Periodic Reporting for period 1 - HyperModeLex (Hyperdimensional Modelling of the Legal System in Digital Society)
Okres sprawozdawczy: 2022-10-01 do 2025-03-31
Can the Law be converted into programming code using AI/ML/LLM and logic formulas without losing legal theory principles, legal linguistic expressivity, Democratic and Constitutional principles? Can a whole Legal System be managed digitally by using knowledge Graphs, Semantic Web techniques, Legal ontologies, AI? Can the translation be made automatically executable using Smart Contracts and immediately enforceable? How does the digitalization of the Law simple, transparent, and accountable for the citizens? Can Legal Design help in this task of communication?
This project aims to answer these questions using interdisciplinary instruments from philosophy of law, constitutional law, legal informatics, including AI&Law, computational linguistics, computer science, HCI, and Legal design. We use Hybrid AI to mitigate the weakness of symbolic AI in semantics.
The project is organised in 5 main sub-projects of research.
1. Analyse post-reductionism/textualism/normativism of philosophy of law in infosphere
Legal text serialization is only one of the multiple representations of the legal normative content (e.g. oral, picture). The Law is a complex set of ingredients: law-making process, legal actions, legal speech acts (e.g. text), political nuances (e.g. intention of the legislator), cultural elements, linguistic implicit concepts (e.g. new terms), different signifieds (e.g. interpretation), logic formula (e.g. commands). We investigate the theoretical relationship between normative statements, textual provisions, logic formula, non-linguistic formats (e.g. icons), programming coding of legal rules with the aim to define, and model, objectivity and subjectivity, explicit and implicit, linguistic and non-linguistic in relation of Legal Sources (e.g. Official Gazette).
2. Include Legal Hermeneutic in eLegislation
We use a new approach based on the interactive dialogue between software applications (text and code) to include the legal interpretation theory in the digital transformation of legal sources, especially in the GenAI era. We implement the explicability of the AI (Guidotti) by relying on argumentation theory to inject interpretation values into the computable system (Sartor, Prakken, Governatori, Rotolo, Boella, van der Torre, van Engers). Computer science techniques like XML, LOD, AI & Law, Legal reasoning, LLM/ML, Knowledge Graph, NLP, HCI and XR Interactive Systems, help to model a framework where to include flexibility of interpretation to prevent the crystallisation of the law in rigid coding.
3. Integrate Legal language role in normativism with computational linguistics models
There is a dense debate on the language of law (Oppenheim, Hart, Bobbio, Ross, Scarpelli, Olivecrona, Conte, Kalinowski, Wróblewski, Marmor, Schauer) and how an abstract statement (norm) is transformed in textual proposition (speech acts) and how it evolves (e.g. interpretation), and finally how logic language can model it formally. We involve computational linguists and semiotic experts working with philosophers of law to investigate the proxy role of the legal language in normative and prescriptive action.
4. Define Constitutional legitimacy of the digital legal sources and its e-enforceability
We investigate the constitutional and parliamentary law perspective (Lupo, Pollicino) in case we support the thesis that the coding law is a legitimate, authentic, and official Legal Source (Official Gazette). It is also important to understand the enforceability of the consumable-Law and whether DLT technology (e.g. smart contract) permits to implement automatic enforceability (e-enforceability).
5. Implement Better Regulation with Legal Design
We investigate which simplification forms we can adopt to visualize the machine-consumable law in a human-consumable manner for implementing simplification, better regulation principles, transparency, and accessibility. Visualization is another form of translation of the legal, so we understand which Human-Computer-Interaction instruments we can use to implement the Legal Design principles and preserve the normative message.
This project aims to answer these questions using interdisciplinary instruments from philosophy of law, constitutional law, legal informatics, including AI&Law, computational linguistics, computer science, HCI, and Legal design. We use Hybrid AI to mitigate the weakness of symbolic AI in semantics.
The project is organised in 5 main sub-projects of research.
1. Analyse post-reductionism/textualism/normativism of philosophy of law in infosphere
Legal text serialization is only one of the multiple representations of the legal normative content (e.g. oral, picture). The Law is a complex set of ingredients: law-making process, legal actions, legal speech acts (e.g. text), political nuances (e.g. intention of the legislator), cultural elements, linguistic implicit concepts (e.g. new terms), different signifieds (e.g. interpretation), logic formula (e.g. commands). We investigate the theoretical relationship between normative statements, textual provisions, logic formula, non-linguistic formats (e.g. icons), programming coding of legal rules with the aim to define, and model, objectivity and subjectivity, explicit and implicit, linguistic and non-linguistic in relation of Legal Sources (e.g. Official Gazette).
2. Include Legal Hermeneutic in eLegislation
We use a new approach based on the interactive dialogue between software applications (text and code) to include the legal interpretation theory in the digital transformation of legal sources, especially in the GenAI era. We implement the explicability of the AI (Guidotti) by relying on argumentation theory to inject interpretation values into the computable system (Sartor, Prakken, Governatori, Rotolo, Boella, van der Torre, van Engers). Computer science techniques like XML, LOD, AI & Law, Legal reasoning, LLM/ML, Knowledge Graph, NLP, HCI and XR Interactive Systems, help to model a framework where to include flexibility of interpretation to prevent the crystallisation of the law in rigid coding.
3. Integrate Legal language role in normativism with computational linguistics models
There is a dense debate on the language of law (Oppenheim, Hart, Bobbio, Ross, Scarpelli, Olivecrona, Conte, Kalinowski, Wróblewski, Marmor, Schauer) and how an abstract statement (norm) is transformed in textual proposition (speech acts) and how it evolves (e.g. interpretation), and finally how logic language can model it formally. We involve computational linguists and semiotic experts working with philosophers of law to investigate the proxy role of the legal language in normative and prescriptive action.
4. Define Constitutional legitimacy of the digital legal sources and its e-enforceability
We investigate the constitutional and parliamentary law perspective (Lupo, Pollicino) in case we support the thesis that the coding law is a legitimate, authentic, and official Legal Source (Official Gazette). It is also important to understand the enforceability of the consumable-Law and whether DLT technology (e.g. smart contract) permits to implement automatic enforceability (e-enforceability).
5. Implement Better Regulation with Legal Design
We investigate which simplification forms we can adopt to visualize the machine-consumable law in a human-consumable manner for implementing simplification, better regulation principles, transparency, and accessibility. Visualization is another form of translation of the legal, so we understand which Human-Computer-Interaction instruments we can use to implement the Legal Design principles and preserve the normative message.
The most important updated achievements are presented in the portal: https://site.unibo.it/hypermodelex/en/results(odnośnik otworzy się w nowym oknie)
The most relevant results are the following.
Analysis of the needs with interdisciplinary methodology
The first result is to create a preliminary analysis of the introduction of AI in Parliament using the instruments of the theory of law, constitutional law, and legal informatics.
HyperModeLex is very active in the legal informatics community (ICAIL, JURIX, XAI) and the hybrid AI methodology (HAIMLA) produces a significant outcome in the state of the art in the legal domain (e.g. Bússola https://www.bussola-tech.co/(odnośnik otworzy się w nowym oknie) Parliaments, Gazette, Governments).
The group participates in many events to improve dialogue with other AI regulations (beyond the AI Act), legal traditions, and technical backgrounds: Washington DC, Congress 10-11 October 2024, Australia Region Rules as Code” Community, Japan Digital Transformation Department of Government of Japan, Albanian Parliament and OSCE-Albania, and European institutions (EU Commissions, European Publication Office, European Parliament). We have a permanent Observatory, commenting and reporting the most important publications in the state of the art.
Results of the experiments
Our research combines a high level of technical knowledge with legal theory and democratic principles, producing a unique product of results, including the theory of language and visual communication methodologies. We have developed Hybrid AI models for managing the structure of legal sources, semantics, and hidden knowledge in legal language (e.g. definitions, normative references, and the similarity of arguments). In particular, we use Framenet to capture the semantics of the legislation without entering into the logic formalization. We have won a “Manifestation of Interest” launched by the Chamber of Deputies of Italy on “AI in Parliament” (see https://interoperable-europe.ec.europa.eu/collection/public-sector-tech-watch/genai4lex-b-ai-powered-legislative-support-italian-chamber-deputies(odnośnik otworzy się w nowym oknie)).
Guidelines for AI in Parliaments integrated with ethics and legal principles
The capacity to have a constructive dialogue with the stakeholders has allowed the group to reach outstanding results also in the elaboration of guidelines for the safe introduction of AI in the Parliament (see i) AI Guidelines for Parliaments, Westminster Foundation for Democracy: https://www.wfd.org/sites/default/files/2024-07/wfd-ai-guidelines-for-parliaments-2024-english.pdf(odnośnik otworzy się w nowym oknie); ii) Guidelines on online publication of case-law and legal knowledge access, CEPEJ, Council of Europe: https://rm.coe.int/da-2024-cepej-plenary-dec-1prov4-en/1680b294d3(odnośnik otworzy się w nowym oknie); iii) Semantic Categorization of Judicial Decisions in the Case Law Databases with Recommendations, TJENI Project, CEPEJ, Council of Europe: https://rm.coe.int/report-on-semantic-categorisation-of-judicial-decisions/1680aeb729(odnośnik otworzy się w nowym oknie)).
Groundbreaking research: LLM and Akoma Ntoso
The group works on training an LLM from scratch using LEONARDO Supercomputer of CINECA. The goal is to teach the LLM the semantic markup language for legal sources Akoma Ntoso to foster legal knowledge as a metalanguage. The group cooperates with the European Publication Office using LLM for markup of the plain text in Akoma Ntoso (AI4XML see https://code.europa.eu/ai4xml/playground(odnośnik otworzy się w nowym oknie)).
Analysis of the needs with interdisciplinary methodology
The first result is to create a preliminary analysis of the introduction of AI in Parliament using the instruments of the theory of law, constitutional law, and legal informatics.
HyperModeLex is very active in the legal informatics community (ICAIL, JURIX, XAI) and the hybrid AI methodology (HAIMLA) produces a significant outcome in the state of the art in the legal domain (e.g. Bússola https://www.bussola-tech.co/(odnośnik otworzy się w nowym oknie) Parliaments, Gazette, Governments).
The group participates in many events to improve dialogue with other AI regulations (beyond the AI Act), legal traditions, and technical backgrounds: Washington DC, Congress 10-11 October 2024, Australia Region Rules as Code” Community, Japan Digital Transformation Department of Government of Japan, Albanian Parliament and OSCE-Albania, and European institutions (EU Commissions, European Publication Office, European Parliament). We have a permanent Observatory, commenting and reporting the most important publications in the state of the art.
Results of the experiments
Our research combines a high level of technical knowledge with legal theory and democratic principles, producing a unique product of results, including the theory of language and visual communication methodologies. We have developed Hybrid AI models for managing the structure of legal sources, semantics, and hidden knowledge in legal language (e.g. definitions, normative references, and the similarity of arguments). In particular, we use Framenet to capture the semantics of the legislation without entering into the logic formalization. We have won a “Manifestation of Interest” launched by the Chamber of Deputies of Italy on “AI in Parliament” (see https://interoperable-europe.ec.europa.eu/collection/public-sector-tech-watch/genai4lex-b-ai-powered-legislative-support-italian-chamber-deputies(odnośnik otworzy się w nowym oknie)).
Guidelines for AI in Parliaments integrated with ethics and legal principles
The capacity to have a constructive dialogue with the stakeholders has allowed the group to reach outstanding results also in the elaboration of guidelines for the safe introduction of AI in the Parliament (see i) AI Guidelines for Parliaments, Westminster Foundation for Democracy: https://www.wfd.org/sites/default/files/2024-07/wfd-ai-guidelines-for-parliaments-2024-english.pdf(odnośnik otworzy się w nowym oknie); ii) Guidelines on online publication of case-law and legal knowledge access, CEPEJ, Council of Europe: https://rm.coe.int/da-2024-cepej-plenary-dec-1prov4-en/1680b294d3(odnośnik otworzy się w nowym oknie); iii) Semantic Categorization of Judicial Decisions in the Case Law Databases with Recommendations, TJENI Project, CEPEJ, Council of Europe: https://rm.coe.int/report-on-semantic-categorisation-of-judicial-decisions/1680aeb729(odnośnik otworzy się w nowym oknie)).
Groundbreaking research: LLM and Akoma Ntoso
The group works on training an LLM from scratch using LEONARDO Supercomputer of CINECA. The goal is to teach the LLM the semantic markup language for legal sources Akoma Ntoso to foster legal knowledge as a metalanguage. The group cooperates with the European Publication Office using LLM for markup of the plain text in Akoma Ntoso (AI4XML see https://code.europa.eu/ai4xml/playground(odnośnik otworzy się w nowym oknie)).