19552 GW Mag 26_Final_singles - Flipbook - Page 28
“The idea that we are still pushing our
profession forward is exciting.”
In a pilot program, Associate Professor of Art Therapy Jordan
Potash helped clients create AI-generated therapeutic images,
like these examples of a comforting dog in a thunderstorm.
(Images used with clients permission.)
— JORDAN POTASH, ASSOCIATE PROFESSOR
OF ART THERAPY
Among them are Alexa Alice Joubin, a professor in the
CCAS English Department, who uses AI platforms to let
students imagine Romeo and Juliet in different lights—from
Victorian interpretations to Black and queer casting; and
Tara M. Sinclair, a professor in the Economics Department,
who has co-created a virtual Federal Reserve model that
features AI-generated “personas” of actual Fed members.
She’s able to simulate their monetary policy-setting meetings
with remarkable accuracy.
In the CCAS Philosophy Department, Professorial Lecturer
Megan Davis trained an AI chatbot to engage students in
Socratic method dialogues; while in the Department of
Organizational Sciences and Communication, Adjunct
Professor Joseph McDonald plans AI student activities that
mimic workplace scenarios—like presentations, public
speaking and team management exercises. “I am on fire
with how AI resources are changing the way I teach and
present my class concepts to our students,” McDonald said.
Creating a “Digital Sandbox”
Beyond her own classroom exercises, Joubin, a
self-proclaimed AI “early adopter,” has also sparked what
she calls a “grass-roots revolution” with a new platform
called Teach Anything. Developed with former computer
science graduate student Akhilesh Ghanshyambhai
Rangani, MA ’26, Teach Anything allows academics to
design and create their own AI applications—customized
chatbots tailored to their specific classroom needs.
Teach Anything “lets professors turn their own course
materials into an AI teaching assistant for their class,” said
Rangani, now a founding engineer at Tambo AI. “Instead of
a generic chatbot that answers everything, the assistant is
limited to what the professor provides. Students can ask
questions, and the AI responds based only on the course
content, not outside sources.”
Since launching Teach Anything in 2026, Joubin has already
shared the free resource with GW faculty members—including
Davis and McDonald—and more than 250 professors worldwide.
28 | www.columbian.gwu.edu
Joubin’s “approach to using AI tools is transformational,”
McDonald said. “She is opening up avenues that inspire
students to think in astoundingly creative ways.”
Joubin said the open-accessed Teach Anything avoids
the data privacy and image sourcing pitfalls surrounding
commercial products—a concern for scholars like Potash
who worry about client confidentiality. Joubin stressed that
Teach Anything employs only open-sourced large language
models (LLMs). It is prohibited from being used for profit or
commercial services. Students are not required to have
accounts and the platform doesn’t store data.
Joubin thinks of Teach Anything as a digital sandbox that is
“built by educators for educators,” enabling AI-hesitant faculty
to get their hands dirty. “It is like a playground,” she said. “We
are giving [professors] a controlled space to experiment.”
From Faux Fed to
Interior Architecture
Sinclair’s faux Fed framework constructed models of each
Federal Reserve member from an archive of information:
speeches, voting records, biographies, policy positions—
even their communication styles. Grounded in real-time
data on their policy deliberations, disagreements and
decisions, the virtual laboratory can mirror the drama inside
the Fed’s D.C. headquarters but within a computer database.
“There’s no way I can get Chair [Jerome] Powell to come over
to GW for a day and let me run him through different scenario
exercises,” said Sinclair, who chairs the CCAS Economics
Department. “But we can create personas with flexible
personalities that give us a close simulation of the real person.”
Prior to a pivotal July 2025 meeting on whether to lower
interest rates, Sinclair adjusted the model’s recipe, adding
