Error-margin Analysis for Hidden Neuron Activation Labels

Understanding how high-level concepts are represented within artificial neural networks is a fundamental challenge in the field of artificial intelligence. While existing literature in explainable AI emphasizes the importance of labeling neurons with concepts to understand their functioning, they mostly focus on identifying what stimulus activates a neuron in most cases, this corresponds to the notion of recall in information retrieval. We argue that this is only the first-part of a two-part job, it is imperative to also investigate neuron responses to other stimuli, i.e., their precision. We call this the neuron labels error margin.

On the Value of Labeled Data and Symbolic Methods for Hidden Neuron Activation Analysis

A major challenge in Explainable AI is in correctly interpreting activations of hidden neurons: accurate interpretations would help answer the question of what a deep learning system internally detects as relevant in the input, demystifying the otherwise black-box nature of deep learning systems. The state of the art indicates that hidden node activations can, in some cases, be interpretable in a way that makes sense to humans, but systematic automated methods that would be able to hypothesize and verify interpretations of hidden neuron activations are underexplored. This is particularly the case for approaches that can both draw explanations from substantial background knowledge, and that are based on inherently explainable (symbolic) methods. In this paper, we introduce a novel model-agnostic post-hoc Explainable AI method demonstrating that it provides meaningful interpretations. Our approach is based on using a Wikipedia-derived concept hierarchy with approximately 2 million classes as background knowledge, and utilizes OWL-reasoning-based Concept Induction for explanation generation. Additionally, we explore and compare the capabilities of off-the-shelf pre-trained multimodal-based explainable methods. Our results indicate that our approach can automatically attach meaningful class expressions as explanations to individual neurons in the dense layer of a Convolutional Neural Network. Evaluation through statistical analysis and degree of concept activation in the hidden layer show that our method provides a competitive edge in both quantitative and qualitative aspects compared to prior work.

Concept Induction using LLMs: a user experiment for assessment

Explainable Artificial Intelligence (XAI) poses a significant challenge in providing transparent and understandable insights into complex AI models. Traditional post-hoc algorithms, while useful, often struggle to deliver interpretable explanations. Concept-based models offer a promising avenue by incorporating explicit representations of concepts to enhance interpretability. However, existing research on automatic concept discovery methods is often limited by lower-level concepts, costly human annotation requirements, and a restricted domain of background knowledge. In this study, we explore the potential of a Large Language Model (LLM), specifically GPT-4, by leveraging its domain knowledge and common-sense capability to generate high-level concepts that are meaningful as explanations for humans, for a specific setting of image classification. We use minimal textual object information available in the data via prompting to facilitate this process. To evaluate the output, we compare the concepts generated by the LLM with two other methods: concepts generated by humans and the ECII heuristic concept induction system. Since there is no established metric to determine the human understandability of concepts, we conducted a human study to assess the effectiveness of the LLM-generated concepts. Our findings indicate that while human-generated explanations remain superior, concepts derived from GPT-4 are more comprehensible to humans compared to those generated by ECII.

Towards Complex Ontology Alignment using Large Language Models

Ontology alignment, a critical process in the Semantic Web for detecting relationships between different ontologies, has traditionally focused on identifying so-called "simple" 1-to-1 relationships through class labels and properties comparison. The more practically useful exploration of more complex alignments remains a hard problem to automate, and as such is largely underexplored, i.e. in application practice it is usually done manually by ontology and domain experts. Recently, the surge in Natural Language Processing (NLP) capabilities, driven by advancements in Large Language Models (LLMs), presents new opportunities for enhancing ontology engineering practices, including ontology alignment tasks. This paper investigates the application of LLM technologies to tackle the complex ontology alignment challenge. Leveraging a prompt-based approach and integrating rich ontology content so-called modules our work constitutes a significant advance towards automating the complex alignment task.

Commonsense Ontology Micropatterns

The previously introduced Modular Ontology Modeling methodology (MOMo) attempts to mimic the human analogical process by using modular patterns to assemble more complex concepts. To support this, MOMo organizes organizes ontology design patterns into design libraries, which are programmatically queryable, to support accelerated ontology development, for both human and automated processes. However, a major bottleneck to large-scale deployment of MOMo is the (to-date) limited availability of ready-to-use ontology design patterns. At the same time, Large Language Models have quickly become a source of common knowledge and, in some cases, replacing search engines for questions. In this paper, we thus present a collection of 104 ontology design patterns representing often occurring nouns, curated from the common-sense knowledge available in LLMs, organized into a fully-annotated modular ontology design library ready for use with MOMo.

On the Psychology of GPT-4: Moderately anxious, slightly masculine, honest, and humble

We subject GPT-4 to a number of rigorous psychometric tests and analyze the results. We find that, compared to the average human, GPT-4 tends to show more honesty and humility, and less machiavellianism and narcissism. It sometimes exhibits ambivalent sexism, leans slightly toward masculinity, is moderately anxious but mostly not depressive (but not always). It shows human-average numerical literacy and has cognitive reflection abilities that are above human average for verbal tasks.

Conversational Ontology Alignment with ChatGPT

This study evaluates the applicability and efficiency of ChatGPT for ontology alignment using a naive approach. ChatGPT's output is compared to the results of the Ontology Alignment Evaluation Initiative 2022 campaign using conference track ontologies. This comparison is intended to provide insights into the capabilities of a conversational large language model when used in a naive way for ontology matching, and to investigate the potential advantages and disadvantages of this approach.

Understanding CNN Hidden Neuron Activations Using Structured Background Knowledge and Deductive Reasoning

A major challenge in Explainable AI is in correctly interpreting activations of hidden neurons: accurate interpretations would provide insights into the question of what a deep learning system has internally detected as relevant on the input, demystifying the otherwise black-box character of deep learning systems. The state of the art indicates that hidden node activations can, in some cases, be interpretable in a way that makes sense to humans, but systematic automated methods that would be able to hypothesize and verify interpretations of hidden neuron activations are underexplored. In this paper, we provide such a method and demonstrate that it provides meaningful interpretations. Our approach is based on using large-scale background knowledge approximately 2 million classes curated from the Wikipedia concept hierarchy together with a symbolic reasoning approach called Concept Induction based on description logics, originally developed for applications in the Semantic Web field. Our results show that we can automatically attach meaningful labels from the background knowledge to individual neurons in the dense layer of a Convolutional Neural Network through a hypothesis and verification process.

A Modular Ontology for MODS -- Metadata Object Description Schema

The Metadata Object Description Schema (MODS) was developed to describe bibliographic concepts and metadata and is maintained by the Library of Congress. Its authoritative version is given as an XML schema based on an XML mindset which means that it has significant limitations for use in a knowledge graphs context. We have therefore developed the Modular MODS Ontology (MMODS-O) which incorporates all elements and attributes of the MODS XML schema. In designing the ontology, we adopt the recent Modular Ontology Design Methodology (MOMo) with the intention to strike a balance between modularity and quality ontology design on the one hand, and conservative backward compatibility with MODS on the other.

An Ontology Design Pattern for Role-Dependent Names

We present an ontology design pattern for modeling Names as part of Roles, to capture scenarios where an Agent performs different Roles using different Names associated with the different Roles. Examples of an Agent performing a Role using different Names are rather ubiquitous, e.g., authors who write under different pseudonyms, or different legal names for citizens of more than one country. The proposed pattern is a modified merger of a standard Agent Role and a standard Name pattern stub.