Modeling the Remote Associates Test as Retrievals from Semantic Memory

Published on June 4, 2022

Imagine your brain as a vast library of word associations. The Remote Associates Test (RAT) is like a treasure hunt through this library, where you’re given three seemingly unrelated words and challenged to find the one word that connects them all. In this study, researchers dive into the mechanics of our semantic memory to understand how we retrieve these relevant associations. By creating a computational model based on human problem-solving, they explore the impact of prior knowledge and memory retrieval processes on matching human behavior. They even expand their investigation to multiple knowledge bases and incorporate concepts from cognitive science. Ultimately, they find that the success of the RAT depends on factors like the strength and direction of associations between prompt words and solutions, as well as our ability to perform multiple retrievals. This research sheds light on the intricate workings of our memory and offers valuable insights for fields like psychology and education! So, grab your thinking cap and delve into the full article to uncover the secrets of how your brain retrieves information.

Abstract
The Remote Associates Test (RAT) is a word association retrieval task that consists of a series of problems, each with three seemingly unrelated prompt words. The subject is asked to produce a single word that is related to all three prompt words. In this paper, we provide support for a theory in which the RAT assesses a person’s ability to retrieve relevant word associations from long-term memory. We present a computational model of humans solving the RAT and investigate how prior knowledge and memory retrieval mechanisms influence the model’s ability to match human behavior. We expand prior modeling attempts by investigating multiple large knowledge bases and by creating a cognitive process model that uses long-term memory spreading activation retrieval processes inspired by ACT-R and implemented in Soar. We evaluate multiple model variants for their ability to model human problem difficulty, including the incorporation of noise and base-level activation into memory retrieval. We conclude that the main factors affecting human difficulty are the existence of associations between prompt words and solutions, the relative strengths and directions of those associations compared to associations to other words, and the ability to perform multiple retrievals.

Read Full Article (External Site)

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes:

<a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>