Predicting Brain Activity Associated with Complex Nouns: Designing an Incentive Compatible Mechanism Open Access

Abstract

Abstract
Predicting Brain Activity Associated with Complex Nouns: Designing an Incentive Compatible Mechanism
By Matthew G. Rubin
New methods of neuroimaging and machine learning have recently been utilized to reveal
underlying neural dimensions of simple noun representation. One bottom-up procedure, multi-voxel pattern
analysis (MVPA), predicts cognitive states by detecting spatial patterns in brain data and correlating
differences in neural activity with behavioral responses. This technique has proved successful with simple
nouns like tools and vegetables, but has never been attempted with complex nouns which incentives may
exist for subjects not to be truthful. We used functional MRI (fMRI) to investigate the neural representation
of Identities, or labels that describe people, and whether an Identity's meaning can be characterized by its
association with actions or attributes. We then trained 3 classifiers to differentiate between Identities based
on subjects' ratings of an Identity's actions and attributes, as well as subjects' sentiments about how good
or bad each Identity was. Although the classifier results varied tremendously by Identity word and had
slightly inflated accuracy level significance because the training and testing data were not independent, the
many innovations that were introduced foreshadow an optimistic future for pattern analysis classification.

Table of Contents

TABLE OF CONTENTS

Introduction ....................................................................................................................... 1
Materials and Methods ........................................................................................................ 8
Experimental Paradigm and Task .............................................................................................. 8
fMRI Data ........................................................................................................................... 9
fMRI Analysis ...................................................................................................................... 10
Voxel Selection ................................................................................................................... 10
Creation of the Questionnaires ............................................................................................... 11
Questionnaire Word Selection................................................................................................. 11
Text Corpus Data ................................................................................................................ 11
Verbs ................................................................................................................................ 12
Adjectives ......................................................................................................................... 12
Word Check ....................................................................................................................... 13
Machine Learning Methods ................................................................................................... 13
Overview .......................................................................................................................... 13
Feature Selection ............................................................................................................... 14
Classifier Training and Testing ............................................................................................... 15
Results ............................................................................................................................ 18
Behavioral Statistics ........................................................................................................... 18
Classifier Performance ......................................................................................................... 19
Binary Good/Bad Classification .............................................................................................. 19
Continuous Verb/Adjective Regression .................................................................................... 21
Continuous Evaluation/Semantic-Combination Regression ........................................................... 23
Discussion ....................................................................................................................... 25
References ...................................................................................................................... 30
Appendix ......................................................................................................................... 34

FIGURES AND TABLES
Figure 1. Experimental Design in the Scanner ............................................................................ 9
Figure 2. Classifier 1 Accuracy .............................................................................................. 19
Figure 3. Classifier 2 Accuracy .............................................................................................. 22
Figure 4. Classifier 3 Accuracy .............................................................................................. 25

Table 1. Mean Behavioral Statistics ....................................................................................... 18
Table 2. Classifier 1: Mean Classification Accuracy ................................................................... 20
Table 3. Classifier 1: Best Classified Identities ......................................................................... 21
Table 4. Classifier 2: Mean Power and Significance ................................................................... 22
Table 5. Classifier 3: Mean Power and Significance ................................................................... 24

About this Honors Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Emory College
Department	Economics and Mathematics
Degree	BA
Submission	Honors Thesis
Language	English
Research Field	Biology, Neuroscience Economics, General Artificial Intelligence
Keyword	Incentive compatibility Neuroeconomics fMRI Machine learning: pattern recognition Experimental economics
Committee Chair / Thesis Advisor	Berns, Gregory S, Emory University
Committee Members	Prietula, Michael J, Emory University Hamilton, Emily, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Predicting Brain Activity Associated with Complex Nouns: Designing an Incentive Compatible Mechanism ()	2018-08-28 13:55:44 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions