What is UW Tacoma doing with AI?

 AI@UW’s inaugural SEED-AI grants will support projects involving UW Tacoma faculty, including efforts to use AI tutoring in math, engineering and computer science courses.

By  Syed Huzaifa Bin Afzal

University of Washington Tacoma researchers are among the recipients of AI@UW’s inaugural SEED-AI grants, a new funding program supporting faculty-led projects that apply artificial intelligence to teaching and learning. 

AI@UW announced 36 funded projects across the UW tri-campus system. According to UW Tacoma News & Information, eight projects involve UW Tacoma researchers, with each funded project receiving up to $50,000. 

The Tacoma-related projects range from AI literacy and teacher preparation to mathematics, engineering, computer science, health care, statistics and faculty workflow. The projects include “Building AI Literacy for UW Tacoma Undergraduate Students,” “Scaling Mastery: An AI-Driven Adaptive Bridge System for Algebra Preparedness in TMATH 109,” “TELL-AI: Collaborating with AI in Teaching English Language Learner Endorsement Program” and “Guided inquiry with AI: Implementing cognitive engagement techniques in engineering education.” 

This article focuses on two classroom-centered projects: Rita Than’s “Scaling Mastery” project in TMATH 109 and the “Guided inquiry with AI” project involving Chris Marriott. A follow-up article will cover projects led by Menaka Abraham and Belinda Y. Louie. 

Associate Teaching Professor Rita Than’s project focuses on TMATH 109, a foundational math course at UW Tacoma. In written responses to The Ledger, Than described TMATH 109 as an interdisciplinary course that connects algebra concepts with practical topics such as personal finance. Than, who designed the curriculum and has taught the course since spring 2021, is working on the project with student researcher Jasmine Sellers, a computer science and electrical engineering student at UW Tacoma. 

The project is designed to give students a more personalized path toward math mastery. Rather than treating a quiz as a final grade, the system would identify where a student struggled and generate practice tied to those errors before the student attempts a retake. 

Than wrote that the project uses a “Targeted Mastery” approach. Under that model, an AI Scenario Generator creates practice problems based on the student’s mistakes, major and interests. After completing the practice and reviewing the related concepts, students can unlock the chance to retake that part of the quiz and improve their score. 

The project also includes a Socratic AI math tutor, which Than described as a 24/7 personal coach that can identify strengths and weaknesses, recognize mistakes and help students decide what to study for exams. 

“Introductory math courses are often ‘gatekeepers’ where a few early mistakes can cause students to fall behind permanently in the course and in making progress on their program,” Than wrote. 

Than identified accumulated knowledge gaps, false confidence after a strong start and the workload of providing individualized support to large classes as some of the challenges the project is meant to address. 

 “Personalized support can help, but it is difficult for one instructor to provide manually to more than 60 students without AI automation,” Than wrote. 

The AI support is not meant to replace student work, according to Than. She described the Socratic math tutor as a tutoring scaffold rather than a shortcut or homework solver. The chatbot is programmed to withhold direct answers and guide students toward finding logic errors in their own work. 

Than also noted that exams and retakes would still be in-person, proctored and human-graded. The project aims to increase pass rates to more than 90%, and Than wrote that initial research showed students who engaged with the retake process performed significantly better on the final exam. 

To reduce risks such as incorrect AI-generated math advice, the project uses a human-in-the-loop approach. Instructors review the AI’s work before deployment, and Python code is used to check problem types and calculations, according to Than. 

Another funded project, “Guided inquiry with AI,” involves Assistant Professor Lorne Arnold, Teaching Professor Chris Marriott and Assistant Professor Cassandra Donatelli from the School of Engineering & Technology. The project will develop an AI teaching assistant for civil and mechanical engineering and computer science courses, according to the UW Tacoma project summary. 

Marriott described the project as an effort to engineer a chatbot that can answer student questions about course material while keeping student learning at the center. One of the project’s main goals is to distinguish between productive struggle and unnecessary struggle. 

“A core goal is to differentiate between material that students benefit from struggling with on their own from material where struggle is not generative,” Marriott wrote. 

“Students can already upload assignments into public large language models and ask for help,” Marriott wrote. “The team’s goal is different: to build course-specific context and instructional guidance into the tool so it reflects the instructor’s intentions instead of simply producing a solution.” 

Marriott wrote that the agent would use the Socratic method to help students build their own problem-solving skills. In practice, that means asking students what they have tried, what worked and where they think the error is, rather than immediately showing the answer. 

The tool is expected to be used primarily in STEM courses, though Marriott also plans to use it in a writing course on ethics. He described the project as an adaptive agent that can adjust to different instructors and teaching styles, rather than a static chatbot placed on a website. 

Marriott also addressed concerns about student work. He wrote that the goal is not to displace student tutors, but to respond to a classroom reality in which human tutors have become underused as large language models have become another option for students. An instructor-adaptive tool, he wrote, may offer better guidance than a general standalone AI tool. 

Marriott placed the project within a broader shift in computer science and engineering. AI-assisted tools, from autocomplete to coding assistants, have been part of a longer technological progression in the field, but newer systems may change how programming work is understood. 

“This is both an opportunity and concern,” Marriott wrote. “Our fields become tremendously more efficient, but we are very uncertain how the fields will restructure themselves around the new tech.” 

Together, the two projects show how UW Tacoma faculty are testing AI as a support tool while trying to preserve the parts of learning that require practice, reflection and problem-solving. The Ledger will continue coverage of UW Tacoma’s SEED-AI projects in a follow-up article focused on undergraduate AI literacy and AI use in teacher preparation for multilingual classrooms.