law.MIT.edu/research

# law.MIT.edu/research ### Images/artifacts folder: https://drive.google.com/drive/folders/1RepEF3-dzLjJ22P75P4lAf8jGnmjrz9v?usp=sharing ## Template **Project Name** **One Sentence Byline** **People** **Short Description** **Links and Artifacts** papers, github, youtube, etc **Preview Image** for the preview image # PROJECTS ## Tedx Talk: **Project Name** The Future of Law: AI Powered Justice? **One Sentence Byline** Robert Mahari explores how computational law research can help us reimagine legal practice in responsible ways. **People** Robert Mahari **Short Description** Almost all sectors of the economy have been reimagined through computation, automation, and AI. Meanwhile, law today is largely the same as it was decades ago. How can law embrace technology to become more efficient and accessible? **Links and Artifacts** https://www.ted.com/talks/robert_mahari_the_future_of_law_ai_powered_justice **Preview Image** (use video thumbnail) ## AutoLAW **Project Name** AutoLAW **One Sentence Byline** Legal precedent prediction using NLP: Identifying sentences of precedent relevant to any legal argument **People (projects lead(s), and collaborators, partners, funders, etc)** Robert Mahari **Short Description** AutoLAW demonstrates how NLP can be used to address an unmet need of the legal community and increase access to justice. The project predicts passages from precedential court decisions relevant to a new legal argument. To this end, we train a transformer model on 530,000 examples of legal arguments contained in U.S. federal judcial opinions. [insert image here: https://drive.google.com/file/d/1mPQQ5JGTw_o0iiRyc-CMf0iU1j4CnZaH/view?usp=share_link] In 96% of unseen test examples the correct target passage is among the top-10 predicted passages. AutoLAW can also predict relevant precedent given a short summary of a complex and unseen legal brief, predicting the precedent that was actually cited by the brief's co-author, former U.S. Solicitor General and current U.S. Supreme Court Justice Elena Kagan. [Other formulation: In 96% of previously unencountered test cases, the accurate target passage is found within the top 10 predicted passages. AutoLAW can also anticipate relevant legal precedents when provided with a brief summary of a complicated and unfamiliar legal document. This includes predicting precedents that were cited by the document's co-author, such as those from former U.S. Solicitor General and current U.S. Supreme Court Justice Elena Kagan.] **Links and Artifacts** https://arxiv.org/abs/2106.16034 **Preview Image** https://drive.google.com/file/d/1yl92eqxXlKkPLOvhNsP_vTAF_1d85rZ_/view?usp=share_link ## Legal Prompt Engineering **Project Name** Legal Prompt Engineering **One Sentence Byline** To make the most of generative AI systems like ChatGPT for law and legal processes inputting the rights prompts requires a blended understanding of both prompt engineering and legal engineering. This project explores methods and design patterns for achieving those results. **People (projects lead(s), and collaborators, partners, funders, etc)** Project Lead: Dazza Greenwood Collaborators: Megan Ma, Robert Mahari **Short Description** **Links and Artifacts** https://github.com/mitmedialab/Legal-Prompt-Engineering **Preview Image** ## Legal Dynamism **Project Name** Legal Dynamism **One Sentence Byline** Imagining a resulatory framework that is designed itteratively in response to real-world data **People (projects lead(s), and collaborators, partners, funders, etc)** Robert Mahari, Sandy Pentland **Short Description** 233 years ago, Thomas Jefferson called on us to reexamine laws in every generation. Advances in computational law allow us to do so in real-time. With this new ability, we have the opportunity to rethink how we define laws — as dynamic systems rather than as static rules. This article presents and explores this idea of legal dynamism. **Links and Artifacts** https://www.networklawreview.org/computational-one/ Related: https://sciencepolicyreview.org/2022/07/mitspr-191618003020/ **Preview Image** https://drive.google.com/file/d/1VeaCIvUaGc-axwrV8lXDS-kUc3uAg-67/view?usp=share_link ## Computational Antitrust **Project Name** Computational Antitrust **One Sentence Byline** As the economy becomes ever more data-driven, we find ourselves at a critical juncture that requires us to revisit antitrust practices to ensure healthy and sustainable competition. **People (projects lead(s), and collaborators, partners, funders, etc)** Robert Mahari, Sandro Lera, Sandy Pentland **Short Description** U.S. antitrust laws have repeatedly responded to the changing needs of the nation’s economy. As the marketplace grows ever more data-driven, we find ourselves at yet another critical economic juncture that requires us to revisit antitrust practices to ensure healthy and sustainable competition. In this article, we propose two new antitrust approaches that fit into the existing regulatory landscape, detect signs of anticompetitive behavior early, and handle the unique nature of the digital marketplace. [Embed Video: https://www.youtube.com/watch?v=oC_UyeQyD9M] **Links and Artifacts** https://law.stanford.edu/wp-content/uploads/2021/04/pentland-computational-antitrust-project.pdf **Preview Image** # MISC FUN here is the results of the IdeaFactory script thingy we played with: Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/ Requirement already satisfied: openai in /usr/local/lib/python3.9/dist-packages (0.27.2) Requirement already satisfied: requests>=2.20 in /usr/local/lib/python3.9/dist-packages (from openai) (2.27.1) Requirement already satisfied: tqdm in /usr/local/lib/python3.9/dist-packages (from openai) (4.65.0) Requirement already satisfied: aiohttp in /usr/local/lib/python3.9/dist-packages (from openai) (3.8.4) Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.9/dist-packages (from requests>=2.20->openai) (1.26.15) Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.9/dist-packages (from requests>=2.20->openai) (3.4) Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.9/dist-packages (from requests>=2.20->openai) (2.0.12) Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.9/dist-packages (from requests>=2.20->openai) (2022.12.7) Requirement already satisfied: frozenlist>=1.1.1 in /usr/local/lib/python3.9/dist-packages (from aiohttp->openai) (1.3.3) Requirement already satisfied: async-timeout<5.0,>=4.0.0a3 in /usr/local/lib/python3.9/dist-packages (from aiohttp->openai) (4.0.2) Requirement already satisfied: multidict<7.0,>=4.5 in /usr/local/lib/python3.9/dist-packages (from aiohttp->openai) (6.0.4) Requirement already satisfied: aiosignal>=1.1.2 in /usr/local/lib/python3.9/dist-packages (from aiohttp->openai) (1.3.1) Requirement already satisfied: yarl<2.0,>=1.0 in /usr/local/lib/python3.9/dist-packages (from aiohttp->openai) (1.8.2) Requirement already satisfied: attrs>=17.3.0 in /usr/local/lib/python3.9/dist-packages (from aiohttp->openai) (22.2.0) Enter your OpenAI API key: ·········· System: We are about to do a sequence of prompts with the intent of creating a great final product. Enter your prompt: Get lawyers to adopt AI powered tool to formulate their arguments such as brief, legal strategies, and oral arguments. The success criteria for this would be that at least a majority of lawyers regularly use these tools for this purpose. ------- User: Get lawyers to adopt AI powered tool to formulate their arguments such as brief, legal strategies, and oral arguments. The success criteria for this would be that at least a majority of lawyers regularly use these tools for this purpose. ------- Assistant: Great! Let's start with the first prompt: 1. What are the benefits of using AI-powered tools for lawyers to formulate their arguments? Some potential benefits could include: - Increased efficiency: AI-powered tools can help lawyers quickly analyze large amounts of data and identify relevant information, saving them time and effort. - Improved accuracy: By using AI to analyze data and identify patterns, lawyers can ensure that their arguments are based on the most relevant and up-to-date information available. - Better outcomes: By using AI to identify potential weaknesses in their arguments, lawyers can strengthen their cases and increase their chances of success. - Cost savings: By using AI to automate certain tasks, lawyers can reduce their workload and potentially lower their fees for clients. What other benefits can you think of? ------- User: Please design a course that would be broadly adopted, including how it achieved the enumerated success criteria. ------- Assistant: Sure, here's a course design that could help lawyers adopt AI-powered tools for formulating their arguments: Title: "AI for Legal Argumentation: How to Use Technology to Strengthen Your Case" Overview: This course is designed to help lawyers understand the benefits of using AI-powered tools for legal argumentation and provide them with the skills and knowledge they need to effectively incorporate these tools into their practice. Learning Objectives: - Understand the benefits of using AI-powered tools for legal argumentation - Identify the types of AI-powered tools available for legal argumentation - Learn how to use AI-powered tools to analyze data and identify relevant information - Understand how to integrate AI-powered tools into legal briefs, legal strategies, and oral arguments - Develop strategies for effectively communicating the use of AI-powered tools to clients and colleagues Course Outline: Module 1: Introduction to AI for Legal Argumentation - Overview of AI-powered tools for legal argumentation - Benefits of using AI-powered tools for legal argumentation - Examples of successful use cases Module 2: Types of AI-powered Tools for Legal Argumentation - Overview of different types of AI-powered tools available - How to choose the right tool for your needs - Case studies of successful tool implementation Module 3: Using AI-powered Tools to Analyze Data - How to use AI-powered tools to analyze large amounts of data - Identifying relevant information and patterns - Case studies of successful data analysis Module 4: Integrating AI-powered Tools into Legal Briefs, Legal Strategies, and Oral Arguments - How to incorporate AI-powered tools into legal briefs, legal strategies, and oral arguments - Best practices for using AI-powered tools in legal argumentation - Case studies of successful tool integration Module 5: Communicating the Use of AI-powered Tools to Clients and Colleagues - How to effectively communicate the use of AI-powered tools to clients and colleagues - Addressing concerns and questions about AI-powered tools - Case studies of successful communication strategies Assessment: - Completion of a final project that demonstrates the use of AI-powered tools in a legal argumentation scenario - A survey to assess the effectiveness of the course and the adoption of AI-powered tools among participants Success Criteria: - At least 80% of participants complete the course - At least 60% of participants report using AI-powered tools for legal argumentation within 6 months of completing the course - At least 50% of participants report increased efficiency, accuracy, and/or better outcomes as a result of using AI-powered tools for legal argumentation ------- User: now please design a completely different course that achieves the result in a completely different way ------- Assistant: Sure, here's another course design that could help lawyers adopt AI-powered tools for formulating their arguments: Title: "AI for Legal Argumentation: A Hands-On Workshop" Overview: This course is designed to provide lawyers with a hands-on experience using AI-powered tools for legal argumentation. Participants will work in small groups to analyze real-world legal cases and develop arguments using AI-powered tools. Learning Objectives: - Understand the benefits of using AI-powered tools for legal argumentation - Learn how to use AI-powered tools to analyze data and identify relevant information - Develop strategies for integrating AI-powered tools into legal argumentation - Gain experience working with AI-powered tools in a collaborative setting Course Outline: Module 1: Introduction to AI for Legal Argumentation - Overview of AI-powered tools for legal argumentation - Benefits of using AI-powered tools for legal argumentation - Examples of successful use cases Module 2: Hands-On Workshop: Analyzing Legal Cases with AI-powered Tools - Participants will work in small groups to analyze real-world legal cases using AI-powered tools - Participants will identify relevant information and patterns using AI-powered tools - Participants will develop legal arguments using AI-powered tools Module 3: Hands-On Workshop: Integrating AI-powered Tools into Legal Argumentation - Participants will work in small groups to integrate AI-powered tools into legal briefs, legal strategies, and oral arguments - Participants will develop strategies for effectively using AI-powered tools in legal argumentation - Participants will present their arguments to the group and receive feedback Module 4: Communicating the Use of AI-powered Tools to Clients and Colleagues - How to effectively communicate the use of AI-powered tools to clients and colleagues - Addressing concerns and questions about AI-powered tools - Case studies of successful communication strategies Assessment: - Completion of a final project that demonstrates the use of AI-powered tools in a legal argumentation scenario - A survey to assess the effectiveness of the course and the adoption of AI-powered tools among participants Success Criteria: - At least 80% of participants complete the course - At least 60% of participants report using AI-powered tools for legal argumentation within 6 months of completing the course - At least 50% of participants report increased efficiency, accuracy, and/or better outcomes as a result of using AI-powered tools for legal argumentation ------- User: compare and contrast the two courses and identify the strongest aspect of each and how well they are likely to achieve the success criteria ------- Assistant: The first course is a more traditional approach to learning, with a focus on providing participants with a comprehensive understanding of AI-powered tools for legal argumentation. It covers a range of topics, from the benefits of using AI-powered tools to the different types of tools available, and provides case studies to illustrate successful implementation. The course also includes a final project and survey to assess the effectiveness of the course and the adoption of AI-powered tools among participants. The strongest aspect of this course is its comprehensive approach to learning. By covering a range of topics and providing case studies, participants are able to gain a deep understanding of AI-powered tools for legal argumentation. The final project and survey also provide a way to measure the effectiveness of the course and the adoption of AI-powered tools among participants. The second course takes a more hands-on approach to learning, with a focus on providing participants with practical experience using AI-powered tools for legal argumentation. Participants work in small groups to analyze real-world legal cases and develop arguments using AI-powered tools. The course also includes a module on communicating the use of AI-powered tools to clients and colleagues. The strongest aspect of this course is its experiential learning approach. By working in small groups and using AI-powered tools to analyze real-world legal cases, participants are able to gain practical experience using these tools. The module on communicating the use of AI-powered tools is also valuable, as it helps participants address concerns and questions about AI-powered tools. Both courses are likely to achieve the success criteria, as they provide participants with the skills and knowledge they need to effectively incorporate AI-powered tools into their practice. However, the first course may be more effective at achieving the success criteria related to adoption, as it provides a more comprehensive understanding of AI-powered tools and includes a final project and survey to assess adoption. The second course may be more effective at achieving the success criteria related to increased efficiency and accuracy, as it provides participants with practical experience using AI-powered tools. ------- User: write a final plan that combines the best elements of each, including a syllabus outline and a example exercise, and describe how the plan achieves the success criteria ------- Assistant: Sure, here's a final plan that combines the best elements of each course: Title: "AI for Legal Argumentation: A Comprehensive and Hands-On Approach" Overview: This course is designed to provide lawyers with a comprehensive understanding of AI-powered tools for legal argumentation, as well as practical experience using these tools. Participants will work in small groups to analyze real-world legal cases and develop arguments using AI-powered tools, and will also learn how to integrate these tools into legal briefs, legal strategies, and oral arguments. Learning Objectives: - Understand the benefits of using AI-powered tools for legal argumentation - Identify the types of AI-powered tools available for legal argumentation - Learn how to use AI-powered tools to analyze data and identify relevant information - Understand how to integrate AI-powered tools into legal briefs, legal strategies, and oral arguments - Develop strategies for effectively communicating the use of AI-powered tools to clients and colleagues - Gain experience working with AI-powered tools in a collaborative setting Course Outline: Module 1: Introduction to AI for Legal Argumentation - Overview of AI-powered tools for legal argumentation - Benefits of using AI-powered tools for legal argumentation - Examples of successful use cases Module 2: Types of AI-powered Tools for Legal Argumentation - Overview of different types of AI-powered tools available - How to choose the right tool for your needs - Case studies of successful tool implementation Module 3: Hands-On Workshop: Analyzing Legal Cases with AI-powered Tools - Participants will work in small groups to analyze real-world legal cases using AI-powered tools - Participants will identify relevant information and patterns using AI-powered tools - Participants will develop legal arguments using AI-powered tools Module 4: Using AI-powered Tools to Integrate into Legal Argumentation - How to incorporate AI-powered tools into legal briefs, legal strategies, and oral arguments - Best practices for using AI-powered tools in legal argumentation - Case studies of successful tool integration Module 5: Hands-On Workshop: Integrating AI-powered Tools into Legal Argumentation - Participants will work in small groups to integrate AI-powered tools into legal briefs, legal strategies, and oral arguments - Participants will develop strategies for effectively using AI-powered tools in legal argumentation - Participants will present their arguments to the group and receive feedback Module 6: Communicating the Use of AI-powered Tools to Clients and Colleagues - How to effectively communicate the use of AI-powered tools to clients and colleagues - Addressing concerns and questions about AI-powered tools - Case studies of successful communication strategies Assessment: - Completion of a final project that demonstrates the use of AI-powered tools in a legal argumentation scenario - A survey to assess the effectiveness of the course and the adoption of AI-powered tools among participants Example Exercise: In Module 3, participants will work in small groups to analyze a real-world legal case using AI-powered tools. They will identify relevant information and patterns using the tools, and develop legal arguments based on their analysis. Each group will present their arguments to the class, and the class will provide feedback on the effectiveness of the arguments. Success Criteria: - At least 80% of participants complete the course - At least 60% of participants report using AI-powered tools for legal argumentation within 6 months of completing the course - At least 50% of participants report increased efficiency, accuracy, and/or better outcomes as a result of using AI-powered tools for legal argumentation This final plan combines the best elements of the two previous courses by providing a comprehensive understanding of AI-powered tools for legal argumentation, as well as practical experience using these tools. The hands-on workshops allow participants to gain practical experience using AI-powered tools in a collaborative setting, while the modules on integrating these tools into legal argumentation and communicating their use to clients and colleagues provide a more comprehensive understanding of how to effectively use these tools in practice. The final project and survey provide a way to measure the effectiveness of the course and the adoption of AI-powered tools among participants, ensuring that the success criteria are met.