EVOLVE AI INSTITUTE

Free faculty toolkit · ARAD 2.0

The ARAD Framework

AI-Responsive Assignment Design for Human Learning

ARAD 2.04 modes · 6 principles · 8 dimensions · GOALS
Get the slides and worksheet Find your mode

New to ARAD 2.0? Read the full framework article for the research, five worked redesigns, and every reference.

Explore the four modes

AI-Restricted

The human act is the point

AI is excluded because the assignment measures the student's independent cognitive performance.

Say to studentsThis task is done without AI because the reasoning is the skill I am assessing, and I need to see yours.

Downloads

Slides, worksheet, and the full package

The full editable deck, a print-ready PDF, and the one-page job aid with the seven-question redesign worksheet.

PowerPoint Webinar slides 46 editable slides with Morph transitions, ready to present or adapt. Download .pptx · 4.5 MB PDF Slides for reading The same 46 slides, readable on any device. Download .pdf · 2.7 MB PDF · one page Job aid and worksheet Four modes, six principles, eight dimensions, GOALS, the seven redesign questions, and an implementation checklist. Download .pdf · 148 KB ZIP All three files Slides in both formats plus the job aid, in one download. Download .zip · 6.0 MB

The question behind the framework

Does your assignment still measure what it claims to measure when a student uses AI?
94%

of AI-generated exam submissions went undetected by markers in a blind university study.

+½

grade boundary higher than real student work, on average.

Scarfe, Watcham, Clarke, & Roesch (2024). PLOS ONE, 19(6), e0305354.

About the session

Design assignments that hold up in the age of generative AI

A 60-minute webinar, including Q&A, on ARAD 2.0. You leave with a redesign of one of your own assignments in note form.

4modes decide the role of AI
6principles filter every design
8dimensions evaluate the evidence
5GOALS steps assemble the redesign

What you'll learn

  1. The four-mode AI continuum for assignment design
  2. Six research-backed principles for preserving human learning
  3. Applying the eight-dimension evaluation rubric
  4. The GOALS process for redesigning any assignment

Four modes · mode finder

Which mode fits your assignment?

The four modes have equal standing, and the learning goal decides among them. Answer up to three questions about one assignment.

Pen-and-ink continuum from less AI to more AI: AI-Restricted, a student writing alone; AI-Limited, a robot offering small help at the page; AI-Collaborative, student and robot reviewing work together; AI-Required, a nurse verifying an AI charting screen against a reference book.
Question 1 of 3

Can students reach the learning target while using AI?

Question 2 of 3

Does AI replace the core thinking the assignment teaches, or support it?

"Replace" means AI could do the exact cognitive work you are trying to build.

Question 3 of 3

Is skilled use of AI itself the learning target, because the field already works this way?

    Your mode

    Start with question 1

    Begin with the human learning target for one assignment, then answer the questions on the left.

      Always: every major course outcome needs at least one AI-Restricted Transfer Gap Check.

      Six principles

      Run every assignment through all six filters

      Each principle is a design commitment while you build and a quality filter while you check. If an assignment fails any one, redesign that part.

      Cognitive Agency

      Who is doing the thinking at each stage?

      The student stays the primary thinker, decision-maker, and judge. List the stages and write the thinker beside each one.

      Learning Integrity

      Does it still measure what it claims if a student uses AI?

      Run the assignment through current models and grade the output on your rubric. If it earns an 87, record the 87.

      Process Visibility

      Can I see how the student got there?

      Require evidence of how the student thought. One well-chosen artifact, such as a baseline memo or prompt log, usually transforms an assignment.

      Evaluative Judgment

      Must the student judge AI output against course standards?

      Verification with teeth: correct at least one AI claim, citing a course reading by page or section.

      Productive Struggle

      Has AI removed the difficulty that makes learning stick?

      AI may absorb extraneous load like formatting and mechanics. Reasoning, analysis, and synthesis stay with the student.

      Equity and Access

      Can every student do this fairly?

      Provide access, teach the tool first, offer an opt-out pathway, and keep AI detection out of grading.

      Eight dimensions

      A rubric that scores the student's thinking

      Choose the three or four dimensions your learning goal needs. Every Exemplary descriptor demands evidence that is hard to fabricate with AI. Open a dimension to see its anchors.

      Performance levels: Beginning → Developing → Proficient → Exemplary

      The Transfer Gap Check

      Did the student learn, or did the tool perform?

      Every major course outcome is assessed at least once under AI-Restricted conditions, on the same rubric dimensions as the assisted task. The difference between the two scores is the finding.

      Pen-and-ink drawing: a student working at a laptop with a small robot, and the same student writing alone in a notebook beside a stack of books.

      Gap of one level or less: the learning transferred, and the student holds the capability independently.

      Gap greater than one level: the learning is tool-dependent. Redesign the upstream task toward a more restricted mode, or raise the weight of the restricted checkpoint.

      The GOALS process

      Five steps, five outputs you can hold

      Once you have done it before, one assignment takes 15 to 30 minutes from target to rubric.

      G

      Goal

      Define the human learning target. What must the student know, do, or judge?

      OutputA one-sentence target statement
      O

      Openness

      Choose the AI mode with four questions, in order, and share the reason with students.

      OutputA mode plus a rationale
      A

      Architecture

      Design for visible thinking: a formative phase, a summative phase, and one of four patterns.

      OutputA phased task sequence
      L

      Look for Evidence

      Align the rubric. A redesigned task graded on the old rubric reverts to the old task.

      Output3 or 4 dimensions with descriptors
      S

      Sustain

      Scaffold AI literacy, study the results, and re-run the assignment through current models every year.

      OutputA literacy plan and a review date

      ↺ Annual review sends S back to G.Patterns: AI Sandwich · Verification Challenge · Comparative Analysis · Generative Challenge

      Keep going

      Practice with the ARAD assistants

      Bring one of your own assignments, or your sprint notes from the webinar. Both assistants run in ChatGPT.

      Custom GPT · by Tim Mousel AI-Responsive Assignment Design (ARAD) Framework Bring one assignment and it walks the GOALS steps with you: goal, mode, architecture, rubric, and sustainment. Custom GPT Create AI-Resistant or AI-Dependent Assignments Purpose-built for the two ends of the spectrum: designs that exclude AI, and designs built around required AI use.
      Chalk drawing of a student thinking through connected ideas toward a checkmark, with a notebook, laptop, and books on the desk.

      Read the framework

      ARAD 2.0: AI-Responsive Assignment Design for Human Learning

      The complete article: theory, four modes, six principles, the rubric, GOALS, five worked redesigns, diagnostics, the programmatic layer, and the full reference list.

      Read the full article →