ACH 2026 · Friday, June 26 · 12:00–13:15 CDT

Building Community-Oriented Infrastructures for AI Experimentation

Matthew K. Gold, Luke Waltzer, Zach Muhlbauer, Azucena García Gutiérrez, and Stephen Zweibel

CUNY Graduate Center · Association for Computers and the Humanities (ACH) 2026

Matthew K. Gold

Approaching AI from a CUNY DH Perspective

CUNY AI Lab · Graduate Center Digital Initiatives (GCDI)

Matthew K. Gold

The CUNY Context

The largest urban public university
Mission: to educate “the children of the whole people” of New York City
The CUNY Graduate Center as a central node

CUNY: 1 University, 26 Colleges across the five boroughs

Matthew K. Gold

CUNY as an Engine of Social Mobility

“Because the elite colleges aren’t fulfilling that responsibility, working-class colleges have become vastly larger engines of social mobility. The new data shows, for example, that the City University of New York system propelled almost six times as many low-income students into the middle class and beyond as all eight Ivy League campuses, plus Duke, M.I.T., Stanford and Chicago, combined.” — David Leonhardt, “America’s Great Working-Class Colleges,” New York Times, January 18, 2017

Matthew K. Gold

DH at CUNY: Building Open Knowledge Infrastructures

Building open-source platforms
Building in response to community feedback and needs
Prioritizing open source and open access
Providing alternatives to enterprise infrastructure

Matthew K. Gold

The Challenge

The Scandal of Digital Humanities — Brian Greenspan, Debates in the Digital Humanities 2019

Matthew K. Gold

AI at CUNY

Embed reflexive critique and a critical approach to technology
Focus on classroom use
Build alternatives to enterprise-level initiatives
Preserve CUNY values

Matthew K. Gold

CUNY Graduate Center AI Guidelines

Expectations around accountability, collective responsibility, data privacy, right of refusal, and transparency
Recommendations that foreground a critical approach to AI
Discussion of ethical issues around commercial AI models
Tools for disclosure that build upon the Generative AI Delegation Taxonomy (GAIDeT)

CUNY Graduate Center AI Guidelines (Draft) — gc.cuny.edu, last revision June 5, 2026

Luke Waltzer

The Critical AI Literacy Institute

CUNY AI Lab · Teaching and Learning Center (TLC), CUNY Graduate Center

Luke Waltzer

The Critical AI Literacy Institute: Origins

CUNY Context
Funding from Google.org
CALI as a Response

Luke Waltzer

CALI: Grounding Scholarship

Critical University Studies
Critical Ed Tech and Infrastructure Studies
DH, Science Education, Educational Development

Luke Waltzer

CALI as an Intervention

Four interconnected areas

Luke Waltzer

CALI Goals

Reasoned Adoption Informed Refusal
Communities of Practice
Research, Tinker, Advocate

Luke Waltzer

CALI 1

22 Faculty from 12 Campuses

Outcomes

Curricula
Survey Responses and Focus Groups
Faculty Reflections
Collaborations

Luke Waltzer

CALI 2

23 Faculty from 12 Campuses

Tracks

Critical Foundations
Ecological Implications
T(h)inkering

Faculty member presenting an 8-class module on critical AI literacy at the CUNY Graduate Center

Zach Muhlbauer

Experimental Learning Infrastructures: The CUNY AI Lab Sandbox

CUNY AI Lab · CUNY Graduate Center

Zach Muhlbauer

Precursors to the CUNY AI Lab

Flowchart: Teaching and Learning Center (TLC) sits to the left of four precursors: Teach@CUNY AI Toolkit (2023), Teaching and Learning Center (TLC) Workshops (2023–2024), Communities of Practice (2024–2025), and Critical AI Literacy Institute (CALI 1, 2025). These converge into the need for institutional capacity to provide shared access to faculty- and staff-made AI tools and resources. Graduate Center Digital Initiatives, Mina Rees Library, and American Social History Project / New Media Lab contribute to that need. The flow then moves to the Proposal (Spring 2025), then CUNY AI Lab Year 1 (Fall 2025–Spring 2026).

CUNY AI Lab homepage shown beside the About Team leadership page

CUNY AI Lab homepage shown beside the About Team fellows and collaborating units page

Zach Muhlbauer

What is the CUNY AI Lab Sandbox?

Step 1: sign in to the CUNY AI Lab Sandbox

Zach Muhlbauer

Transparency and Privacy by Design

Zach Muhlbauer

Evaluating and Comparing Open-Weight Models

Model comparison in the Sandbox, prompt one

Model comparison in the Sandbox, prompt two

Zach Muhlbauer

Supporting Purpose-Built, Custom Models

Zach Muhlbauer

Communities of Practice within the Sandbox

Diagram titled Communities of Practice w/in Sandbox, showing the CUNY AI Lab Sandbox as shared infrastructure for Spanish courses, T(h)inkering fellows, CCNY faculty development, TLC workshops, grad-instructor inquiry groups, and research partnerships

Zach Muhlbauer

T(h)inkering with Faculty Fellows

T(h)inkering fellows across five disciplines

Zach Muhlbauer

Workshopping Project Pilots: AmigAI

Azucena García Gutiérrez

AmigAI Pilot: Classroom Implementation

City University of New York (CUNY) · The Graduate Center

Azucena García Gutiérrez

Classroom Context

Intermediate Spanish · SPAN 204 · Queens College, CUNY

A mixed classroom of heritage (HL) and second-language (L2) learners
Spanish-language textbooks lean on technologies that ignore diverse Latinx and Afrolatinx realities (Padilla & Vana, 2022; Vana & Padilla, 2024)
Equip students to critically analyze pedagogical texts and build a Critical Language Awareness (CLA) (Leeman, 2018)
Use AmigAI to help students challenge texts from their own textbook

Azucena García Gutiérrez

Excerpted System Prompt

"You are a guide for conducting discourse analysis of educational texts in Spanish for intermediate-level college students (B2 level on the CEFR scale). You challenge the student to analyze the text critically. Your students include both Spanish L2 learners and heritage language learners. You speak Spanish at an intermediate level (B2 level on the CEFR scale). You never use praise, compliments, or validation in your responses. You guide the student only through questions."

The AmigAI 204 configuration open in the CUNY AI Lab Sandbox interface

Azucena García Gutiérrez

Interactions

Texts and guided questions

Three texts from the Spanish textbook: immigration, El Caribe, and Guatemalan Indigenous populations
AmigAI surfaces biases, stereotypes, omissions, and different realities within the text through guided questions

AmigAI

Te pido que mires de nuevo el primer párrafo. En las líneas 3 a 6 dice que 'con mayor frecuencia el latino llega a los Estados Unidos con un nivel de estudios más alto y mejor preparado para ejercer trabajos bien remunerados'.

Student A03

¿Esta afirmación del texto, de que los trabajos son 'bien remunerados', coincide con la realidad de muchos inmigrantes latinos que conoces o de la que has oído hablar?

Azucena García Gutiérrez

Classroom Application

Fundamental Components

Reading comprehension of the reading materials
Explanation of the task with AmigAI; onboarding instructions
Individual interaction with AmigAI
Debriefing discussion as a class

Illustration of a student in thought, surrounded by wordless speech bubbles

Azucena García Gutiérrez

First Impressions

Positive Components

More intentional

"…overall, this was a very positive interaction. It asked very specific questions, and it asked more about, like, in conversation." — Student 1

"I didn't expect it to be so detailed with these questions. And I thought it was very cool — quotes directly from the text." — Student 2

"When you have someone, or something else, to ask questions to think deeper, it kind of gives you a deeper understanding of what you're reading." — Student 4

"A before" and "an after" the readings

"It feels informative, but it doesn't portray the negative side." — Student 1

"The chatbot actually helped me see the real side of immigration that the text might not have wanted to cover." — Student 2

"The text never talks about the push back on immigration or negative attention…" — Student 2

AI Critical Awareness

"It makes me think of how ChatGPT is being trained — what are their perspectives, what kind of information are we actually reading?" — Student 3

"Los libros de texto cuentan con formas de enseñar relacionadas a una agenda." — Student 3

Azucena García Gutiérrez

First Impressions

Negative Components

Awareness of their learning process through AI

"Analyzing text… it's hard to do that on an AI-chat basis model." — Student 2

"There's a lot more opportunity for the student to voice their input in a natural way when it's in a real-life setting, like a classroom." — Student 2

"I feel like with AI, you lose that personal touch." — Student 2

"And doing it with the chatbot, you lose that humanity." — Student 2

"The false positivity feels like you are talking with a robot, a screen." — Student 1

Monolingual Approach

"It would not translate for me. It just repeats the question." — Student 1

"If it's not gonna help me understand the text, what's the point?" — Student 1

Azucena García Gutiérrez

Conclusions

A pedagogical tool for AI critical-literacy discussions
Challenging to use under translingual approaches and at lower levels
Conversations that develop critical awareness
Technical issues and invisible labor remain real costs

{
  "id": "amigai--204",
  "base_model_id": "deepseek.v3.2",
  "name": "AmigAI 204",
  "params": {
    "system": "Eres un guía para hacer un análisis del discurso de textos educativos en español (nivel B2). Nunca usas elogios, cumplidos ni validación; guías al estudiante únicamente a través de preguntas y lo retas a analizar críticamente el texto. [… 11 pasos: del perfil del estudiante al feedback lingüístico …]",
    "max_tokens": 168,
    "top_k": 40,
    "top_p": 0.9,
    "temperature": 0.3
  },
  …
}

Azucena García Gutiérrez

Examples of Interactions

AmigAI conversation transcript analyzing the text 'Corriente latina,' with the student's written response

Azucena García Gutiérrez

Thank you

CUNY, The Graduate Center · azucena.garciagutierrez39@gc.cuny.edu

Stephen Zweibel

Harnessing Agentic AI: Tools for Open University Infrastructure

CUNY AI Lab · Mina Rees Library, CUNY Graduate Center

Stephen Zweibel

What the Lab Builds and Runs

How faculty and students build and run their own AI tools
The impact of agentic AI on research and libraries
From building tools to publishing them live

Stephen Zweibel

Tools Portal

tools.ailab.gc.cuny.edu — standalone applications
Media tools: transcription, PDF accessibility, image description; plus a disclosure tool and agentic studios
Zero-retention: the providers store nothing from prompts or outputs

Animated demo of the CUNY AI Lab Multilingual Transcription Suite showing an audio upload and generated transcript

Stephen Zweibel

Working with Agentic AI

Agent Studio and Site Studio for research tasks
A shift from conversation toward task delegation
Faculty and students build tools with minimal coding

Site Studio

zach-muhlbauer Publish

Both updates are done:

Bio shortened — trimmed to one clean sentence that captures your research focus
Photo updated — your actual photo now shows in place of the placeholder

The subtitle still says “Assistant Professor, Department of Sociology” — want to update that to reflect your actual role and department?

PhD candidate in English / Co-Founder of the CUNY AI Lab

file index.html

Done! Subtitle now reads “PhD Candidate in English · Co-Founder, CUNY AI Lab.” Want to update any of the nav links (Email, CV, Research, Teaching) with real URLs?

Ask a follow-up…

Zach Muhlbauer

PhD Candidate in English · Co-Founder, CUNY AI Lab

My research explores digital public spheres, literacy technologies, critical AI studies, creative coding, and electronic literature — with a particular focus on how generative AI is reshaping composition and coding classrooms, and what students’ everyday literacy practices reveal about our shifting knowledge landscape.

EmailCVResearchTeaching

Agent Studio

What would you like to work on?

Search APIs, analyze data, create visualizations, or build tools.

Ask anything or describe what you want to build…

Try these

Search papersAnalyze trendsFind booksBuild a tool

Built for the CUNY community

Stephen Zweibel

Teaching Librarians to Build with AI

Agentic AI for Library Practice: a 16-week program for CUNY librarians, no programming background needed
Build working tools — Primo dashboards, LibGuides interfaces — with Claude Code, then investigate what it built
Spot hallucinations and errors, verify against the docs; better equipped for work with patrons and students

Stephen Zweibel

Tools CUNY librarians built

Harvested 133,000+ records from CUNY Academic Works
A daily-rebuilt RSS feed for a library journal
Turned an OER spreadsheet into a searchable guide
Flagged at-risk database access after one campus lost HeinOnline

Stephen Zweibel

An accessible scholarly record

Students can run an accessibility check before depositing a dissertation
Staff remediate library PDFs at scale, with AI doing the first pass
Built and run by the lab for the Mina Rees Library

CUNY AI Lab PDF Accessibility app showing a PDF upload area and checks for OCR, structure, alt text, tags, validation, and fidelity

Stephen Zweibel

Kale Workbench

A coding agent in the browser, like Claude Code or Codex — on the lab's Cloudflare infrastructure, with open models
The lab runs it, so it sets the model and the limits
Nothing is pushed or published without your approval; hands off to Kale Deploy when ready

Stephen Zweibel

Kale Deploy

Agents deploy it themselves, through an MCP server
Build with an agent, and it's live at <project>.cuny.qzz.io — the first CUNY home for live web projects
Hosts many projects at once for very little; users run their own

Stephen Zweibel

From Sandbox to a Live Web App

Sandbox → Workbench → Kale Deploy
A complete path, from an open model to a live app
Hosted by the lab, meant for everyone at CUNY

Stephen Zweibel

CUNY-Run AI Infrastructure

Together they let CUNY run its own AI infrastructure
Built so the people who use it can inspect and operate it

CUNY AI Lab · ailab.gc.cuny.edu

Q&A & Discussion

ACH 2026

Links & Readings

LinkCUNY AI Labcuny.is/ai LinkGC Teaching and Learning Centercuny.is/teaching LinkCritical AI Literacy Institutecuny.is/cali LinkGC Digital Initiativescuny.is/gcdi LinkAmerican Social History Projectashp.cuny.edu LinkGC Librarylibrary.gc.cuny.edu

ReadingBender & Hanna.The AI Con. Harper.

ReadingBrier & Fabricant, 2016.Austerity Blues. Johns Hopkins University Press.

ReadingCary & Hull.Comprehensive AI Literacy.

ReadingCottom, 2025.Daring Ideas for the Future Series.

ReadingCrawford, 2021.Atlas of AI. Yale University Press.

ReadingFreire & Macedo, 2000.Pedagogy of the Oppressed. Continuum.

ReadingGold & Klein, 2023.Debates in the Digital Humanities 2023. University of Minnesota Press.

ReadingGuest et al., 2025.Against the Uncritical Adoption of “AI” Technologies in Academia. Zenodo.

ReadingHao, 2025.Empire of AI.

ReadingHeeg & Avraamidou, 2026.Demystifying AI. PROSPECTS.

Readinghooks, 1994.Teaching to Transgress. Routledge.

ReadingIn Defense of the Arts, Humanities, and Social Sciences, 2025.Journal of College Science Teaching.

ReadingLeeman, 2018.Critical language awareness and Spanish as a heritage language. Routledge.

ReadingMader & Dertien, 2016.Tinkering as Method in Academic Teaching.

ReadingNewfield, 2018.The Great Mistake. Johns Hopkins University Press.

ReadingPadilla & Vana, 2022.The “Other” Latinx. Journal of Identity, Language, and Education.10.1080/15348458.2021.2014845

ReadingPetrich, Wilkinson, & Bevan, 2013.It Looks Like Fun, But Are They Learning?

ReadingResnick & Rosenbaum, 2013.Designing for Tinkerability.

ReadingSano-Franchini, 2026.What’s “Critical” about “Critical AI”? WAC Clearinghouse.

ReadingSuarez et al., 2025.Critical AI Literacy.

ReadingVana & Padilla, 2024.Tales of (under)representation. Hispania, 107(2).10.1353/hpn.2024.a929127

ReadingWatters, 2021.Teaching Machines. MIT Press.