A Survey on Verification and Validation, Testing and Evaluations of Neurosymbolic Artificial Intelligence

TL;DR

This survey addresses the challenge of testing, validating, verifying, and evaluating neurosymbolic AI by examining two prominent taxonomies (Kautz and Yu) that categorize how symbolic and sub-symbolic components interact. It maps V&V concepts to symbolic logic, knowledge graphs, and sub-symbolic models, and reviews current V&V techniques for both parts, assessing their applicability to neurosymbolic architectures. The analysis shows that symbolic components can enhance the V&V of sub-symbolic predictions, but some architectures require novel or adjusted methods and dedicated testing frameworks. The work highlights opportunities such as safety-focused reinforcement learning, policy extraction for interpretability, and knowledge-graph-centric validation, while also outlining open challenges and the need for domain-specific testing frameworks to advance trustworthy neurosymbolic AI.

Abstract

Neurosymbolic artificial intelligence (AI) is an emerging branch of AI that combines the strengths of symbolic AI and sub-symbolic AI. A major drawback of sub-symbolic AI is that it acts as a "black box", meaning that predictions are difficult to explain, making the testing & evaluation (T&E) and validation & verification (V&V) processes of a system that uses sub-symbolic AI a challenge. Since neurosymbolic AI combines the advantages of both symbolic and sub-symbolic AI, this survey explores how neurosymbolic applications can ease the V&V process. This survey considers two taxonomies of neurosymbolic AI, evaluates them, and analyzes which algorithms are commonly used as the symbolic and sub-symbolic components in current applications. Additionally, an overview of current techniques for the T&E and V&V processes of these components is provided. Furthermore, it is investigated how the symbolic part is used for T&E and V&V purposes in current neurosymbolic applications. Our research shows that neurosymbolic AI as great potential to ease the T&E and V&V processes of sub-symbolic AI by leveraging the possibilities of symbolic AI. Additionally, the applicability of current T&E and V&V methods to neurosymbolic AI is assessed, and how different neurosymbolic architectures can impact these methods is explored. It is found that current T&E and V&V techniques are partly sufficient to test, evaluate, verify, or validate the symbolic and sub-symbolic part of neurosymbolic applications independently, while some of them use approaches where current T&E and V&V methods are not applicable by default, and adjustments or even new approaches are needed. Our research shows that there is great potential in using symbolic AI to test, evaluate, verify, or validate the predictions of a sub-symbolic model, making neurosymbolic AI an interesting research direction for safe, secure, and trustworthy AI.

Figures (8)

• Figure 1: Contents of this paper.
• Figure 2: Flowchart of the Learning for Reasoning type of neurosymbolic AI. The sub-symbolic component is used to limit the search space for the symbolic part. Therefore, it is accelerating the process recent_advances.
• Figure 3: Flowchart of the Learning for Reasoning type of neurosymbolic AI. In this version of Learning for Reasoning, the sub-symbolic part transforms the knowledge that can be obtained from data to symbols recent_advances.
• Figure 4: Flowchart of the Reasoning for Learning type of neurosymbolic AI. Here, the symbolic part can guide or constrain the sub-symbolic part recent_advances.
• Figure 5: Flowchart of the Learning-Reasoning type of neurosymbolic AI. Here, the characteristics of the other architectures are combined and the two parts are in constant interaction recent_advances.