Backpack — Rutgers AI Study Companion

00 — THE VISION

A supportive academic and emotional companion for students.

Backpack helps students stay organized, confident, and stress-free — built right into the Rutgers dashboard they already open every day.

THE BRIEF

Rutgers wanted to enhance the student learning experience using AI tools.

THE PRODUCT

An AI study companion embedded as a component of the Canvas dashboard.

MY ROLE

Designer on a team of four — research, flows, wireframes, and high-fidelity UI.

01 — THE PROBLEM

The dashboard shows information — but offers no personalized study help.

Students rely on the Rutgers dashboard daily, yet it lacks a simple way to ask doubts, access missed notes, or get guided study support.

So they feel overwhelmed, fall behind easily, and struggle to reach the grades they're aiming for.

THE CORE TENSION

"Students struggle to understand why their grades don't match their effort."

Students feel overwhelmed, disorganized, confused, burnt out and inconsistent

What's currently not working

The current Rutgers Canvas dashboard — information, but no personalized study help

Disorganization Confusion Burnout Inconsistency

02 — RESEARCH & INSIGHTS

What students actually wanted from AI.

Research covered the tools and challenges of traditional learning, how students use AI to cope, and the issues with off-the-shelf AI tools — grounded in interviews with 9 undergraduates.

Project overview — Rutgers wanted to enhance student learning with AI; trust the process

Context inquiry — interviewed 9 undergraduate students

Navya — a curious learner who believes AI should enhance, not replace, human thinking

Key learnings

LEARNING 01

Trust is fragile. Conflicting AI responses and lower grade outcomes reduced students' trust in AI-only learning.

LEARNING 02

People still matter. Students preferred in-person support — lectures, office hours, TA sessions — especially before exams.

LEARNING 03

Off-the-shelf falls short. Generic tools don't know the course, the syllabus, or what was actually taught.

Key learnings — conflicting AI responses reduced trust; students still preferred in-person support

The team mapping insights together during a workshop

03 — WHO I DESIGNED FOR

Meet Ivy — overloaded, conscientious, and cautious about AI.

PRIMARY USER

Ivy Hubber · 21

// Cell Biology & Neuroscience major
// Creative artist, active in clubs
// Balances heavy coursework + labs
// Uses AI sparingly — environmental values
// "AI's helpful, but I can't fully trust it."

Her struggle: balancing effective learning with ethical, sustainable AI use.

Ivy faces information overload from 300-page lecture notes, and 900+ minutes of recordings to review. She feels overwhelmed and unheard when uploads are delayed, and study groups are inconsistent.

A day in Ivy's study life

01 · AT HOME

Reviews notes

Debates whether to use AI or just re-read. 😟

02 · LECTURE NOTES

300 pages

Notes are dense and hard to recall. 😵

03 · RECORDINGS

900+ minutes

Too long to review; doubts go unanswered. 😩

04 · STUDY GROUP

Inconsistent

Different levels; hard to meet, easy to feel lost. 😞

The full journey map

Ivy Hubber user journey map — timeline, actions, thoughts, feelings, touchpoints and opportunities

04 — OPPORTUNITIES FOR DESIGN

Four bets for a calmer way to learn.

The research pointed to clear openings — each one a chance to cut overwhelm and rebuild trust.

Key takeaway: Ivy's journey and the opportunities for design

BET 01

AI-powered summaries

Combine professor notes and self-notes into short, clear pages.

BET 02

Explain it differently

Short video explanations and interactive quizzes for faster comprehension.

BET 03

Show the reliability

AI reliability indicators plus cross-referenced textbook explanations.

BET 04

Learn together

A collaborative space where students quiz each other and discuss misunderstood concepts.

05 — THE SOLUTION

Backpack — built into the dashboard, not bolted on.

A supportive space to revisit concepts, catch up on missed notes, and get clear, reliable guidance whenever classroom help isn't available. Students onboard by choosing courses and setting simple study goals.

Onboarding — Backpack is the go-to study companion

Welcome back, Tiffany — your desk is ready

The Backpack dashboard — Hi, I am BackPack, with four smart actions

ASK A QUESTION

When students hit their limit, the Rutgers-built helper breaks the problem into small, guided steps.

UPLOAD FILE

Upload lecture notes; Backpack extracts, summarizes, and explains them in multiple formats.

PEP TALK

A reach-out tool to connect with a TA, advisor, or campus counselor.

REVISE TOPIC

An assessment-driven approach that surfaces weak points needing more practice.

Inside the experience

Upload successful — color-coded reliability of every result

Analysis complete — a clear summary and a choice of how to learn

Video explanation with a detailed, color-coded breakdown

Video explanation playing a concept video

TRUST, MADE VISIBLE

Every answer is color-coded: green = verified from university resources, yellow = cross-check recommended, red = uncertain (scraped from forums).

Generate flashcards Take a test Video explanation Study guide Optional voice greeting

06 — PROCESS & EVALUATION

From low-fi flows to a heuristic-checked experience.

Low- and mid-fi wireframes mapped the end-to-end flow — log in, add a link, generate a summary, upload documents — to validate the core interactions before adding visuals.

End-to-end process — research, wireframing, design, iteration, testing

Sketches & low-fidelity wireframes

Lo-fi storyboards and sketches taped up during ideation

Low-fidelity wireframes mapping the end-to-end flow

Heuristic evaluation

Running a heuristic evaluation gave me a structured way to find what wasn't working before testing with students — unclear labels, missing feedback states, and moments where students might feel lost.

Usability testing

I ran moderated sessions with students, walking each one through core tasks — onboarding into Backpack, uploading a lecture file, asking a question, and reading a color-coded answer. Watching where people hesitated showed me which labels were unclear and where a little reassurance was needed. The biggest signal: students trusted the tool far more once the reliability colors and sources were visible, and they moved faster when each screen surfaced just the next best action. Those rounds of feedback directly shaped the final, calmer flow.

07 — END-TO-END & REFLECTION

Reducing cognitive load is a design decision — not a default.

The throughline across research, wireframing, design, and iteration was calm: surface only what the student needs next, and make the AI honest about what it knows. Trust came from restraint and clear reliability cues, not from more features.

Next: quantitative usability testing — task success rate, errors, time-on-task, and SUS — to validate clarity and efficiency with real students.