Machine Understanding of Human Implicit Intention

Abstract

The project aimed at understanding implicit (un-represented or hidden) human intention. The implicit intention domain consists of two axes: the sympathy for one's represented intention and the sympathy for one's counterpart. This project focuses on the latter. The subjects were asked to read Korean statements on the screen and reply Yes (Sympathy/Agreement to the statement) or No (Non-sympathy/Disagreement). Korean has subject-object-verb structure and negation comes at the end. Thus, whether the statement is affirmative or negative is only known at the very end of sentence. The basic assumption is that the subjects make decision before reading to the end of sentence, which corresponds to the implicit intention. EEG, fMRI and pupil dilation signals were measured during this experiment. Experimental results indicate that there is clear difference between the activation levels of the measured signals, and it is possible to predict the subject's response with the accuracy of about 80% by SVM.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
May 18, 2013
Accession Number
ADA587007

Entities

People

  • Soo-Young Lee

Organizations

  • KAIST

Tags

Communities of Interest

  • Human Systems
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Accuracy
  • Agreements
  • Brain
  • Cognition
  • Cognitive Science
  • Computational Science
  • Decoding
  • Information Processing
  • Information Science
  • Kernel Functions
  • Language
  • Machine Learning
  • Neural Networks
  • Neurosciences
  • Psychology
  • Recognition
  • Supervised Machine Learning

Readers

  • Brain and Cognitive Science; Experimental Psychology; Cognitive Neuroscience
  • Computational Linguistics
  • Neural Network Machine Learning.