Thumbnail Image
Item

Exploring the Criteria for Artificial Consciousness: Implementing and Evaluating Agentic Software Systems in Virtual Environments

Le, Bach
Citations
Altmetric:
Contributor
Photographer
Author
Le, Bach
Artist
Editor
Advisor
Keywords
Artificial intelligence—Philosophy, Consciousness—Computer simulation, Intelligent agents (Computer software), Machine learning, Cognitive science
Local ID
Collections
Honors Papers
Show all metadata
Files
Thumbnail Image
2026_honors_thesis_le.pdf
Adobe PDF, 3.86 MB
URI
https://hdl.handle.net/10267/37259
Abstract
This honors thesis investigates the possibility of artificial consciousness in software agents by evaluating and implementing systems based on a set of proposed “indicator properties” derived from neuroscientific theories of consciousness. The study analyzes contemporary reinforcement learning algorithms to determine which properties they satisfy, compares their behavioral performance, and explores both bottom-up and top-down approaches to designing conscious-like systems, including modifications to existing models and the use of the LIDA cognitive architecture. Findings reveal a significant gap between theoretical criteria for consciousness and their practical realization in computational systems, as agents that satisfy more indicator properties do not necessarily perform better in task-oriented environments. The project concludes that while the indicator property framework offers a promising structure for evaluating artificial consciousness, it remains difficult to operationalize and requires further refinement to be practically useful in AI development.
Description