Word Associations: How Computer Science can Help us Gain Insight into the Human Brain

by Jule Schatz 

What is the first thing you think of when you see the word “mouse”?
What about when you see “dairy”?
Most likely you all gave pretty different answers for the two word prompts above.
Now what about “cake”? What word comes to mind?
I can guess that most of you thought of “cheese”.

How come I was able to guess? This is a psychological effect called priming. The connection you made for “cake” was influenced by the words that you had previously seen, “mouse” and “dairy”. Because you were thinking about “mouse” and “dairy”, your mind was already thinking about “cheese”. So, when I asked you to connect something with “cake”, that’s where your mind went.

Word associations can be a fun way to learn about how people think. People have also used word associations to create fun games. One popular example is the game Codenames (pictured below). It involves coming up with word associations and trying to guess what associations other people will make.

Another game involving word associations is the Remote Associates Test (the RAT). In this game, a player is presented with three words and has to come up with a fourth word that relates to all three. Here is one example I think we can all get correct after our earlier priming!
Cake, cottage, Swiss. . . .
The answer is cheese! Here are some more for you to try.
Man, glue, star
Dust, cereal, fish
Way, board, sleep
See the answers at the bottom of the article.

These can be fun brain teasers, but they also bring up questions of how humans solve the RAT. Did you find one of these examples to be harder than the others? Why was that? What parts of the brain are involved? Does priming relate to how people perform on the RAT?

One way to approach these questions is to use computer science. This can be done by creating a model of the human brain on a computer and seeing what settings and directions make it behave the most human like.

Soar is a cognitive architecture that can be used to model the human brain that can perform task such as the RAT. Soar allows us to test the hypothesis that priming helps humans solve the RAT. In Soar, you can create a program that uses priming and another that doesn’t when solving RAT problems. By comparing these results to human performance, we gain information on how humans solve the RAT.

From running such experiments in Soar we found that priming does result in more human like performance on RAT problems. This is evidence that humans do rely on priming when solving RAT problems. This is a concrete example of how we can use computer science to gain insights into the human brain.

If you want more information on my research in the use of Soar for solving RAT problems, check out the paper.
You can also go to the github where all the code for the project lives!

More information on priming: https://en.wikipedia.org/wiki/Priming_(psychology)

More information on Soar: https://soar.eecs.umich.edu/

More information on General AI and cognitive architectures: http://www.cogsys.org/pdf/paper-1-2.pdf

RAT Answers: super, bowl, walk