Student-developed AI chatbot opens Yale philosopher’s works to all

The public is often closed off from scholarly perspectives on the potential benefits of generative artificial intelligence (AI). Studies often reside behind pricey paywalls. And even if they are accessible, they are frequently written in esoteric language that non-academics struggle to parse.

Nicolas Gertler, a first-year student in Yale College, saw a potential solution to these obstacles, through generative AI’s own capabilities.

Gertler, a research assistant at Yale’s Digital Ethics Center (DEC), has spearheaded an experiment using generative AI to make bodies of knowledge broadly accessible. With Rithvik “Ricky” Sabnekar, a high school junior and skilled developer from Texas, he created the Luciano Floridi Bot, also known as LuFlot, a free AI-powered online educational tool designed to foster engagement with the works of Yale philosopher and DEC Director Luciano Floridi, a pioneer in the philosophy of information and one of the most-cited living philosophers.

The developers believe it’s the first time a chatbot has been trained on an academic’s corpus of literature and released to the public for free.

“The idea was to democratize access to Professor Floridi’s work,” Gertler said, who undertook the project after discussing its possibilities with Floridi. “The issues he has written about touch everyone’s lives, and more and more people are becoming aware of AI. LuFlot provides an AI-driven platform for much broader engagement with the ethical questions surrounding this transformative technology.”

Meant to facilitate teaching and learning, the chatbot is trained on all the books that Floridi has published over his more than 30-year academic career. Within seconds of receiving a query, it provides users detailed and easily digestible answers drawn from this vast work.

It’s able to synthesize information from multiple sources, finding links between works that even Floridi might not have considered.

The tool features in-text citations, allowing users to trace the origins of the information provided directly to the original texts. If asked a question outside its knowledge base, the bot will politely respond that the query falls outside the scope of Floridi’s expertise. But relevant questions receive prompt and thorough answers. For example, a question asking how to use AI ethically quickly generated a clear, eight-point response with sources cited.

The interface also allows users to ask follow-up questions.

“Anyone, regardless of their knowledge of AI, can visit the website, ask a question, and have a conversation with the founder of the philosophy of information,” Gertler said. “I think that’s remarkable.”

The tool warns users to evaluate its answers critically, as it may generate incorrect or biased information.

Floridi, professor in the practice in the Cognitive Science Program in Yale’s Faculty of Arts and Sciences, is impressed by the chatbot and its young developers.

“The center is focused on the impact of digital technology and the ethics of AI, so it makes a lot of sense to have a bot available to answer people’s questions,” Floridi said. “The bot is an amazing tool. Nicolas and Ricky deserve all the credit. I’m just the supporting band.”

He has also become a user. For background on a paper he is writing, Floridi asked the chatbot about asymmetry between good and evil.

“It gave me an amazing answer, accurately referencing concepts and ideas I had completely forgotten that I’d written about,” he said. “It can instantly draw interesting connections between something I published last year and something I published in 1991, which is incredible.”

Dynamic knowledge

Gertler, who is from Los Angeles, became interested in AI’s potential to benefit society about five years ago after watching YouTube videos on the topic.

“I love thinking about the societal and ethical implications of new technologies and how they can be leveraged to help people, especially those in underserved and marginalized communities,” he said.

He is involved with two youth-led organizations devoted to harnessing new technology for the greater good: He serves as AI and education advisor of Encode Justice, an organization of more than 1,000 high school and college students across the world that advocates for steering AI in directions that benefit society. He also is vice president of Fidutam, a civil-society group that mobilizes its more than 1,600 members to advocate for and build responsible technology.

At the DEC, Gertler explores how to use generative AI to create educational frameworks that can provide enriched learning experiences. He also serves as the first-ever “AI student ambassador” at Yale’s Poorvu Center for Teaching and Learning, engaging with faculty on integrating AI into instruction and developing resources for students and faculty on the ethical use of AI in pedagogy.

The LuFlot project came together fast. Earlier this year, Gertler showed Floridi a chatbot he had created in his cognitive science class, intended to help students learn the course material. After a few conversations, the idea for LuFlot was born.

Gertler and Floridi decided to build from scratch a custom chatbot with its own user interface, rather than rely on existing tools like ChatGPT. Gertler brought on board Sabnekar, who attends Plano East Senior High School in Plano, Texas, to help develop the bot. The two met through Fidutam, where Sabnekar leads the technical development of the organization’s projects. 

“Ricky is technically brilliant,” Gertler said.

The pair founded Mylon Education, a startup company seeking to transform the educational landscape by reconstructing the systems through which individuals generate and develop their ideas. LuFlot is the startup’s first project. 

“Our goal is finding ways to harmonize human agency and creativity with AI-enabled structural support,” Gertler said. “That involves incorporating AI tools into the writing process without inhibiting people’s agency. It’s not about using the chatbot to write essays. It’s about using this technology to deepen your knowledge and sharpen your creativity and critical thinking skills.”

Generative AI and other innovations are changing how people learn, Floridi said. So there needs to be new approaches to teaching.

LuFlot demonstrates that it is cheap, feasible, and efficient to train an AI chatbot on a scholar’s corpus and have it produce high-quality answers to users’ prompts, he said. Chatbots could be trained on other scholars’ work or on an instructors’ entire course content, which could help students learn and retain information. 

“It’s much more useful than just making my lecture notes available online,” he said.

Floridi says he is reminded of Plato’s criticism of the invention of writing — that it is not a dynamic means of sharing knowledge because reading it always conveys the same answer.

“The bot is dynamic,” he said. “It will not give you the same answer if you ask the same question even slightly differently. And you can ask it to be more specific. And it will grow as you train it with more information.”

Visit this link to converse with LuFlot about the ethics of digital technologies.