AI Proficiency and Literacy: The Next Frontier in Education

Throughout the summer, faculty gathered weekly to explore AI together. The sessions ranged from fundamentals to advanced applications:

• Understanding basics: What is a large language model? What is generative AI?
• Advanced use cases: Faculty sharing how they use AI in their own classes and research
• Project brainstorming: Developing use cases for university and K-12 educators

"It was fun. We had a blast this summer. I don't always say that when I'm hanging out at the office in the summer, but we had fun."

The newly opened center serves as a hub for incubating interdisciplinary research projects. Current plans include:

• AI workshops and talk series throughout the year
• Point-counterpoint discussions: Bringing together AI enthusiasts and skeptics to articulate strengths, weaknesses, and ways forward

The goal is not to create fights, but to have substantive discussions about how educators should navigate these powerful tools.

Emerging Technologies for Learning (Spring semester):

• Focus on AI and extended reality (VR, AR)
• Understanding strengths and weaknesses from an educational perspective

Philosophical Course on AI and Knowledge (Ed Psych):

• How AI tools are causing us to rethink the nature of knowledge itself
• Examining human learning and what it means to know

AI in Education (Historical Perspective):

• Looking at all waves of AI: neural networks, intelligent tutoring systems, expert systems, symbolic AI
• Understanding that AI isn't just ChatGPT - it's a rich history of tools supporting education

"I will only do it if I can take a historical look at AI... In fact, it's hot, and then it's not hot, and then it's hot, and then it's not hot. We're in another hot right now, we'll see."

Co-directed with Dr. Rebecca Quintana, this program prepares designers for technology-based learning environments:

• Residential program: Students work with learning designers across campus units
• Online MOOC series on Coursera: Learning Experience Design certificate available globally
• AI integration: Exploring how AI can support the design process itself

This led to a second Coursera series: "Exploring Generative AI for Learning Design" - brainstorming ideas to support designers using AI.

1. Design Partner:

Can LLMs work collaboratively with students during design activities? Not giving answers, but helping brainstorm, throwing questions back to help students think through problems.

2. Scaffolding Agent:

Using LLMs as tools to determine what kind of scaffolding to generate and show learners - and potentially to better fade scaffolding throughout system use.

3. Collaborator:

Training models to act as collaborators - understanding what it means to collaborate rather than just generate.

"We've been putting in NSF grants to explore these ideas about large language models being this sort of intermediary, helping, being a tool to think with rather than a tool to just generate answers for you."

Some students are already using LLMs as thought partners rather than answer machines:

• Getting ideas but still judging which direction to go
• Asking "what direction sounds good and why?" then deciding themselves
• Using generated multiple choice questions selectively: "This one's pretty good, I'll use that. This was just wrong. This was pretty good, but let me change it."

The key: students remain in charge, using AI as a brainstorming partner while maintaining their own judgment.

Era 1: Newspapers as Curators

News organizations synthesized, curated, and presented information to the public. Different newspapers offered different lenses, but there was a managed flow of information.

Era 2: Search Engines (~1995)

AltaVista and other search engines gave students direct access to vast information. The famous "storm front" story: students researching weather typed "storm front" and got a white supremacist website instead.

"We thought, oh, maybe search engines, we can't just let kids loose on these things."

This led to the concept of information literacy - teaching students to judge, trust, and synthesize information.

Era 3: Large Language Models (Now)

AI can rapidly generate tons of information - some helpful for brainstorming and learning, but also content that is false, misleading, and harmful.

A conversation with a literacy colleague revealed a growing concern:

"What I fear is that these tools are creating this illusion that I can get the right answer without having to struggle. But it's in the struggle that the learning happens."

Students increasingly want extremely detailed rubrics - not to learn, but to know exactly what points to hit for their grade. They focus on the end product rather than the process.

The problem: In education, especially graduate school, it's the process that matters. These tools may be shortchanging the cognitive struggles we want students to engage in.

From the discussion, a German participant observed:

"Humans tend to believe in something first, and then they try to find a rational for it. AI can help with this: 'Hey ChatGPT, if I want to do this, what is a good reason to do this?'"

Graduate students may already know what they want to say, then ask ChatGPT for supporting references - the opposite of proper research methodology where you read literature first, then form conclusions.

The old norm: "You're not supposed to cite anything you haven't read."

The new danger: People won't care if information is right or wrong, as long as it helps them convince others.

Different institutions are setting different boundaries:

• Journals: Most moving toward self-disclosure requirements for AI use
• NSF: "You can use it in your proposal, but if you screw up, that's on you"
• Reviewers: Not allowed to use AI in peer review
• Syllabi: Range from "no LLM use" to "tell me exactly what you did with it"

An interesting finding: Many students feel they can use generative AI for their own work, but get angry if they find out faculty are using it for something related to their work.

Equity concern: Different students have access to different tools - some can afford paid versions with better capabilities, while K-12 schools vary widely in what they can provide.