Building Bridges with Bots: A Law Scholar’s Journey into Data Science Mentorship

When most people think of data science, they don’t imagine a lawyer leading the charge. But for one Stanford Law scholar, Takuma Iwasaki, that unexpected intersection sparked not only new research questions but also a chatbot.

From Courtrooms to Code

Before coming to Stanford, Takuma worked as a lawyer in Japan, focused on M&A, antitrust, and international trade. While traditional legal work rarely intersects with data science, his firm took a different approach: combining legal reasoning with empirical analysis. Collaborating with economists and data scientists, Takuma discovered the power of using data—not just doctrine—to construct legal arguments.

“I realized the legal field often lacks a data-driven perspective,” Takuma explained. “That experience made me want to explore the intersection of law and data science in a more academic setting.” Stanford, with its interdisciplinary ethos and expertise in data science, was a natural choice.

A Chatbot is Born

One of Takuma’s experiences with the data science community began with a challenge: how to prepare for Stanford’s Data Science Mentoring program without prior mentorship experience or an available mentee.

Takuma’s solution? Build a chatbot.

Inspired by the idea of simulating real-world mentoring interactions, Takuma created a tool to help future mentors explore common challenges in data science mentorship. “It was my first time building a chatbot, and it started just as a personal project. Then I realized other mentors were struggling with the same lack of practice, so I thought, why not share it?”

The chatbot—still in beta—offers an intuitive interface and flexibility to adapt to many scenarios, thanks to the versatility of AI. “Data science is such a broad field. Mentees have different interests, majors, and needs. That’s where a chatbot shines—it can simulate many types of conversations.”

While the tool hasn’t been widely publicized, Takuma hopes to expand access, enabling others to use the bot as a framework for building their mentoring tools tailored to specific fields or communities.

The chatbot isn’t the only way Takuma has engaged with mentorship. As part of the Data Science Mentoring Program, he worked directly with community college students interested in data science, and the experience, Takuma says, was eye-opening.

“These students come from completely different backgrounds than Stanford undergrads. Learning how they think about data science, what they hope to do with it—it broadened my perspective as an educator.”

Even after the program officially ended, Takuma has continued mentoring his mentees. “It’s been a win-win relationship.”

Discovering New Research Directions

Takuma’s time with Stanford Data Science (SDS) has also shaped his academic trajectory. Weekly research talks and informal gatherings nicknamed Data & Donuts have exposed him to fields he had never considered, including public health.

“One of my favorite moments was realizing that public health shares so much with empirical legal studies—we use similar methods, we care about similar societal impacts.” That insight led to a new research direction: exploring the intersection of law and public health.

Takuma’s favorite SDS experience, the Stanford Causal Science Conference, introduced him to new ideas, scholars, and potential collaborations. “It was inspiring to see so many disciplines come together around a shared interest in causal inference.”

Looking Ahead

With one foot in law and the other in data science, Takuma is carving out a unique space that challenges traditional boundaries. Whether it’s through scholarship, mentoring, or even AI tools, his goal is clear: to make data-driven thinking more accessible across disciplines, including in fields where data science perspectives are currently uncommon.

“Stanford Data Science has helped me grow both as a researcher and as an educator,” Takuma said. “I wouldn’t be where I am now without it.”

Useful Links

• Mentorship program overview
• Paper on data science mentorship: Near-peer mentoring in data science: Two experiences at Stanford University
• Recently-published paper - Financial Firepower: School Shootings and the Strategic Contributions of Pro-Gun PACs

Data Science Mentorship Bot "Assistant" & Getting Started Tips

• Data Science Mentorship Bot "Assistant" & Getting Started Tips
• Chatbot users act as data science mentors, beginning each session with a greeting. When a user indicates that the mentorship session is finished (e.g., by saying “end”), the bot should provide detailed feedback summarizing the overall session.
• Mentors are expected to explore the mentee’s background, aspirations, and questions to offer meaningful, personalized guidance.

Navigating Mentorship in Data Science: A Generative AI Approach

By Takuma Iwasaki

Data science is inherently interdisciplinary. It draws on theoretical insights from a variety of fields and supports a wide range of applications—from public health and environmental studies to economics and even classics. It also bridges the gap between academic research and real-world industry challenges. This rich interdisciplinarity is a major strength, but it also presents unique challenges, when it comes to mentorship. Data science mentors must navigate questions that range from highly theoretical to deeply applied; the field’s diversity often places mentors in unfamiliar situations.

My experience working on the Data Science Mentoring project for community college students, led by Professor Sabatti, highlighted just how multifaceted effective mentorship can be. The mentees came from diverse backgrounds, with different skill sets, academic interests, career aspirations, and expectations of the mentorship process. There is no one-size-fits-all model for mentorship in data science. Mentors must be equipped to support a broad spectrum of mentee needs, which often requires experience that is difficult to acquire beforehand.

This challenge led me to explore how generative AI could help. Building on my involvement with the mentoring project, I developed a chatbot designed to assist in training new mentors. The tool, available here, simulates real-world mentorship scenarios. I started by creating five fictional but representative mentee profiles, each reflecting different educational paths, experiences, and goals. I then used generative AI to analyze these profiles, identifying likely needs, common challenges, and strategies for effective mentor responses. To make the training experience more authentic, I intentionally designed the chatbot to occasionally respond in imperfect or unpredictable ways, which helps expose users to nuanced (and sometimes messy) realities of mentoring conversations.

Although the chatbot is specifically tailored for new PhD mentors working with community college students, the broader concept has great potential for scalability. It offers a flexible, accessible tool for preparing mentors across a wide array of disciplines, especially in fields characterized by diversity in background and goals. In this way, generative AI can help bridge an important gap, making mentorship more inclusive, adaptive, and capable of meeting the needs of a growing and varied talent pool.