Reproduce “A Measure for Level of Autonomy Based on Observable System Behavior” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jason M. Pittman
Researchers at the University of Maryland Global Campus propose a theoretical framework for measuring the Level of Autonomy (LoA) of autonomous systems based solely on observable runtime behavior. This framework uses an edit distance function to compare observed system actions against a human behavior baseline, where a greater edit distance implies a higher inferred level of autonomy, enabling real-time assessment without access to internal system data.
Key insights
- The level of autonomy in an observed system can be inferred by quantifying the dissimilarity between its external actions and a predefined human behavioral baseline.
- A greater edit distance between observed system actions and the human reference actions correlates with a higher predicted level of autonomy, suggesting divergence from direct human mimicry.
- It is possible to "blindly compare" autonomous systems at runtime and assess their autonomy without requiring prior knowledge of their internal states, design parameters, or proprietary information.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.