Skip to content Skip to sidebar Skip to footer

Example of Fluid Intelligence and Crystallized Intelligence Psychology

Let's test your intelligence. So, what is the capital of France? Why, Paris! Mais oui! OK, now complete the sequence: A 1 B 2 C. Did you say 3? Yeah, that's pretty easy. But that wasn't the real quiz. Here's what I really want to know: which form of thinking did you use to solve each question?

Well, recalling previously acquired information like a capital city is a form of intelligence known as "Crystallized Intelligence". Whereas the use of reasoning and logic to deduce the next character in a sequence is known as "Fluid Intelligence".

Example of Fluid Intelligence and Crystallized Intelligence Psychology
image source:

Crystalized intelligence refers to knowledge that comes from the previously acquired information. It's dependent on a person's knowledge, on their skills, on their expertise developed over a lifetime of learning and experience. 

Crystalized intelligence is fact and experience-based. Whereas fluid intelligence is the capacity to think logically and solve problems in new and unfamiliar situations independent of acquired knowledge, fluid intelligence represents a person's ability to problem solve using reasoning and by also using logic.

When you come across a new problem that your current knowledge can't address your call on fluid intelligence to resolve it. Now, the notion of fluid and crystallized intelligence dates back to 1963, when it was first formalized by psychologist Raymond Cattell. Now you might be asking yourself, "Why am I talking about psychology in an IBM Technology video?" Well, these two forms of intelligence that we use to solve problems every day can also play an important role in machine learning.

For example, consider an AI system like IBM Watson. To answer a question it can sift through like a million books per second. And perhaps the answer to your question is right there in one of these books, like our Paris example. In that case, it's a simple case of natural language understanding to answer the question, "What is the capital of France?" That's crystallized intelligence.

But most problems aren't that simple. And to solve them, we may need to combine both crystallized and fluid intelligence together. So, for example, say we wanted an AI travel system. Let's build one. An AI travel system. And what we want it to do is to build us an itinerary of the best way to spend the day in Paris, so the output of this is an itinerary.

We need to use our knowledge of Parisian geography and cultural history to build a corpus of what's available. That's the crystalized intelligence part, but we also need to apply those options to a derived understanding of the types of activities that we like to do. So, these are the sorts of things that we'd be interested in doing.

What do we normally do when we visit a new city? Other comparable options here in Paris? Can we tailor this to our personality, our budget, and our willingness to try new things? All of that stuff, well, that's the fluid intelligence side of things. So from there, a system could derive a subset of all the possible things to do in Paris that day, distilled into a tailored, personalized itinerary.

For us humans, our crystalized intelligence is the knowledge we acquire, and our fluid intelligence is how we apply it. For an AI system, you can think of a systems model as being crystalized intelligence because it teaches itself to do one task really well by training on massive datasets of prior experience.

Then you can think of its ability to solve new problems as being fluid intelligence because it can apply that model to a previously unseen problem. And as for me, my crystallized and fluid intelligence is telling me that morning croissants under the Eiffel Tower should be top of my list. If you have any questions, please drop us a line below.