Most “learning paths” are playlists with better typography.
A stack of modules. A progress bar. A badge at the end. You can finish the path and still not know whether you can do the work under pressure—because the product never asked you to.
A real journey has verbs.
Map the goal → simulate the moment → adapt from feedback → prove the trail.
That is the Learning Journeys spine in ProfilEd. Learner acts first. AI scaffolds second. If the model answers before you attempt, you are not on a journey. You are watching someone else’s homework.
Map — find yourself on a continuum
Start on the Curriculum Graph.
A map is not a course catalogue. Learning progressions are research-based continuums of how people develop deeper, broader understanding over time—they articulate a hypothesis about how learning typically moves (Hess, 2012). The Graph is that idea as product: position, prerequisites, next node.
Pick a goal before any AI. Target role conversation, skill family with an outcome, problem type you keep failing. Vague destinations produce vague practice.
If you skip Map, every Session is tourism.
Simulate — rehearse under constraints
Then practise like the work will feel.
Study Sessions, Story Quests, and Socratic AI Interviews are the Simulate stage—deliberate practice under stated conditions, not content consumption. Ericsson and Harwell (2019) restate the old criteria clearly: tasks you can perform, “immediate informative feedback,” and repeated similar attempts. Simulation-based practice with those conditions outperforms exposure-only training for skill goals (McGaghie et al., 2011). Map the medical research carefully: we are talking professional judgement and career performance, not clinical procedures—but the mechanism travels.
| Stage | Surface | You do first | System role |
|---|---|---|---|
| Map | Curriculum Graph | Commit to a goal / node | Show progression & prerequisites |
| Simulate | Sessions · Quests · Interviews | Attempt under constraints | Scenario, probe, immediate feedback |
| Adapt | Feedback → next choice | Interpret the gap; choose next attempt | Scaffold, re-prompt, suggest—not decide |
| Prove | Shared Footprint | Review trail; optional share | Persist attempts + feedback deltas |
Sim-only coaches that stop at a session score leave you with a number and nowhere to put it. Badge pathways that only stack completions never required an attempt. Simulate has to feed the next two verbs.
Adapt — change the next attempt (you still decide)
Feedback that does not change Friday’s practice is decoration.
Black and Wiliam’s line still holds: assessment becomes formative when the evidence is actually used to adapt teaching and learning to meet the need. Goal, present position, way to close the gap—together. A score dump is not Adapt.
Here is where most “AI adaptive learning” goes wrong. Optimization without agency feels helpful until the learner can no longer diagnose their own gap. Human–AI Symbiotic Theory names Agency-Preserving Scaffolding: AI as development partner, not replacement; co-regulation that keeps decision authority with the human (HAIST, 2026). Related co-agency work is blunt: decision-making remains a human endeavor; humans stay in the loop at assessment (HACC-E, 2026).
In ProfilEd terms: the system may scaffold, re-prompt, and suggest. You choose the next node, the harder Quest, or the Interview focus. Learner action before AI is not branding. It is how Adapt stays learning instead of autopilot.
Prove — leave a trail someone can review
Mastery paths should be outcome-bound, not seat-time-bound. Bloom’s mastery cycle is formative check → corrective practice → retry (cite the mechanism; soft-pedal folklore about effect-size magnitude). Advancement on the Graph should follow demonstrated standard.
Prove is the durable residue of that cycle: attempts, feedback that mattered, reflection, next action—held in Shared Footprint. Not a completion certificate. Not an awareness heatmap. A reviewable journey trail a mentor or career team can skim.
For how the practice→feedback→progress loop closes inside the footprint, see the Evidence Footprint guide. For what belongs in a career evidence record, see the checklist. For why galleries lose context, see From Portfolio to Proof. This piece owns the itinerary—the four verbs—not those theses.
Anti-patterns (kill these early)
- Completion theatre — finished the playlist, still cannot perform
- AI answers first — no attempt, no evidence
- Badge stacks without practice — icons without a trail
- Dashboards without next action — awareness that never adapts
- Adaptation that removes diagnosis — the model decides; the learner forgets how to think
Walk one cycle this week
- Map — one Graph goal, one active node.
- Simulate — one Session or Quest or Interview under a real constraint.
- Adapt — write the next attempt before you close the tab.
- Prove — open Shared Footprint; check that something moved.
If step 4 is empty, you consumed. You did not travel.
Running this with a cohort or campus team?
Sources (research brief only)
- Hess (2012) — learning progressions
- Ericsson & Harwell (2019) — deliberate practice
- McGaghie et al. (2011) — simulation + DP
- Black & Wiliam (1998) — formative assessment
- Bloom (1984) — mastery formative→corrective→retest (mechanism)
- HAIST (2026) — Agency-Preserving Scaffolding
- HACC-E (2026) — co-agency; human decision at assessment
Ops note
- Category flagship:
learning-journeys(R4). - CTA:
/welcome+/campaigns/pilotper Content Ops GO. - CMS: confirm insights vs tutorials for slug.
- No fake metrics; soft-pedaled Bloom 2σ.
Map → Simulate → Adapt → Prove
Most “learning paths” are playlists with better typography.
A stack of modules. A progress bar. A badge at the end. You can finish the path and still not know whether you can do the work under pressure—because the product never asked you to.
A real journey has verbs.
Map the goal → simulate the moment → adapt from feedback → prove the trail.
That is the Learning Journeys spine in ProfilEd. Learner acts first. AI scaffolds second. If the model answers before you attempt, you are not on a journey. You are watching someone else’s homework.
Map — find yourself on a continuum
Start on the Curriculum Graph.
A map is not a course catalogue. Learning progressions are research-based continuums of how people develop deeper, broader understanding over time—they articulate a hypothesis about how learning typically moves (Hess, 2012). The Graph is that idea as product: position, prerequisites, next node.
Pick a goal before any AI. Target role conversation, skill family with an outcome, problem type you keep failing. Vague destinations produce vague practice.
If you skip Map, every Session is tourism.
Simulate — rehearse under constraints
Then practise like the work will feel.
Study Sessions, Story Quests, and Socratic AI Interviews are the Simulate stage—deliberate practice under stated conditions, not content consumption. Ericsson and Harwell (2019) restate the old criteria clearly: tasks you can perform, “immediate informative feedback,” and repeated similar attempts. Simulation-based practice with those conditions outperforms exposure-only training for skill goals (McGaghie et al., 2011). Map the medical research carefully: we are talking professional judgement and career performance, not clinical procedures—but the mechanism travels.
| Stage | Surface | You do first | System role |
|---|---|---|---|
| Map | Curriculum Graph | Commit to a goal / node | Show progression & prerequisites |
| Simulate | Sessions · Quests · Interviews | Attempt under constraints | Scenario, probe, immediate feedback |
| Adapt | Feedback → next choice | Interpret the gap; choose next attempt | Scaffold, re-prompt, suggest—not decide |
| Prove | Shared Footprint | Review trail; optional share | Persist attempts + feedback deltas |
Sim-only coaches that stop at a session score leave you with a number and nowhere to put it. Badge pathways that only stack completions never required an attempt. Simulate has to feed the next two verbs.
Adapt — change the next attempt (you still decide)
Feedback that does not change Friday’s practice is decoration.
Black and Wiliam’s line still holds: assessment becomes formative when the evidence is actually used to adapt teaching and learning to meet the need. Goal, present position, way to close the gap—together. A score dump is not Adapt.
Here is where most “AI adaptive learning” goes wrong. Optimization without agency feels helpful until the learner can no longer diagnose their own gap. Human–AI Symbiotic Theory names Agency-Preserving Scaffolding: AI as development partner, not replacement; co-regulation that keeps decision authority with the human (HAIST, 2026). Related co-agency work is blunt: decision-making remains a human endeavor; humans stay in the loop at assessment (HACC-E, 2026).
In ProfilEd terms: the system may scaffold, re-prompt, and suggest. You choose the next node, the harder Quest, or the Interview focus. Learner action before AI is not branding. It is how Adapt stays learning instead of autopilot.
Prove — leave a trail someone can review
Mastery paths should be outcome-bound, not seat-time-bound. Bloom’s mastery cycle is formative check → corrective practice → retry (cite the mechanism; soft-pedal folklore about effect-size magnitude). Advancement on the Graph should follow demonstrated standard.
Prove is the durable residue of that cycle: attempts, feedback that mattered, reflection, next action—held in Shared Footprint. Not a completion certificate. Not an awareness heatmap. A reviewable journey trail a mentor or career team can skim.
For how the practice→feedback→progress loop closes inside the footprint, see the Evidence Footprint guide. For what belongs in a career evidence record, see the checklist. For why galleries lose context, see From Portfolio to Proof. This piece owns the itinerary—the four verbs—not those theses.
Anti-patterns (kill these early)
- Completion theatre — finished the playlist, still cannot perform
- AI answers first — no attempt, no evidence
- Badge stacks without practice — icons without a trail
- Dashboards without next action — awareness that never adapts
- Adaptation that removes diagnosis — the model decides; the learner forgets how to think
Walk one cycle this week
- Map — one Graph goal, one active node.
- Simulate — one Session or Quest or Interview under a real constraint.
- Adapt — write the next attempt before you close the tab.
- Prove — open Shared Footprint; check that something moved.
If step 4 is empty, you consumed. You did not travel.
Running this with a cohort or campus team?
Sources (research brief only)
- Hess (2012) — learning progressions
- Ericsson & Harwell (2019) — deliberate practice
- McGaghie et al. (2011) — simulation + DP
- Black & Wiliam (1998) — formative assessment
- Bloom (1984) — mastery formative→corrective→retest (mechanism)
- HAIST (2026) — Agency-Preserving Scaffolding
- HACC-E (2026) — co-agency; human decision at assessment
Ops note
- Category flagship:
learning-journeys(R4). - CTA:
/welcome+/campaigns/pilotper Content Ops GO. - CMS: confirm insights vs tutorials for slug.
- No fake metrics; soft-pedaled Bloom 2σ.