system is given many examples, like thousands of emails labeled spam or not spam, and it finds the patterns itself. As more good examples arrive, it can improve, without anyone writing each new rule.
Comparison
Handling New Situations Over Time
Now compare them on new situations. A rule-based system fails, or guesses, when no rule applies. A learning system can handle cases similar to ones it has seen. Each has a main weakness. Fixed rules can't adapt, and they're costly to write and update. A learning system is only as good as its data, and may inherit its gaps and bias. Over time, a fixed system stays at one level, while a learning system can keep improving with more good examples, though the gains eventually slow.
Analogy
Three Kinds Mapped to AI Systems
Perkins' three kinds also map, roughly, onto the parts of an AI system. Treat this as an analogy, not an exact match. Neural intelligence is like the hardware and design: processing power, memory, and the structure of the learning system. Experiential intelligence is like the training data, and what the system learned from it. But learned patterns aren't understanding, and gaps in the data become gaps in ability. Reflective intelligence is like self-checking: testing outputs, estimating uncertainty, choosing strategies, and planning. That's often the weakest part. Many systems can't tell when they're wrong, and people supply most of the checking.
Quick Check
A Chess Program on Faster Hardware
One more. A chess program's code and training stay exactly the same. It's moved onto a much faster computer, and now it looks further ahead in the same time. In Perkins' map, which kind of intelligence did the move improve? Pause, and decide. Neural. Faster hardware is the machine version of the engine: more raw processing power. Experiential is tempting, since chess skill is about patterns. But the training didn't change, so it knows nothing new. And nothing was added to check its own moves, so it isn't reflective. A stronger engine helps, but only within what the system already knows.
Bringing It Together
Knowing When to Doubt an Answer
Back to the two students from the start. They had the same knowledge. The difference was a reflective habit, pausing to check, and that habit can be learned. Machines face a similar gap. Fixed programming gives a system one person's knowledge, at one moment. Machine learning lets it build experiential knowledge from data. But knowing when to doubt an answer is still a major challenge: for machines, and for people who never learn to pause and check.
Key Takeaways
Neural, Experiential, and Reflective
To sum up. Perkins describes three kinds of intelligence: neural, experiential, and reflective. Neural intelligence barely changes, but knowledge and thinking habits grow, so effective intelligence can grow. Thinking is taught by making it visible, naming it, practicing with feedback, building habits, and transferring on purpose. And machines can build experience from data, but checking their own answers is still the hard part.
Reflection
A Skill You Improved Through Practice
Here's something to think about. Pick a skill you improved through practice: a sport, an instrument, or a subject you once found hard. Which of the three kinds did you strengthen most? And what would be missing from a system that learned from data, but never checked its own mistakes?
About Introduction to Artificial Intelligence at UPI Study
Introduction to Artificial Intelligence (Computer Science 373) is a self-paced online course of 69 video lessons in 9 chapters, from intelligent agents and search to game strategy, neural networks, logic and reasoning, machine learning, language, robotics and AI ethics. It carries a NCCRS credit recommendation for 3 semester credits.
Every lesson transcript is free on pages like this one, and the videos are free with a free UPI Study account. Enrolling adds a practice quiz after each lesson, the graded work and final exam that earn the credit, lifetime access, and an official transcript on completion. UPI Study has helped over 130,000 students with college credits. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
Enroll now → Course details Download the syllabus All Computer Science courses Every course with PRO →
Compare: AI vs Current Trends in CS and IT
Why take it here
Why students take Introduction to Artificial Intelligence with UPI Study
Where the credit goes
Universities that work with NCCRS
Introduction to Artificial Intelligence carries a NCCRS credit recommendation. These universities work with NCCRS; some are UPI Study's signed partners. ACE and NCCRS have over 2,100 cooperating colleges that will consider accepting credits; the final decision is with the universities.
Thomas Edison State UniversitySigned partner
University of the PeopleSigned partner
Charter Oak State CollegeSigned partner
Strayer UniversitySigned partner
Columbia Southern UniversitySigned partner
Indiana TechSigned partner
Penn StateWorks with NCCRS
Southern New Hampshire UniversityWorks with NCCRS
SUNYWorks with NCCRS
+many more cooperating universitiesFind my college →Find my college → Not listed? Ask Athena in the corner of this page, or book a 15-minute call.
FAQ
Questions students ask us
About Learnable Intelligence: Insights from Perkins' Framework
What are Perkins' three kinds of intelligence?
What does Perkins mean by mindware?
What are hugging and bridging in teaching for transfer?
How does Perkins' theory of intelligence apply to artificial intelligence?
About the Introduction to Artificial Intelligence course
How many college credits is Introduction to Artificial Intelligence?
How much does Introduction to Artificial Intelligence cost and what is included?
Which universities accept UPI Study credits?
How long does Introduction to Artificial Intelligence take?
How are the exams done?
How do my credits reach my college?
Ready for the full course? Earn 3 credits in Introduction to Artificial Intelligence.
69 video lessons, a quiz after each one, and the graded work that earns the credit. Recommended for credit by NCCRS.
3 credits · NCCRSEnroll →