By Robert Baird, CITL Senior Associate Director

Vishal Sachdev, Clinical Associate Professor of Business Administration at Gies College of Business, spent last summer and this fall at the University of Illinois teaching his students with the help of course “bots,” generative AI chatbots that help students with course-specific questions and learning objectives, from a syllabus-explainer Q&A bot to a one-on-one tutor bot, to a debate bot, to a discussion bot partner “with infinite memory.”

Long an innovator in designing technology-enhanced learning environments for students, Vishal helped pioneer (with Dr. Aric Rindfleisch) the world’s first 3D printing MakerLab within a business school. Vishal’s work with course bots was supported by a Provost’s Initiative on Teaching Advancement grant (PITA) which he and Assistant Professor Unnati Narang used to help explore whether AI helper bots worked better as one-on-one tutors for individual students or as course helpers within online group discussions. Titling his grant “Working With AI Instead of Hiding From It,” Vishal utilized a course he frequently teaches: IT for Networked Organizations (BADM 350). Teaching the course last summer with a smaller enrollment of 30 students, Vishal was able to experiment with teaching bots and advance his AI skills and to work with a developer to create bots that helped with specific course learning objectives and targeted goals.

Vishal got started with AI when he realized “We have to figure out what to do with this because students were probably the earliest adopters. So they figured out that maybe this is a shortcut. So, you've got to figure out how to not use it as a shortcut . . . And just as we allow Google as a tool or . . . calculators as a tool, you're probably going to have to allow this as a tool as well. So, the only way to figure it out is to try it. So that's where I started.” Vishal was inspired by educator Ethan Mollick, who categorizes bots into seven teaching objectives: mentor, tutor, coach, teammate, student, simulator, and tool (see “Assigning AI: Seven Ways of Using AI in Class”).

Q&A Bot

The first and simplest bot Vishal started with was a question and answer bot, which works by providing the bot with specific texts that it can then use to respond to users questions in natural language, simplifying the typical strategy where a user must search an index or scan a frequently-asked-questions page: “So we did the course syllabus and course outlines, policies and all that and that was sort of the warm up exercise. So, students get to talk to a bot rather than a TA for Q&A . . . If you can't figure it out, ask the bot and see how it does. So, you know [we had] some good outcomes from that . . . I just said, you know, talk to the bot for all the policies and stuff.”

One of the goals of the pilot was to identify ways that AI could scale up and help instructors and programs provide more frequent and higher-quality feedback and support to courses with

hundreds of students. With course bots helping with common, basic questions 24x7 and guiding students through self-help, the hope was that instructors could jump in and help students with more complex challenges.

Student Engagement Bot

One example of a good teaching practice Vishal frequently uses is an “8 Nouns” ice breaker, where students select eight nouns that best describe themselves and explain their choices, a type of activity that is known for enhancing teacher-to-student and student-to-student engagement. The challenge, Vishal noted, is that “If you’ve got 30 people or 60 people or 600 like we have for the iMBA, and you want to have an ice breaker and you want to make sense of it you can’t.” Beyond managing the activity, Vishal had some ideas about how to improve it, wondering if a course bot could help connect student responses with course content and career options. The bot developed for the 8 Nouns activity summarized and linked each student’s response to chapters and discussion topics of the course as well as various career options. For Vishal, “language models can assimilate all that language and key words, find patterns and create a summary. . . and now it’s possible to create a customized motivational message relevant to their [students] passions, music, football, family, whatever, and relate it to one particular career option which is inspired by the particular course you are in.”

Debate Bot

Another highly specialized bot for Vishal’s class was “the bot as a debate partner. So, again, you give it a set of instructions . . . and this was a debate about Does IT matter as a source of competitive advantage?” Students are given a position on a topic and then “assigned to a bot that takes the opposite position.” Students are encouraged “to fine-tune your argument with the bot and the bot will argue against your position and see if you can argue for your position.” Each student in the class was then able to “get some more ideas of where your argument is weak or strong and then use that output to debate with a human,” their student peers. Watch this short video where Vishal discusses creating the Debate Bot.

Discussion Bot

For many years Vishal’s online course discussion forums have been creatively designed to fully integrate course learning goals and topics with state-of-the-art online discussion features and applications, such as YellowDig, an online discussion tool designed to enhance student conversations and collaborations. In another course on database design, Vishal ran an experiment to see if bots could help raise student’s “novice” awareness of a topic to uncover elements with which they were unaware and unfamiliar, and whether the bot was most effective in one-on-one or group discussion contexts. For Vishal, “there is reasonable evidence to suggest that novices in a particular domain don’t know what questions to ask. . . If you put a novice in front of a bot and the questions are still going to be bad because they don’t know what they don’t know and, thus, don’t know what to ask.” Could a bot serve, then, as a guide for novice students and help deepen and elevate students’ online discussions. An experiment to test this hypothesis is underway to see if a group of students working with a bot, work better to overcome the I-don’t-know-what-to-ask problem vs an individual working with a bot.

Getting Started With AI

For instructors just beginning to consider generative AI, Vishal believes “the most successful activity . . . is probably using it as a brainstorming partner. So, if you have activities where you can encourage students to brainstorm with the bot and then synthesize, I think that's a sure winner. It'll work every time. So that's sort of a beginner level activity that everybody can try.”

Evolving Teacher Role in Age of AI

Regarding the challenges of generative AI to education, Vishal feels that “the biggest challenge is not plagiarism or academic integrity . . . I think the bigger challenge right now for me is, which I don't have an answer for, is that if you assume an infinite memory knowledge source, which can solve almost all problems on your phone, what should you be teaching students? So that's the challenge I'm dealing with.” Referencing the problem that novice learners do not know enough to know what they don’t know, Vishal hypothesizes that teachers should be spending more time teaching students how to identify and ask better questions: “the challenge still remains that in order for people to traverse along that path [successful learning], they still need to ask questions. And if they don't ask good questions, they don't get good answers.”

Sidebars & Resources

“Advancing the Learning Teaching Experience With AI.” James Gray podcast interview with Vishal Sachdev. 

“A sage, a guide, or a curator — Pioneering AI Use in the Business Classroom,” Oct. 2, 2023. University of Illinois, Gies College. 

“Assigning AI: Seven Ways of Using AI in Class.” June 12, 2023. Ethan Mollick.