# Sequence-to-sequence models with attention mechanistically map to the architecture of human memory search

## TL;DR

This study shows that machine learning models with attention mechanisms work similarly to how humans search their memory, offering new ways to understand and model memory.

## Contribution

The study reveals that sequence-to-sequence models with attention mechanistically align with human memory search architectures.

## Key findings

- Sequence-to-sequence models with attention mirror the CMR model of human memory.
- The model captures both average and optimal human memory behaviors using a free recall dataset.
- The model's performance emerges from interactions between its components.

## Abstract

Past work has long recognized the important role of context in guiding how humans search their memory. While context-based memory models can explain many memory phenomena, it remains unclear why humans develop such architectures over possible alternatives in the first place. In this work, we demonstrate that foundational architectures in neural machine translation – specifically, recurrent neural network (RNN)-based sequence-to-sequence models with attention – exhibit mechanisms that directly correspond to those specified in the Context Maintenance and Retrieval (CMR) model of human memory. Since neural machine translation models have evolved to optimize task performance, their convergence with human memory models provides a deeper understanding of the functional role of context in human memory, as well as presenting alternative ways to model human memory. Leveraging this convergence, we implement a neural machine translation model as a cognitive model of human memory search that is both interpretable and capable of capturing complex dynamics of learning. We show that our model accounts for both averaged and optimal human behavioral patterns as effectively as context-based memory models using a publicly available free recall experiment dataset involving 171 participants. Further, we demonstrate additional strengths of the proposed model by evaluating how memory search performance emerges from the interaction of different model components.

This study shows that foundational architectures in machine learning, sequence-to-sequence models with attention, mirror mechanisms of human memory. They can serve as alternative memory models, capturing behavior and aiding performance understanding.