Better AI through Logical Scaffolding

TL;DR

This paper introduces the concept of logical scaffolds to enhance AI software quality, demonstrating their application in perception, prediction, and agent behavior systems to improve reliability and performance.

Contribution

It proposes logical scaffolds as a unifying framework for improving AI systems across various domains, extending existing runtime monitoring ideas.

Findings

Logical scaffolds can be applied to perception systems for better reliability

They are useful for general prediction and agent behavior models

The approach unifies and extends runtime monitoring concepts

Abstract

We describe the concept of logical scaffolds, which can be used to improve the quality of software that relies on AI components. We explain how some of the existing ideas on runtime monitors for perception systems can be seen as a specific instance of logical scaffolds. Furthermore, we describe how logical scaffolds may be useful for improving AI programs beyond perception systems, to include general prediction systems and agent behavior models.