← All work

Dartmouth · Learning Design and Innovation · Capstone

From Canvas Classroom to Coursera

How do you give feedback when there is no instructor? I’m designing Module 2 of a self-paced Coursera course on implementation science, where learners propose an intervention, check it with AI, and get judged by peers, not the machine.

  • Instructional design
  • AI-supported feedback
  • Live prototypes

Module 2 overview

Module 2 overview page with course navigation, a captioned welcome video, and progress and help panels

AI review

Example AI review with a verdict, scores for six criteria, challenges and next steps

The problem

A Canvas course built on live sessions and instructor grading has to work for thousands of self-paced learners with no instructor.

My focus

The first high-stakes decision: choosing an evidence-based intervention (EBI), where learners most need feedback.

The design

Worked examples, an AI reviewer that challenges first, and peer grading of the learner’s own reasoning.

4objectives

7step journey

2live prototypes

01The challenge

The Dartmouth course runs on Canvas for four weeks, with live Zoom sessions, paid textbooks and instructor grading. None of that works in a self-paced Coursera course with many learners. After benchmarking the Johns Hopkins Introduction to Implementation Science course on Coursera and reviewing research on online learning, three needs stood out.

1

Feedback without an instructor

MOOCs struggle to give feedback on complex, higher-order work (Bates, ch. 5).

2

Media built for online

Engagement drops after about 6 minutes of video (Guo, Kim & Rubin, 2014); choose the medium for the need (Bates, ch. 8).

3

Presence and structure

Learners need to feel the instructor and peers are there, and structure to manage their own learning (Community of Inquiry; UIS).

02Who it’s for

The online course will serve a broader, less prepared and less supported audience than the Canvas master’s cohort, so the design must explain more, show more examples and give feedback without an instructor.

Working health professionals and graduate students worldwide, aged about 25–55, at an intermediate level. Many study on a phone or a slow connection, outside US time zones, around jobs and families. They want to apply implementation science to a real problem at work (Watkins & Marsick, 2020).

How the audience shaped the design

Audience factDesign response
No instructor to catch mistakesAI check, self-check lists, model answers
Some have no project topicA worked sample project to follow
Mixed prior knowledgeStart-here lesson and plain definitions in M1
Busy, global learnersShort items, free readings, captions, plain English
Different pacesPeer review only in M2, M6 and M8
Canvas cohort vs. online learners

Canvas facts come from the syllabus; the online column is an assumption to check.

Canvas cohortOnline learners assumption
WhoMaster’s students in one programProfessionals, researchers, students, global
Prior knowledgeShared baselineFrom none to expert
TimeFixed 4-week scheduleAround a job, start and stop
SupportInstructors, TA, live sessionsNo instructor
MaterialsLibrary and paid textbooksFree and open only
Four likely learner groups

Not yet confirmed by enrollment data. The logic model project serves both practitioners and researchers.

GroupWantsNeeds
Health practitionersMake a program work in their settingPractical steps and tools
Researchers and grad studentsDesign studies, write grantsFrameworks and the logic model
Global health program staffAdapt programs to low-resource settingsNon-US examples, free readings
Prospective degree studentsTry the field before applyingA clear overview
Misconceptions and barriers to design for
  • “Any intervention that worked in one study is an EBI.” They may confuse an EBI, an implementation strategy and a guideline.
  • “If the AI says it’s fine, it’s fine.” Or the opposite: “Using AI is cheating.”
  • “Peer feedback isn’t real feedback.”
  • Varying access to AI tools and different levels of digital literacy.

03What learners will do

By the end of Module 2, learners will be able to:

Propose

an evidence-based intervention for a health problem in their own setting, using a structured 8-part template.

Appraise

the evidence behind it: strength and quality, fit with their setting, and where it sits on the research-to-practice “subway line.”

Use AI critically

telling what the AI drew from their text, what is general knowledge, and what still needs checking, then checking it themselves.

Revise and justify

their EBI choice in writing, and give useful feedback on a peer’s reasoning.

Each objective is a skill a learner could put on a resume, following the “goals as skills” idea from the Johns Hopkins benchmark and competency-based education (Krauss, 2017).

04Design decisions

Collins & Kapur

Scaffold the hard decision

Worked examples model the task, the template breaks it into steps, the AI coaches, and a reflection log asks learners to explain and look back.

Rigolizzo · Kegan & Lahey

Make thinking safe to examine

The AI is told to “challenge first, then strengths,” pushing learners to question why they chose their EBI, and making growth the norm.

Deloitte 2026

Keep the human judgment

AI appears only in ungraded steps. Peers grade the learner’s own reasoning, never the AI’s output.

05The learner journey

Propose → Check with AI → Peer review, in seven steps. Each step uses one cognitive apprenticeship method. Click a step, or use the arrows.

Step 1: Module page

Step 1 of 7

Module page

The two new items sit after the lessons and before the quiz and the peer-graded submission, so learners see the whole path first.

Cognitive apprenticeshipSequencing · global before local

06Try the prototypes

Both run live. In the module overview, press “Show design rationale” to see the reading behind each choice. In the AI check, fill in the boxes and watch the reviewer prompt build itself.

Open full screen ↗
Design rationale for the overview page, by section
Section of the pageDesign choiceReading
Welcome video + transcriptShort and personal; the instructor tells a story about their own first mistakeCOI teaching and emotional presence; Guo et al. 2014
“Why this module matters”Course why: the gap between evidence and practiceCOI existential presence
ObjectivesVerb-first skills, each tied to an itemKrauss 2017 (CBE)
Item list (Learn → Do → Show)Videos cut to 5–6 min; peer discussion; no-AI optionGuo et al.; Bates ch. 8; Collins & Kapur; Gronseth (UDL)
How the pieces connectLearner’s own project; AI never gradedFalbel; Watkins & Marsick; Deloitte 2026
Plan your weekPlan before the task, look back afterwardsCOI regulatory presence; Rigolizzo 2012
Help + ways to learn panelsCaptions, transcripts, text descriptions, low bandwidth, keyboard accessInstructure a11y (POUR); UIS

Everything Module 2 includes

Module overview page

Purpose, objectives, time per item, a welcome video script and plan-your-week prompts.

Short videos, 6 min or less

“What counts as an EBI?” plus re-cut guest-talk clips, each with a framing question.

Static graphics

Three steps for finding EBIs, the subway-line diagram and a comparison table, with alt text.

“Propose your EBI” template

Eight numbered boxes, each with a worked example.

“Check with AI” tool

Builds the reviewer prompt from the learner’s answers and copies it to any AI.

Peer-graded submission

A 4-question reflection log, a rubric and calibration examples.

Plan-and-review prompts

“How will you find 3–5 articles?” before the task, “Did your plan work?” after.

07What’s next

Before the gallery walk

  • Finalize objectives, item list and time estimates; write the welcome video script.
  • Revise the AI check: keyboard access, labels that don’t rely on colour alone, a verification checklist.
  • Build graphics with alt text; write the rubric and two calibration examples.
  • Think-aloud test with 2–3 users, then revise.
  • Run the gallery-walk station and send the module to the Dartmouth team.

Assumptions to check

  • Ask Coursera or edX for learner demographics from similar courses.
  • Survey or pilot one module with 5 to 10 target learners.

Open questions

  • Equity: a non-AI self-check with the same 6 criteria for learners without a good AI account.
  • Open decisions: 8 or 7 modules; whether Coursera can host the tool; video rights; certificate model.

Long term, this becomes the full 8-module Coursera course, built as one project chain from defining implementation science to the capstone logic model.

Read the reflection: Who Gives Feedback When No One Is There? →

Want to hear more?

I’m happy to walk through any of this work in detail.

Get in touch