1 of 22

SUNY CIT 2026 · STONY BROOK UNIVERSITY

How to Empower Learners

with Human-Centered AI

Junghyun Ahn · Science and Math, FIT

Kyunghee Pyun · History of Art, FIT

2 of 22

FRAMING

What is Human-Centered AI?

IBM Research, 2022

AI that "amplifies and augments rather than displaces human abilities." The principle: human + AI beats either alone.

Stanford / Fei-Fei Li, 2025

AI should "understand, serve, and empower people rather than replace them."

For schools, 2025

Augmentation over automation · literacy over policy · design over technology · vision over decisions.

The core inversion

Traditional AI asks "what can the system do?" Human-centered AI asks "what does the person need — and how does the system serve that?" The user is not a data point. The user is the point.

Sources: IBM Research; Stanford HAI (Fei-Fei Li); human-centered AI school frameworks, 2025.

3 of 22

FRAMING

Three Principles of Human-Centered AI

Augmentation, not replacement

AI extends human reach — more time with students, deeper research, faster prototyping. It does not substitute for human presence, expertise, or judgment.

Transparency & contestability

Users must understand what the AI is doing, why it produced an output, and be able to push back. Black-box outputs disempower learners.

Ethical grounding & equity

Systems must respect privacy, deliver fair outcomes across populations, and account for whose data trained them. Equity is a design choice, not an afterthought.

4 of 22

WHY THIS MATTERS FOR STUDENTS

AI and the Future of Our Students

World Economic Forum · Future of Jobs Report 2025

170M

new jobs by 2030

92M

jobs displaced

39%

of core skills change

The #1 skill employers now want: analytical & critical thinking.

70% call it essential. The question for our students isn’t "will AI take my job?" — it’s "can I do the part of the job AI can’t?" That is exactly what human-centered teaching builds: judgment, creativity, and ethics.

Source: World Economic Forum, Future of Jobs Report 2025.

5 of 22

WHY THIS MATTERS FOR STUDENTS

What Are the New Jobs, Actually?

Concrete roles emerging in 2025–26 — across technical, ethical, and creative fields.

TECHNICAL & OPERATIONAL

• Prompt Engineer

• AI Engineer / ML Specialist

• AI Integration Specialist

• Forward-Deployed Engineer

• Chief AI Officer (CAIO)

ETHICS, POLICY & OVERSIGHT

• AI Ethicist

• AI Privacy Officer

• AI Policy & Compliance Specialist

• AI Auditor (bias & fairness)

• AI Risk Manager

CREATIVE & HUMANITIES-ADJACENT

• AI Creative Technologist

• AI-Powered Content Curator

• Digital Archivist (AI tools)

• Museum AI Research Scholar

• AI UX Designer

→ The Metro Politan Museum of Art (2026–27): Schmidt Sciences Research Scholars in AI & the Humanities — humanities scholars hired to investigate AI applications in museum information, conservation, and material culture.�The non-STEM AI career is already here.

Sources: WEF Future of Jobs 2025; LinkedIn 2024 Workforce Report; The Met / NYFA, March 2026.

[What Are the New Jobs] Make the abstract concrete. This is the slide that quiets the room.

170 million new jobs sounds great in the abstract. But what are they, actually? Three categories.

Technical roles — prompt engineer, AI engineer, integration specialist, forward-deployed engineer, chief AI officer. LinkedIn data shows over 140 percent growth in AI engineer postings.

Ethics, policy, and oversight roles — AI ethicist, privacy officer, compliance specialist, bias auditor, risk manager. Every company adopting AI now needs these.

And the category most relevant to our students: creative and humanities-adjacent roles — AI creative technologist, content curator, digital archivist, museum AI scholar, AI UX designer. The Met just launched a Schmidt Sciences program in 2026 hiring humanities scholars to investigate AI in museum work. The non-STEM AI career isn't coming. It's already here.

TIP: "When I show this to my art-history students, they realize there are art-history jobs in AI right now, at the Met, today. That's the moment empowerment becomes practical."

6 of 22

OUR APPROACH

PI

Why "Empower"? The Human Need

Students felt helpless

After the pandemic, many students felt discouraged and found it hard even to attend class. We wanted them to regain autonomy and control.

A motivated community

A positive emotional response is needed to build a learning community of motivated students.

Faculty felt vulnerable too

Teaching, research, and advisement reduced to numbers; experts treated like customer-service reps; students positioned as consumers.

Human agents disappearing

In the machine-learning narrative, human interaction risks becoming obsolete — exactly what we set out to resist.

7 of 22

OUR APPROACH

PI

Why "Empower"? Defined Outcomes

KEY EMPOWERMENT ENHANCEMENTS

Active Leadership. Students as active co-creators of AI ethical principles — not just finding the conversation "meaningful."

Agency & Skill. Equipping students with the agency to navigate industry changes.

Student-Led Practices. "Student-led, AI-assisted educational practices" — emphasizing their autonomy.

KEY ACTIVE-LEARNING FOCUSES

Dynamic Lab. An active, collaborative space for building ethical principles — not a passive discussion.

Live Case Studies. Industry changes as immediate problem-solving opportunities to analyze and dissect.

Field-Tested Outcomes. Final publications stem from applied, hands-on student practice and active research.

8 of 22

THE PROJECT

PI

Process

A year-long collaboration

Faculty and staff met weekly in summer and winter, monthly during semesters. Two goals: develop each researcher’s individual project, and prepare the FIT symposium.

Two grants in succession

A 2025 proposal got a SUNY CID grant; SUNY IITG then let us expand scope and bring more faculty into the seminar group.

What we built together

Case studies on AI-assisted collections management. Students curated and analyzed images, art objects, and archival documents. NirVana Rizvi served as research assistant across all summer meetings.

9 of 22

THE PROJECT

The Two Grants: CID and IITG

SUNY CID · 2025

Curriculum and Instructional Development

Seed grant for "Human-Centered AI: Knowledge, Authority, and Creativity." A small-team exploration of how non-STEM faculty could engage AI critically — the entry point for asking "how do we teach this?"

SUNY IITG · 2025–26

Innovative Instructional Technology Grant

Funded the multi-institutional expansion: FIT, Old Westbury, Buffalo, U. of Memphis, Monroe CC. Goal: develop case studies in AI-assisted collections management for arts and cultural organizations, plus a capstone symposium.

Why human-centered AI specifically? Most AI funding goes to technical capability. CID and IITG let us fund the human side — faculty development, student literacy, ethical case studies, and cross-disciplinary conversation. The grants made human-centered AI a teachable framework, not just a research term.

10 of 22

THE PROJECT

PI

Our Method & Mission

METHOD

Rather than celebrate capability, our research highlighted

AI’s vulnerability on issues of ethics and societal values.

MISSION

Across FIT, Old Westbury, Buffalo, and Monroe CC: empower non-STEM students and instructors through AI literacy — so they can critically engage AI-assisted research and assess its impact on sensitive messages about race, poverty, discrimination, and disability.

Recurring question: how does AI-driven analysis adapt to

social justice, decolonization, and the ethical values of diverse learning communities?

11 of 22

CASE STUDY

PI

The Schomburg Center Study

Kyunghee Pyun (FIT) with Prof. Eunseo Choi’s team (University of Memphis)

AI systems processed text- and image-based datasets using deep-learning algorithms (PyTorch) to analyze patterns in visual and textual data.

What: early-20th-century photographs of African American students learning trade skills — testing how AI reads historically sensitive images of race, education, and the U.S. workforce.�How: a custom computational tool, then commercial platforms (Gemini, ChatGPT) in summer 2025, revisited Jan–Feb 2026.�Finding: in just 5–6 months the commercial AI improved dramatically — which amazed the team.

Photographs: The New York Public Library, Schomburg Center for Research in Black Culture, Digital Collections.

[Schomburg Study / PI] Let the photographs do work. Pause and let people look.

Concrete example of that mission. Prof. Pyun's case study with Prof. Eunseo Choi's team at the University of Memphis, using photographs from the Schomburg Center for Research in Black Culture — the NYPL's major archive of Black history up in Harlem.

What you're looking at: early-20th-century photographs of African American students learning trade skills. A classroom. Young women presenting a "Conserving Materials" project. A children's rhythm band — see "Lillie Spiller's Band" on the drum.

The team tested how AI reads these historically sensitive images. Custom computational tool first, then ChatGPT and Gemini in summer 2025. The finding — on the next slide — surprised everyone.

TIP: Pause 2–3 seconds on the photographs before talking. This is your most visually arresting slide.

12 of 22

CASE STUDY · WHAT WE FOUND

PI

Schomburg: What the Team Found

The research question was about bias — the documented finding is about the pace of change.

THE QUESTION

Could AI surface a bias toward diversity in the U.S. workforce through early-20th-century photographs of African American students learning various trade skills?

Prof. Choi tested this with his own

computational tool analyzing visual clues, then with commercial AI (Gemini, ChatGPT) in June–August 2025.

THE FINDING

Revisiting the project in January–February 2026, commercial AI platforms had improved dramatically in 5–6 months — a pace that amazed the team.

The finding isn’t a verdict on bias. It’s about

how fast the tools are changing — which is itself a lesson for any educator planning curriculum.

Open question: when AI “improves,” does it improve at fairness — or just at sounding more confident?

13 of 22

PEDAGOGY

What This Teaches Our Students

"Who is missing,

who is flattened,

what context

is lost?"

The questions every student should ask of every AI output.

Bias is not a future problem

It’s in every output our students will read, write, or submit. Recognizing it is a core literacy skill.

The critical-thinking move

Treat AI outputs as drafts to interrogate, not answers to accept — name what is missing, flattened, or stripped of context.

Schomburg as a model

Bring AI to a sensitive archive, compare its readings against expert and community knowledge, document where it fails. That is human-centered AI in action.

14 of 22

THE PROJECT

PI

Our Project Website

humancenteredai.fitnyc.edu/

Built gradually over the project — a home for our team, projects, resources, and the 2026 symposium.

15 of 22

CREATIVE FUTURES · FIT · APRIL 10, 2026

Our One-Day Symposium

Framed around Socially Responsible AI (SRAI) and Human-Centered AI (HCAI) — a growing interdisciplinary field, with new graduate programs at Buffalo and beyond.

16 of 22

AUDIENCE FEEDBACK · BY THE NUMBERS

How It Landed

Post-symposium survey (N=12 responses)

3.83 → 4.58

Self-rated understanding of AI, creativity, culture & ethics (1–5 scale, before → after)

12 / 12

rated the multi-institutional mix “Extremely” or “Very” valuable

10 / 12

said the pacing was “about right”

Top requests for next year: hands-on workshops (7) · industry partnerships (5) · take-home toolkits (5) · cross-campus events (4)

17 of 22

AUDIENCE FEEDBACK · IN THEIR WORDS

What They Found Valuable

“In one or two sentences, what was the most valuable part of today for you?”

“Hearing students and faculty discuss together — not just about AI itself but its impacts and how we study them (environmental, psychological).”

— FIT faculty

“Learning about what other SUNY campuses are thinking and doing with AI — colleagues and students from different departments and professional fields.”

— FIT faculty

“Getting a better appreciation of both the negative and positive aspects of AI, particularly its effects on the environment.”

— FIT faculty

“Much effort from professors to teach AI future to students — was so impressive.”

— Guest

18 of 22

SPEAKER SPOTLIGHT · INDUSTRY & ETHICS

What They Actually Said

Kathleen O’Brien — IBM

“Fear to fluency.” AI accelerates every stage of design — research, ideation, production, validation. But empathy, taste, and judgment stay human. Bring AI in early, not as a finishing touch.

Evie Joselow — FIT, Art Market Studies

“Creative Credit.” When an AI-assisted work sells, who gets credit and value — the artist, the model, the dataset, or the system? Attribution is the central ethical and market question of AI art.

Marco Tedesco — Columbia (Keynote)

“The Climate Justice–AI Paradox.” AI models climate change at unprecedented resolution — yet training large models consumes enormous energy and water. The diagnostic tool is also a contributor.

U.S. infrastructure and climate-data networks — from Tedesco’s keynote.

[Spotlight — Industry & Ethics] Three substantive arguments — land each one.

Kathleen O'Brien from IBM — FIT alum — called her talk "Fear to Fluency." AI speeds up every stage of design work, but the human judgment about what's good, what's appropriate, and when to throw the AI output away never gets automated. Use it early, stay in charge.

Evie Joselow from FIT asked the central ethical question of AI art today: "Creative credit."When an AI-assisted artwork has value, someone has to answer who actually made it and who gets paid — the artist, the model, or the people whose work trained it. Attribution is the unsolved question of the AI art market, and we can't pretend it's settled.

And the keynote, Marco Tedesco from Columbia, on the Climate Justice–AI Paradox. AI is both the best tool we have for understanding climate change and a growing contributor to it through its energy and water use. The takeaway: you can't talk about AI's benefits without also counting its environmental cost.

TIP: "After Tedesco's keynote, the room went quiet. People realized: every classroom ChatGPT use has an energy footprint. At scale, that's a conversation we owe our students."

19 of 22

SPEAKER SPOTLIGHT · EDUCATION (1 of 2)

What They Built — Tools for Understanding

LLM4Art — Liu, Li / SUNY Buffalo

Multimodal LLMs answer “what am I looking at?” with structured visual, historical, and cultural analysis. Designed for learners and visitors without access to expert docents.

AI Card Deck Game — Ghidiu, Kural, Lee / Monroe CC

A physical, analog deck that lets students and faculty explore human-centered AI values without technical intimidation. Paired with MCC’s tiered framework — Levels 1–5 of AI use, from “No AI” to “Full AI.”

Left: LLM4Art interpretation interface, from the speakers’ own slides.

[Spotlight — Education 1 of 2] Two tools that scaffold understanding.

LLM4Art from Dancheng Liu and Zhaohui Li at Dr. Xiong's lab at Buffalo. That screenshot is the actual interface — a learner uploads an artwork, and the system gives back structured visual, historical, and cultural analysis. In the example it's Raphael's School of Athens, situated in the Italian Renaissance. AI as a bridge to art — not a replacement for looking. AI can be a patient docent — it helps a learner who's standing alone in front of an artwork answer "what am I looking at?" The message is AI as a bridge into art, scaffolding understanding rather than replacing the act of looking.

On the right, the AI Card Deck Game from Monroe Community College — Katie Ghidiu, Melis Kural, and Christina Lee. A physical deck of cards, no coding required, no technical intimidation. You don't need to code to think critically about AI. The message is that a low-stakes, analog, playful entry point lets non-STEM students and faculty build real AI judgment without the technical intimidation.

TIP: "The Card Deck is the most non-STEM-friendly intro to AI ethics I've seen. If you teach humanities, take a picture of this slide."

20 of 22

SPEAKER SPOTLIGHT · EDUCATION (2 of 2)

What They Built — Voices from Students & Research

Student Panel — Rizvi, Hodges, Pantano / FIT

“The Ethics of AI and Art History.” Students interrogating AI in museums and naming what humanities skills can do that AI cannot — empowerment as the confidence to question.

Chloé Martin — FIT Social Sciences

“AI Help-Seeking through Stress & Coping.” Studies how students’ AI use relates to self-efficacy and psychological distress — turning evidence into policy and curriculum guidance.

Images: speakers’ own symposium slides.

[Spotlight — Education 2 of 2] NirVana by name — this is her moment.

On the left, the student panel. NirVana Rizvi — our research assistant all summer — with Sophie Hodges and James Pantano. They asked, point-blank, whether AI belongs in museums at all. A humanities education is exactly what equips you to interrogate AI — the most empowered move a student can make isn't using the tool, it's having the grounding to ask whether it should be used at all. They named what humanities skills can do that AI cannot: contextual reasoning, ethical judgment, cultural sensitivity. For me, this was the most powerful moment of the day.

Empowerment isn't teaching students to use AI — it's giving them the confidence and the critical grounding to question whether AI belongs in a given context at all.

On the right, Chloé Martin from FIT Social Sciences. Her research is on AI help-seeking and its relationship to stress and coping in college students — the wellbeing piece of human-centered AI that nobody else is talking about, and she's asking the question with data.

TIP: "NirVana came into this project as an RA. She left it as a symposium speaker who challenged tenured professors. That's what empowerment looks like."

21 of 22

WHAT’S NEXT

PI

Future Directions

22 of 22

TAKEAWAYS

Empowering Learners — What It Takes

1

Center the learner: students as co-creators, not consumers

2

Design for non-STEM learners — assume no technical background

3

Treat AI as a scaffold for critical thinking, not a shortcut

4

Build communities of practice across disciplines and institutions

5

Ground everything in ethics, equity, and social responsibility (SRAI + HCAI)

humancenteredai.fitnyc.edu Kyunghee Pyun · Junghyun Ahn With thanks to SUNY IITG & CID

Thank you — questions welcome.