Syracuse University Libraries will roll out a free, hands-on workshop series this fall that cuts through the hype surrounding generative AI, guides students through the secure use of Microsoft Copilot, and opens the door to large-scale text mining with ProQuest’s TDM Studio. The program, announced for the Fall 2025 semester, pairs rapid skill-building sessions—like a Zotero citation bootcamp—with deep dives into AI literacy that many universities have struggled to deliver.

The line-up comes at a moment when campus IT teams and faculty are scrambling to keep pace with tools that students already use, often without understanding the privacy trade-offs or the risks of fabricated citations. By weaving together practical technology instruction with frank discussions about model hallucination and data governance, the libraries are staking out a model that other research institutions should watch closely.

Workshop roster: A 10-week sprint through modern research literacy

The “Learn! at Syracuse University Libraries” program kicks off with foundational sessions for new graduate students and quickly adds advanced topics. Organized by librarian Giovanna Colosi, the workshops are all free, open to SU students, faculty, and staff, and require pre-registration.

  • Getting Started with Research — offered both in-person and via Zoom, this session teaches database strategies, citation tracking, and how to incorporate AI tools into discovery without losing critical thinking.
  • Saving, Organizing, and Citing Your Sources with Zotero — a practical walkthrough of the open-source reference manager, covering browser connectors, group libraries, and the Word plugin.
  • Using SU Libraries as an Online or Distance Student — remote access to collections, interlibrary loan, and 24/7 chat support.
  • A Student’s Guide to Using Microsoft Copilot for Coursework and Research — a dedicated online session on September 26 that shows how to draft, summarize, and study with Copilot while keeping academic integrity intact.
  • Demystifying AI: What’s Really Inside the Black Box? — an approachable explainer on how generative models work, why they hallucinate, and what responsible use looks like.
  • Introduction to ProQuest Text and Data Mining (TDM) Studio Visualization Dashboard — a no-code interface for topic modeling, geographic trends, and sentiment analysis across massive news and scholarly corpora.

Colosi said the workshops are “especially relevant for new graduate students, providing practical skills, research strategies and exposure to SU Libraries’ vast collection.” But the inclusion of Copilot and TDM Studio signals that the libraries are targeting not just newcomers but also researchers hungry for scalable tools.

Why Windows users should care

For Windows-centric students and researchers, these workshops address a gap that often goes unnoticed: the tools they’re expected to use on campus are deeply integrated with the Microsoft ecosystem, yet guidance on safe configuration lags behind. The Copilot session will walk through features like Copilot Notebooks, study-guide generation, and drafting assistants inside Word and Teams—all of which Microsoft is pushing to education customers. Meanwhile, TDM Studio’s workbench environment (Jupyter-backed, Python/R) can complement local Windows development setups, but comes with strict data export limits that many researchers trip over.

By attending, Windows users can learn to harden their workflows: use institutional Entra ID accounts instead of personal logins for Copilot, save prompts and outputs alongside Zotero-managed references, and plan TDM projects that stay within platform quotas. These are not abstract concerns; security researchers have already flagged vulnerabilities in AI-integrated productivity suites, and default biometric enrollment in some Teams features has triggered real privacy alarms. The libraries’ program provides a timely blueprint for threading the needle between productivity and protection.

Inside the Copilot workshop: Promise meets policy

The September 26 session explicitly targets students who want to use Copilot for coursework—drafting essays, summarizing readings, creating flashcards—but who may not realize that the tool’s behavior changes dramatically depending on account type. Here’s what attendees should expect and what they should grill instructors about.

The institutional account safety net

Microsoft differentiates between consumer Copilot (web, mobile) and Copilot used with an organizational Entra ID. Data from Entra ID-signed-in users is excluded from model training, and commercial data protections may apply in Microsoft 365 Copilot. For students logged in with an @syr.edu account, this means prompts and outputs theoretically stay within the university’s control. The workshop must make this distinction crystal clear: using a personal Microsoft account could send sensitive research prompts into a less protected pipeline.

The Center for Democracy and Technology and other watchdogs have raised concerns about student data in AI products, and without explicit institutional guidance, many learners default to the easiest login—often personal. The libraries’ session is an opportunity to flip that behavior.

Provenance and the hallucination problem

Copilot can produce polished text, but it also invents citations with alarming confidence. The workshop should pair every demonstration of drafting or summarization with a live example of output verification: taking a Copilot-generated paragraph, extracting the claimed sources, and running them through library databases to see if they exist. If instructors don’t do that, they risk sending students away with a tool they trust too much.

One mitigation: save all Copilot prompts and outputs in a Copilot Notebook or OneNote, then cross-link each generated reference to a Zotero item. That way, when a professor asks, “Where did this come from?” the student has a chain of evidence. The Zotero session, which covers the Word plugin (“Cite while you write”), becomes the perfect partner workshop.

Security vulnerabilities to watch

Recent disclosures have shown that AI add-ins can become vectors for data exfiltration. Reports of flaws in Copilot Studio and Microsoft 365’s AI integrations, along with incidents like Teams’ default auto-enrollment in facial recognition, underscore that these are not hypothetical problems. The workshop should mention that campus IT can implement conditional access policies, but that students still bear responsibility for not throwing unpublished manuscripts or personal identifiers into a chat window.

Demystifying AI: Beyond the “black box” trope

Too many AI literacy sessions stop at “models are trained on data.” The SU Libraries’ Demystifying AI workshop appears designed to go further, unpacking tokenization, probability-based generation, context windows, and failure modes. That technical grounding is essential for Windows power users who may later fine-tune local models or run Python-based analysis alongside Copilot.

Key concepts that should appear in the session:

  • Hallucination vs. unjustified specificity: Why a model confidently claims a fake paper has a DOI.
  • Bias amplification: How training data skews outputs in ways that affect citation recommendations or summarization.
  • Prompt engineering basics: Not as a magic spell, but as a way to understand model limitations.
  • Verification workflows: Treating AI output as a draft to be fact-checked against primary sources.

When a library teaches these concepts alongside a tool-specific Copilot session, it creates a mental model that outlasts any product update. Students leave knowing that “AI said so” is never enough.

ProQuest TDM Studio gives SU researchers access to millions of news articles and scholarly papers for text mining—without the copyright infringement risk of scraping the open web. The Libraries’ workshop will walk attendees through the Visualization Dashboard, a no-code interface that can generate topic models, geographic heatmaps, and sentiment timelines. For coders, the Workbench environment offers Jupyter notebooks with pre-installed libraries.

But scale comes with strings:

  • Export limits: Syracuse’s TDM policy caps dataset sizes and restricts what can be downloaded. Researchers planning a thesis-scale analysis need to request workbench access early and design projects within platform guardrails.
  • Model integration: Some TDM workbenches now allow LLM-based NLP tasks, but that can send data outside the institutional bubble. The libraries must clarify whether external model calls are permitted, given that ProQuest emphasizes local compute and rights-cleared content.
  • Seat scarcity: Workbench spots are limited and time-bound, so team-based projects require coordination. The workshop should push students to consult the Data Services team before they hit a wall.

For Windows users who might otherwise spin up a local Python environment and hit a paywall or a licensing tangle, TDM Studio offers a sanctioned alternative. The workshop’s real value is in teaching researchers what questions to ask before embarking on a large-scale analysis.

Zotero: The glue that holds it all together

Among the flashier AI and data-mining sessions, the Zotero workshop might seem basic—but it’s the lynchpin. Zotero’s browser connectors capture metadata from databases and publisher sites with a click. Its Word plugin inserts citations in real time, and group libraries allow co-authors to share references without version-control chaos.

However, Windows installations can trip over security software that blocks the Word add-in. The Zotero.dotm file needs to land in Word’s Startup folder, and institutional machines with aggressive Group Policy settings sometimes prevent automatic placement. A quick troubleshooting demo—showing students how to manually place the file under an IT-administered exception—would pay dividends.

Pairing Zotero with Copilot creates a workflow that’s hard to beat: use AI to generate a rough outline or summarization, but build every citation from a Zotero-managed source. The library’s own guides (researchguides.library.syr.edu) reinforce this, and the fall schedule mirrors that philosophy.

Practical playbook for attendees

  1. Register early — sessions fill up, and attending the full Copilot-AI-Zotero triad ensures a coherent workflow.
  2. Use your @syr.edu account for all Copilot work; confirm with IT that commercial data protections are active.
  3. Save everything: prompts, Copilot outputs, and notes go into a searchable notebook, linked to Zotero records.
  4. Request TDM Studio workbench access now if you’re planning a project; know the export caps before you design your analysis.
  5. Test Zotero’s Word plugin on your Windows machine before the session; have IT’s number handy if the add-in doesn’t appear.
  6. Keep Windows and Office updated — patches can alter plugin behavior and security defaults.

What the libraries get right — and where they must deliver

Strengths: The schedule balances beginner and expert tracks, mixes in-person and hybrid formats for distance learners, and acknowledges that AI literacy is not a separate skill but woven into research integrity. Offering Copilot and AI literacy back-to-back is smart: students learn both how to use the tool and why they should be skeptical of it.

Gaps that could undermine impact:
- If the Copilot workshop doesn’t explicitly address model hallucination and fabricated citations, it risks legitimizing a tool that can actively harm scholarly work.
- Rapid product changes in Microsoft’s education roadmap mean workshop content could become outdated within months; version notes and links to vendor docs are non-negotiable.
- TDM Studio’s complexity demands office hours or follow-up labs; a single demo won’t be enough for first-time users.
- Security briefings must be specific: “Don’t share sensitive data” is meaningless without examples of what constitutes sensitive data in a research context.

The bottom line

Syracuse University Libraries has assembled one of the more thoughtful workshop programs in higher education for Fall 2025. It doesn’t just teach tools—it frames them within principles of verification, privacy, and compliance that Windows-centric researchers need. For anyone on campus who wants to use Copilot without embarrassment, explore big data without legal risk, and manage citations without panic, the sequence of sessions from late August into October is a low-cost, high-impact bet.

Attendees will walk away with more than tips: they’ll have a mental model that treats AI as a collaborator to be interrogated, not a oracle to be trusted. In a year when academic integrity guidelines are still being written in real time, that’s the best outcome a library workshop can deliver.