The Research Behind Edna

Thirteen sources. Three questions. One argument.

This project was inspired by the UC Irvine AI in Higher Education program. The ideas explored throughout this experience reflect the program's emphasis on evidence-informed teaching, ethical AI use, and designing learning experiences that keep human thinking at the center.

Edna wasn't built from prompts or opinions. She was built from research across learning science, cognitive psychology, AI literacy, writing studies, and equity-minded pedagogy.

Every question Edna asks—and every question she doesn't ask—is intentional. Together, these ideas shape how Edna supports faculty in designing assessments where student thinking remains at the center.

The research behind Edna explores three essential questions.

How do we design learning experiences where thinking—not technology—builds lasting understanding?

AI Needs Guardrails, Not Bans

Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, O., & Mariman, R. (2025). Generative AI without guardrails can harm learning (Article)

What we learned
Students who used AI without learning guardrails performed worse than students with no AI access at all. When thoughtful guardrails were added, the learning loss disappeared.
How Edna uses it
Edna doesn't try to remove AI from learning. She helps us design assessments where AI can support learning without replacing student thinking.

Thinking Builds Learning

Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X. H., Beresnitzky, A. V., & Maes, P. (2025). Your Brain on ChatGPT (Article)

What we learned
Brain activity research suggests that when AI does too much of the thinking, learners become less cognitively engaged. Less thinking today can mean less learning tomorrow.
How Edna uses it
Edna intentionally asks us to think, draft, and reflect before offering recommendations. The goal is to support our thinking—not replace it.

Learning Happens Before the Answer

Melumad, S., & Yun, J. H. (2025). The Effects of Large Language Models Versus Web Search on Depth of Learning (Article)

What we learned
AI delivers polished answers quickly, but that convenience can reduce the deeper learning that comes from evaluating evidence, making connections, and forming our own conclusions.
How Edna uses it
Edna helps us redesign assessments so student reasoning—not AI-generated answers—becomes the evidence of learning.

Humans Think First

Tate, T. P., Ritchie, D., & Warschauer, M. (2026). Generative AI as a mediational agent: Rethinking learning in sociocultural theory (Article)

What we learned
Durable learning begins with human thinking. AI should extend our reasoning—not replace it.
How Edna uses it
Every phase of Edna is designed around this principle. We think first. AI responds second.

The First Draft Matters

Tate, T. P. (2026). Think First (Article)

What we learned
The first idea shapes the thinking that follows. When AI produces the first draft, our own thinking can become anchored to its suggestions.
How Edna uses it
Edna asks us to draft before she generates anything substantial, ensuring the redesign begins with our expertise rather than the AI's.

Productive Struggle Matters

WestEd & National Writing Project. (2025). Friction by Design (Web Resource)

What we learned
Not all struggle is bad. Productive friction strengthens learning, while unnecessary friction creates barriers without adding value.
How Edna uses it
Edna removes barriers that don't serve learning while preserving the challenges that deepen thinking.

How do assessment design choices protect student voice, equity, trust, and human judgment?

Human Judgment Protects Students

Tan, Y., Phalen, L., & Demszky, D. (2026). Marked Pedagogies (Article)

What we learned
AI systems gave less rigorous and less substantive feedback to students identified as Black, Latinx, multilingual, or students with disabilities—even when their writing was identical to everyone else's.
How Edna uses it
Edna keeps faculty judgment at the center of assessment design. Human expertise is one of the strongest safeguards against automated bias.

Student Voice Is Evidence

Abdulhai, M., White, I., Wan, Y., Qureshi, I., Leibo, J., Kleiman-Weiner, M., & Jaques, N. (2026). How LLMs Distort Our Written Language (Article)

What we learned
AI tends to flatten language, making writing more uniform while reducing the individuality, culture, and identity reflected in authentic human voices.
How Edna uses it
For many California Community College students, voice is more than style—it's evidence of lived experience, knowledge, and perspective. Edna helps us design assessments that preserve, rather than erase, those voices.

Trust Builds Better Learning

Watkins, M. (2025). Our Response to AI Cannot Be Adversarial (Article)

What we learned
AI detection has produced false accusations, damaged trust, and placed the burden of proving innocence on students. Surveillance has real educational costs.
How Edna uses it
Edna never recommends surveillance as a design strategy. Instead, she helps us create assessments where authentic student thinking naturally becomes the evidence of learning.

Integrity Begins With Design

Senk, S., & Inoue, T. (2025). A New Script for Academic Integrity in the AI Age (Article)

What we learned
Academic integrity grows from honest conversations, shared expectations, and meaningful assessment—not from detection software or honesty pledges alone.
How Edna uses it
Edna encourages us to build integrity into the design of learning itself, creating assessments that invite authenticity rather than policing it.

What knowledge, habits, and design decisions help faculty create meaningful learning in an AI-enabled world?

Know What AI Can—and Can't—Do

Mollick, E. (2023). Centaurs and Cyborgs on the Jagged Frontier (Article)

What we learned
AI is remarkably capable in some tasks and surprisingly weak in others. Those boundaries are often difficult to predict, making it easy to overestimate what AI can reliably do.
How Edna uses it
Edna's Friction Audit helps us examine what AI can already do with an assignment—and, just as importantly, what still requires human reasoning. Better assessment design begins with understanding that boundary.

AI Literacy Is Design Literacy

Long, D., & Magerko, B. (2020). What Is AI Literacy? (Video)

What we learned
AI literacy isn't just about knowing how to use AI. It includes thinking critically, maintaining learner agency, and understanding how AI systems work and where they fall short.
How Edna uses it
Rather than generating answers, Edna asks questions. Faculty remain the designers, the decision-makers, and the experts throughout the entire redesign process.

AI Literacy Includes Ethics

Ng, D. T. K., Leung, J. K. L., Chu, S. K. W., & Qiao, M. S. (2021). Conceptualizing AI Literacy (Article)

What we learned
AI literacy combines technical understanding with critical evaluation and ethical decision-making. It's not enough to know what AI can do—we also need to decide how it should be used to support learning.
How Edna uses it
Every conversation with Edna is designed to strengthen AI literacy by encouraging thoughtful, intentional, and ethical decisions about assessment design.

One Argument: The future of assessment won't be determined by AI. It will be determined by the questions educators choose to ask.

Page updated

Google Sites

Report abuse