An ontological approach to foster the convergence, interoperability and operationalization of frameworks for Trustworthy AI
This paper introduces the AI-Ethics Ontology (AI-EO), a semantic infrastructure built on web technologies that aims to unify, interoperate, and operationalize various Trustworthy AI frameworks through a dynamic, iterative development process validated by case studies.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are trying to build a massive, global library for "Good AI."
Right now, we have a problem: Different countries and organizations are writing their own rulebooks on how to make Artificial Intelligence safe, fair, and trustworthy.
- Australia has written a list of 8 rules.
- The European Union has written a much longer, more complex law with pillars of "Lawful, Ethical, and Robust."
- Tech companies have their own internal checklists.
The problem is that these rulebooks are written in different languages, use different words for the same things, and are organized differently. If a developer in Australia wants to check if their AI is "Fair," they might look at a definition that doesn't match the EU's definition of "Fair." It's like trying to build a house where the blueprints from the architect, the electrician, and the plumber all use different symbols for "light switch." You end up with a confusing mess.
This paper introduces a solution called the "AI-Ethics Ontology" (AI-EO).
Think of AI-EO not as a new rulebook, but as a universal translator and a master filing system for all these different AI ethics rules.
The Core Idea: The "Rosetta Stone" for AI Rules
The author, Salvatore Pileggi, created a digital "map" (called an Ontology) that connects all these different rulebooks together.
Here is how it works, using a simple analogy:
1. The "Dictionary" Problem
Imagine the EU says, "You must be Accountable." Australia says, "You must be Responsible."
To a computer, these are two totally different words. To a human, they mean almost the same thing.
- Without AI-EO: A computer checking for "Accountability" would miss the Australian rule about "Responsibility."
- With AI-EO: The system knows that "Accountability" and "Responsibility" are twins. It links them together. Now, if you ask the system, "Is this AI accountable?" it checks both the EU and Australian rules automatically.
2. The "Living Map" (Iterative Evolution)
Usually, when you write a rulebook, it's static. Once it's printed, it's done. But AI changes every day. New types of AI (like "Agentic AI" that acts on its own) appear constantly.
- The Old Way: You wait years to update the law.
- The AI-EO Way: The author designed this system to be agile. It's like a Google Map that updates in real-time.
- Step 1: The system takes the Australian rules and maps them.
- Step 2: It takes the EU rules and maps them.
- Step 3: It finds where they overlap and where they differ, creating a "federated view" (a single, unified dashboard).
- Step 4: If a new rule comes out tomorrow, the system can just add it to the map without breaking the whole thing.
3. The "Smart Filing Cabinet"
The paper describes a process where the system uses AI to read these rulebooks, pick out the important keywords (like "Fairness," "Privacy," "Safety"), and file them into a structured database.
- Human Supervision: Humans act as the librarians to make sure the AI files things correctly.
- Semantic Enrichment: The system doesn't just file the word; it adds notes, links to the original law, and explains why it matters.
Why Does This Matter? (The "So What?")
Imagine you are a company trying to launch a new AI chatbot. You want to make sure you aren't breaking any rules anywhere in the world.
- Before AI-EO: You hire a team of 50 lawyers to read every single AI ethics document on the planet. It takes months and costs a fortune.
- With AI-EO: You plug your AI into this "Universal Translator." It instantly scans your code against the unified map of all global rules. It tells you: "Hey, your chatbot is great for the EU, but you missed a specific requirement for 'Transparency' that Australia requires."
The "Secret Sauce": Descriptive vs. Prescriptive
The author emphasizes that this system is descriptive, not prescriptive.
- Prescriptive is like a strict teacher saying, "You must do it this way."
- Descriptive is like a helpful guide saying, "Here is how everyone else is doing it, and here is how your idea fits in."
This makes the system flexible. It doesn't force everyone to think exactly the same way; it just helps them understand each other better.
Summary
In simple terms, this paper presents a digital bridge.
- The Problem: AI ethics rules are scattered, confusing, and hard to compare.
- The Solution: A smart, computer-readable map (Ontology) that translates between different rulebooks.
- The Result: A world where we can build safer, more trustworthy AI faster because everyone is speaking the same "ethical language," even if their original rulebooks were written differently.
It's like turning a chaotic room full of people speaking different languages into a room where everyone has a universal translator, allowing them to finally agree on how to build a safe future together.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.