As part of the project, students will build a conversational AI system that helps students plan to fulfill the requirements of the new data science certificate in SAS. https://datascience.sas.rutgers.edu/
The rules for this certificate are a little complicated and there are many ways for students to find courses that match their interests. So we are expecting that it will be useful for students to get help. The basic idea of a conversational system is that it can answer questions about the program and manage the creation of a study plan that fulfills the program requirements. (Note that most of the programming and analysis for this project is actually domain independent, meaning that it could be used for a wide range of information and planning tasks, and that makes this exact application somewhat flexible.) You’ll be able to start from infrastructure that my graduate students have created and that we will be extending over the summer, and there will be opportunities to interact with other faculty as well as me who have interests in this kind of technology.
The focus of IRTs is collaborative problem solving. You’ll learn how to do it yourselves in the project, and in the course of this project you’ll also learn how AI researchers have taken insights from collaborative problem solving studies and used them to design interactive computer systems with some of the same capabilities. I have a fun set of readings to make this possible, including papers from the philosophy of mind on teamwork, papers from linguistics on cooperative conversation, papers from computer science on cooperative planning and interaction, and even some work from the management literature on effective business teams. For CS students, the work of the project involves creating software tools as well as implementing a collaborative planner. A collaborative planner looks at the state of problem solving, which might be a partial plan or a set of partial plans, and makes suggestions about conversational steps that could help the user extend the current plan or decide between those plans. Programming this involves figuring out how to represent plans with appropriate data structures, and how to recognize that an action can help flesh out a plan using a generic rule 4 base. The tools involved enable students without CS background to contribute to the project.
For example, a tool based on logic might let a philosophy student write down precisely what the requirements of a plan are, which the planner can then use automatically to figure out how to meet those requirements. Or a tool can implement some ideas of discourse theory to let a linguistics student write down how to vary the system’s language so that it better fits the context. Everybody on the team will collaborate to collect and look at sample conversations which set the scope of the project and allow you to make sure that you have meaningful coverage of useful patterns in conversation, but this aspect of the project is an opportunity for students with interests in human-computer interaction, cognitive science and communication to take the lead.