Using modified incremental chart parsing to ascribe intentions to animated geometric figures.

Area of Science:

• Cognitive Science
• Artificial Intelligence
• Psychology

Background:

• Humans spontaneously attribute intentions based on observed behavior.
• This attribution occurs even with simple geometric shapes in 2-D animations.
• Object movement in animations provides critical information for inferring intentions.

Purpose of the Study:

• To develop a computational model for inferring intentions from observed movement.
• To utilize spatiotemporal constraints and motion information for intention ascription.
• To adapt natural-language processing techniques for analyzing visual behavior.

Main Methods:

• Modified incremental chart parsing from natural-language processing.
• Developed a system using spatiotemporal constraints of figure movement.
• Employed a rule-based system to propose and rank candidate intentions or causes.
• Implemented confidence scoring and dynamic revision of candidates based on incoming observations.

Main Results:

• The model successfully proposes candidate intentions or non-agentive causes for observed movements.
• Confidence scores allow for ranking of potential interpretations.
• The system dynamically updates interpretations as new motion data becomes available.
• A pilot study confirmed that human perception aligns with the model's predictions.

Conclusions:

• Motion information is a key factor in attributing intentions to observed behavior.
• Computational models can effectively simulate human-like intention ascription.
• Adapted natural language processing techniques show promise in analyzing visual motion for cognitive modeling.