The Center for Data Innovation spoke with Renee Henson, visiting assistant professor of law at the University of Missouri School of Law, who created a custom chatbot called Toby to help law students improve their argumentation skills and prepare for real-world scenarios involving complex negotiations. Hensen discussed how she programmed Toby to be aggressive and challenging and how AI can boost productivity in the legal profession.
Martin Makaryan: What inspired you to create a chatbot specifically tailored for law students?
Renee Henson: At the University of Missouri School of Law, I teach a course called “Lawyering: Problem-Solving and Dispute Resolution” that aims to equip first-year law students with important practical skills they need to be successful when practicing law. My background as a business and commercial litigator, particularly in products liability and intellectual property, exposed me to the challenges of dealing with difficult opposing counsel. Understanding the difficulties of dealing with opposing counsel who may use various tactics to intimidate or push their agenda during complex negotiations led me to explore using AI to simulate such interactions in my lawyering class. I wanted students to engage in a realistic negotiation exercise, encountering the kind of adversarial personality they’re likely to meet in practice. That’s how the idea for a chatbot came about.
Makaryan: How did you create Toby and customize it to be aggressive and adversarial?
Henson: I built a customized application powered by OpenAI’s GPT model, which I named Toby after the character that everyone disliked in the popular show The Office and customized it to embody an adversarial lawyer by using tailored prompts and uploading data related to the course. I programmed Toby to prioritize his client’s interests and his own ego, behaving as an obstinate, high-pressure negotiator. To make it realistic, I uploaded my class materials to train Toby in line with the tactics I teach, like identifying opening offers and bottom lines during a negotiation to settle the case. I also conducted a few rounds myself to ensure Toby stayed in character and adhered to the negotiation structure before deploying it in class. It took some fine-tuning to get Toby to respond consistently as a seasoned lawyer rather than breaking into generic responses or breaking character.
Makaryan: Can you describe the classroom setup for using Toby?
Henson: We devoted almost an entire class session to a negotiation exercise with Toby. I divided the students into four groups representing the plaintiff in a hypothetical case, while Toby represented the defendant. Each group had confidential case information, much like in a real-life negotiation, which drew their red lines and summarized the objectives in the negotiation. Similarly, I uploaded confidential information for Toby in the same case to help the chatbot engage in the kind of adversarial exchange we would expect in a real negotiation setting. I displayed Toby on a large screen for all to see in class, and the students engaged in a back-and-forth to negotiate settlement. Each group presented their offers, which Toby responded to in real-time. The goal was to reach a realistic settlement by balancing client interests and handling Toby’s challenging responses. To my surprise, despite Toby’s consistent and realistic responses and hard negotiating posture, in the end, the students were able to reach a compromise and settle the case.
Makaryan: What impact did this AI-powered exercise have on learning outcomes for students?
Henson: I received an overwhelmingly positive response about using Toby in class. This exercise enabled the students to think on their feet, a skill difficult to replicate outside real-world situations. The students appreciated the realistic complexity Toby brought to the table, especially the emotional and tactical aspects of negotiation, which are sometimes the hardest to navigate in real-life negotiations. Afterwards, students frequently referenced Toby in later classes, drawing parallels between the exercise and real-world negotiation strategies. They found it beneficial not only in understanding negotiation techniques but also in familiarizing themselves with how AI technology can assist them in legal practice.
Makaryan: How can AI boost learning outcomes in law school?
Hensen: AI has a vast potential to make the legal practice more efficient, especially by streamlining or automating tasks like document review, drafting, and research. For law schools, I think AI can serve as a supplementary tool to aid learning. At the same time, it’s crucial to ensure that students use it to enhance their skills rather than to perform the analytical work for them, in which case they would be unable to develop their own analytical skills essential to perform the duties of a lawyer in the future. As future lawyers, they need to understand AI’s capabilities and limitations, making informed choices about its use.
