My colleague, Renee Henson, wrote an important new article, My Negotiation Chatbot Toby and Pedagogical Governance, about a chatbot that she developed for her first-year Lawyering: Problem-Solving and Dispute Resolution course at the University of Missouri Law School.
Her article provides a thoughtful analysis of how faculty can use AI to improve legal education. It also illustrates something broader: AI can be used in a “centaur” mode in which humans and AI contribute complementary capabilities. Working together, they can produce better results than either could produce independently while humans retain judgment and responsibility.
Meet Toby
Renee created Toby to give students practice negotiating with a difficult counterpart lawyer. That is a familiar challenge in practice but hard to reproduce in a law school classroom. Students in dispute resolution courses tend to be too darn cooperative. So Renee trained Toby to be intentionally obstinate. (Dwight Golann developed a simulation bot for the same reason.) Simulations with Toby require students to apply negotiation concepts and force students to respond in real time.
Renee compares herself to an athletic trainer where Toby is the equipment that students use to practice. The equipment provides opportunities for repeated practice, but it does not decide what students should learn or whether they are learning it well. She designs the exercise, sets the boundaries, evaluates what happens, corrects problems, and changes the design when appropriate.
Sometimes Toby gets things wrong. In one simulation, for example, it proposed an unrealistic solution. Renee used the mistake to explain why the response was problematic and to discuss practical implications.
That episode reflects an important feature of her approach. Because AI tools regularly make mistakes, it’s important for students to recognize their limitations and exercise judgment about using their outputs.
Students Use AI in Centaur Mode
When students negotiate with Toby, they must interpret what it says, decide what is important, choose how to respond, apply the concepts they have learned, and adjust as the interaction proceeds. Toby contributes speed, responsiveness, unpredictability, and endless willingness to play the role. The students practice legal and negotiation analysis, strategy, judgment, and decision-making. This is an excellent illustration of how students can use AI in centaur mode.
That distinction matters in debates about whether AI helps or harms students’ development of analytical skills. Much of the discussion about AI focuses on whether someone “uses AI” or not. But AI use involves many different activities. A student who asks AI to do the intellectual work and submits the result with little thought is doing something quite different from a student who uses AI to create a demanding environment in which the student must do more thinking.
Renee makes this distinction concrete in other parts of her course as well. She permits students to use AI to help identify parties’ interests, rights, and power dynamics, while requiring them to develop their own mediation plans and reflective analyses. The pedagogical question therefore becomes more precise: Which parts of the work can AI usefully assist with, and which parts should students perform themselves to develop the capabilities the course is intended to teach?
Renee Uses AI in Centaur Mode
Renee’s own work with Toby is another example of using AI in centaur mode. She did not simply decide to “use AI in class.” She identified a pedagogical problem, selected a tool, defined what she wanted it to do and not do, refined the instructions, tested the bot with difficult questions, instructed students how to use it, observed the results, and continued revising it.
She developed a six-step framework for using AI in education:
- identify a pedagogical gap
- select a platform
- define objectives and limits
- refine the prompt
- test the tool, including unusual or difficult situations
- deploy it with instruction about its limitations and proper use
This is a useful model for faculty who want their students to benefit from learning with AI – while also learning how to use AI. Using AI effectively requires experimentation, feedback, correction, practice, and continuing human judgment.
There is an interesting parallel. Renee uses AI to create an educational experience, and students use AI to learn through their own interactions with it. In both contexts, humans actively manage their interactions with AI. Renee manages the learning environment, and students make their own negotiation decisions. AI expands what Renee and her students can do, while providing opportunities for them to exercise their own judgment.
A Useful Way to Think About AI in Legal Education
Renee takes the risks of using AI in her course seriously. Her article discusses errors, overreliance, misuse, bias, unequal access, and concerns that AI could weaken the development of critical thinking. She advocates responsible pedagogical governance to address these risks. She argues that faculty should plan AI use to achieve educational objectives, establish limits, supervise it, teach students about the risks, and revise the approach as experience reveals problems.
This provides a useful way to move beyond overly simplistic arguments about whether AI is good or bad for legal education. Faculty and administrators should ask more specific questions, including:
- What are humans trying to accomplish?
- What do humans contribute?
- What does AI contribute?
- What does the combination make possible that would be difficult for either to accomplish as well alone?
- Who exercises judgment and responsibility?
Renee’s development and use of Toby provide good answers to those questions. She supplies the pedagogical purposes, course knowledge, design, supervision, correction, and judgment. Toby supplies a readily available, responsive, and sometimes maddening negotiating counterpart. Students supply the thinking and decision-making required to deal with it. The combination of these contributions creates real educational value.
Renee has used Toby with students for several years, and she reports both benefits and problems. Her experience suggests that the issue for legal educators is not simply whether students will use AI. It’s how we can use it to help students learn to exercise good judgment.
