Mental Models, Alignment & Opportunity Mapping

Mental Models Are Executable Artifacts, Not Workshop Sketches

Teams often treat mental models as inspirational wallpaper. Somebody runs a workshop, draws towers on a wall, and everybody nods at the idea that users think differently from the org chart. Then the diagram disappears because nobody can attach it to delivery work. ASD fixes that by treating mental-model outputs as artifact contracts rather than facilitation residue.

There are two closely related runtime forms in play. mental-model-diagramming produces a mental-model-map, which is the skyline: towers, spaces, and the durable cognitive structure of a task domain. mental-model-alignment produces a journey-map, because the alignment question is inherently comparative and staged: what the customer thinks is happening, what frontline staff think is happening, where the two diverge, and which touchpoints express that divergence. Those two forms complement each other. One captures structure. The other captures the clash between structures in lived service flow.

The point of keeping both as artifacts is practical. A mental model then stops being a metaphor and starts being a dependency. An opportunity matrix can consume it. A content-model change proposal can cite it. A journey map can express its gaps explicitly. A Critic can veto it when a divergence lacks evidence from both sides.

This is why the language of belief, question, and task matters so much. If a model says, "The platform organizes grades by class," that is system language. If it says, "I need one picture of how my child is doing," that is user logic. The whole alignment exercise depends on preserving that distinction. Otherwise I end up redrawing the product architecture in a softer font and calling it empathy.

The taxonomy discipline is load-bearing. A belief is a claim the user treats as true about how the service works. A question is an information need. A task is an intended action. If I blur those categories, the model turns mushy fast. "The spinner stopped so my work is saved" is a belief. "Did my grade actually publish" is a question. "Resume yesterday's grading" is a task. Those differences are exactly what later content-model and interaction changes must answer.

That is why the research team at Lumen extracted and reviewed mental-model items before they talked about screens. The point was not to make the workshop more poetic. The point was to stop the design team from collapsing cognition, behavior, and interface preference into one bucket called feedback.

The Alignment Grid Makes Divergence Concrete

Once the cognitive demand side is explicit, I need a way to compare it with the service as actually performed. That is the alignment grid. Think of it as a disciplined collision table. One column names the user belief or question. One column names the system truth or frontline interpretation. One column names the touchpoint where the clash becomes visible. One column names the consequence. Evidence sits underneath all of it.

The method does not ask for abstract empathy. It asks for paired claims. Customer quote on one side. Frontline or system quote on the other. Shared stage. If both sides are not present, the divergence is not ready. That is exactly what the mental-model-alignment contract enforces. The Critic's gate instruction is explicit: reject any divergence marked without a quote from both sides.

The Lumen EdTech case shows why this matters. Parents thought by child. Teachers thought by grading sitting. Students thought by due date. The platform, meanwhile, thought by course and reverse chronology. Each actor was rational on their own terms. The failure demand emerged because the service forced all three groups to think in the platform's ontology. The alignment grid made that mismatch visible as service work rather than as user confusion.

{
  "method": {
    "id": "mental-model-alignment",
    "name": "Mental Model Alignment",
    "phase": "synthesis",
    "reviewGate": "critic"
  },
  "steps": [
    {
      "role": "mapper",
      "instruction": "Build two parallel journeys for the same process: the customer's and the frontline staff's, stage by stage."
    },
    {
      "role": "synthesizer",
      "instruction": "Mark every stage where the two mental models diverge, citing the quote from each side."
    },
    {
      "role": "critic",
      "instruction": "Reject any divergence marked without a quote from both sides.",
      "gate": true
    }
  ],
  "acceptance": [
    {
      "rule": "min-nodes",
      "min": 3
    },
    {
      "rule": "min-evidence-per-node",
      "min": 2
    },
    {
      "rule": "field-non-empty",
      "field": "emotion"
    },
    {
      "rule": "field-non-empty",
      "field": "touchpoint"
    }
  ],
  "outputKind": "journey-map"
}

The grid changes the quality of debate. I am no longer hearing vague claims like "parents find it confusing." I am seeing a named divergence, a touchpoint, paired evidence, and a measurable cost. That makes the next step possible.

Opportunity Mapping Needs Evidence Weights, Not Design Charisma

Once misalignments are visible, teams get excited and immediately start ideating. That is understandable. It is also dangerous. Not every gap deserves the same sequence, and not every emotionally resonant gap has the same business gravity. Opportunity mapping is the step that stops the loudest problem from stealing the roadmap by drama alone.

I score opportunities against four questions. First, how strong is the evidence? Second, how severe is the user harm? Third, how much failure demand or delivery risk does the gap generate? Fourth, how reversible is the likely intervention? The score is not meant to pretend objectivity where none exists. It is meant to make trade-offs explicit enough that design, product, engineering, and operations can argue on the same page.

Evidence weight is the anchor. A gap supported by ticket volume, usability failure, interview evidence, and frontline corroboration should outrank a clever but weakly evidenced idea. In Lumen, the by-child parent view, grading-session continuity, and due-date ordering all cleared that bar. A handful of attractive side opportunities did not. They were kept visible but sequenced later.

id,opportunity,evidence_weight,user_harm,delivery_impact,reversibility,total
X,parent-child-aggregation,5,5,5,3,18
Y,grading-session-state,5,5,4,2,16
Z,due-date-primary-feed,4,4,4,4,16
G4,notification-bundling,3,2,3,5,13
G5,teacher-copy-cleanup,2,2,2,5,11

The point is not that every team must use my exact weights. The point is that ASD gives those weights somewhere durable to live. If the opportunity later turns out weaker than expected, I can see whether the original problem was in the evidence, the score, or the execution. Without that memory, prioritization becomes folklore within a quarter.

There is a governance benefit too. When an opportunity score cites evidence ids and an artifact id, leadership can challenge the weighting without denying the existence of the underlying gap. That is a healthier argument. It keeps the fight on sequencing rather than reality itself.

Lumen's strongest opportunities also illustrate the difference between interface tweaks and ontology changes. A child-level summary is not a filter bolted onto a course model. A grading session is not a tooltip explaining autosave. A due-date-primary feed is not a cosmetic sort order if the entire student's planning behavior depends on it. Opportunity mapping, done properly, tells me which gaps demand structural change and which can be handled with lighter service moves.

A Valid method-run Walkthrough for Mental-Model Alignment

The platform contract matters most when I can actually run it. So here is the shape of the workflow using the real CLI surface. I create the project, capture evidence rows from both sides of the relationship, inspect the method contract, write the alignment artifact, send it to the Critic, and only then run the deterministic replay. The commands below are deliberately plain. No pseudo-tool names. No catch-all shortcut command. Just the shipped interface.

asd project-create --name "Lumen Alignment"   --client "Fictional composite — Lumen EdTech"   --sector education   --service "gradebook and parent communication"

asd evidence-add --projectId lumen-alignment-9b3f5d2a   --source parent-12 --kind quote   --body "I just want one picture of how my child is doing, not eleven separate class pages."

asd evidence-add --projectId lumen-alignment-9b3f5d2a   --source teacher-07 --kind quote   --body "When the spinner stops, I assume the grading sitting is safely saved for tomorrow."

asd evidence-add --projectId lumen-alignment-9b3f5d2a   --source support-queue-q1 --kind telemetry   --body "Parent grade-finding tickets 3720 per quarter; grading-draft loss tickets 2640; missed-deadline comprehension tickets 1800."

asd evidence-add --projectId lumen-alignment-9b3f5d2a   --source content-model-v2 --kind document   --body "The platform organizes grade history by course and stores append-only draft changes per assignment column."

asd method-run --methodId mental-model-alignment   --projectId lumen-alignment-9b3f5d2a

asd artifact-write --projectId lumen-alignment-9b3f5d2a   --methodId mental-model-alignment   --title "Parent and frontline alignment journey"   --nodes '[
    {
      "id":"find-grades",
      "label":"Find the latest school picture",
      "fields":{"emotion":"2","touchpoint":"parent portal"},
      "evidence":[
        {"evidenceId":"38ff95bb5223d05c6eeafae8d15819e50c4cb27ba22809da1958be2bb5a465cf"},
        {"evidenceId":"4a24df6d58ec99e275dcc98f0dd65e0d1c64133eaa7f66d89ad1263f23d77e64"}
      ]
    },
    {
      "id":"resume-grading",
      "label":"Resume a grading sitting",
      "fields":{"emotion":"2","touchpoint":"gradebook"},
      "evidence":[
        {"evidenceId":"a24b51bac317fb3baa60f8d5e5cbcde0ab3f62c0487e51321e4191855c0ea9de"},
        {"evidenceId":"4a24df6d58ec99e275dcc98f0dd65e0d1c64133eaa7f66d89ad1263f23d77e64"}
      ]
    },
    {
      "id":"see-what-matters-next",
      "label":"See what is due next",
      "fields":{"emotion":"3","touchpoint":"student feed"},
      "evidence":[
        {"evidenceId":"c872dc90f8078c90d74c7ae71fbd27f545d81ca0092d284211c208b2c901068f"},
        {"evidenceId":"4a24df6d58ec99e275dcc98f0dd65e0d1c64133eaa7f66d89ad1263f23d77e64"}
      ]
    }
  ]'

asd review --projectId lumen-alignment-9b3f5d2a   --artifactId 7bfbea7ad1f34940bd46a6a66c7eaa22   --verdict pass   --rows '[
    {
      "rowKind":"proposal",
      "severity":"medium",
      "rule":"source-diversity",
      "message":"Keep the support-queue telemetry visible alongside the paired quotes on all three divergence nodes.",
      "evidenceId":"c872dc90f8078c90d74c7ae71fbd27f545d81ca0092d284211c208b2c901068f"
    }
  ]'

asd factcheck --projectId lumen-alignment-9b3f5d2a

Read that carefully and a few truths jump out. method-run gives me the contract; it does not write the artifact for me. artifact-write still needs nodes that satisfy the method's acceptance rules. The review step still needs evidence-cited rows, even on pass. And the final check is still project-wide, because the platform cares about the state of the store, not merely the confidence of the last agent who touched it.

The ids in the shell block are illustrative content-addressed examples, included so the walkthrough stays mechanically faithful to the real CLI shape without pretending this chapter is a live workspace receipt. That distinction matters. A tutorial can be illustrative without becoming fake.

Case Spine: Lumen EdTech and the Cost of Ontology Mismatch

Lumen EdTech looked, at first glance, like a classic UX clean-up. Support volume was ugly. Parent trust was low. Teachers complained about lost grading drafts. Students missed deadlines. The executive reflex was to reskin the interface. The alignment work showed something harsher: the service was faithfully implementing the wrong model of the world.

Parents did not think in courses. They thought in children. Teachers did not think in assignment columns. They thought in grading sittings that start, pause, and resume. Students did not think in posting chronology. They thought in urgency. The platform's content model violated all three. That is why tickets kept coming even when individual screens looked competent.

The mental-model work created three durable actor models and then forced a comparison against the system model. That yielded the three major gaps in the previous table plus smaller secondary ones. Opportunity scoring then made the sequencing obvious. The first structural move had to be parent-by-child aggregation because it had the heaviest combination of evidence weight, ticket load, and daily user harm. The teacher draft-state gap followed close behind because it was eroding expert trust. The feed-order gap mattered enough to sit alongside it because students were paying the price in real missed work.

From there the design work became less mystical. The team created a unified Communication Hub rather than three disconnected portals. It introduced child aggregation across courses, explicit grading-session objects with visible save and conflict states, and a due-date-primary student feed. Those are not cosmetic fixes. They are service-architecture responses to named cognitive mismatches.

Measurement closed the loop. Lumen did not declare victory because the new architecture sounded cleaner. The pilot re-ran task-success tests, monitored support-ticket categories, and checked whether first-click paths were converging on the intended actor models. That kept the alignment story empirical. If the new child-level summary still sent parents hunting across class pages, the model-match metric would expose it. If the grading-session object still leaked drafts, teachers would tell us quickly and the ticket queue would confirm it.

That outer loop matters because mental-model work is never a one-shot artifact. Schools change reporting practice. Terms change. Product sprawl creeps back in. A durable alignment capability therefore needs periodic replay, not one triumphant workshop and a framed diagram.

The pilot result mattered because it proved the diagnosis had teeth. Support tickets fell. Parent task success rose sharply. Teachers trusted the gradebook again because saving meant what they thought it meant. Students could see the week ahead in a structure that matched their planning behavior. The service had not simply become prettier. It had become more legible to the people using it.

That is the larger lesson of this chapter. Mental models are not soft context. They are a way of discovering where a service is asking people to think like the provider rather than like themselves. ASD makes that discovery durable by giving it artifacts, evidence, review, and replay. Once I do that, opportunity mapping stops being a brainstorm and becomes a disciplined argument for which structural changes deserve to happen first.

That durability is the real win. The model, the divergence, the score, and the resulting change proposal all stay connected instead of dissolving into meeting memory.

That is how a cognitive diagnosis becomes engineering work, service work, and eventually measurable operating performance rather than a one-off insight artifact.

That traceability is what keeps the walkthrough honest enough to be copied into real practice.

Run the alignment exercise on a cadence, not once. Services drift toward provider logic as back-office constraints change, so the mental model that was accurate at launch quietly stops being accurate two quarters later. I revisit the grid whenever telemetry shows new drop-off patterns, whenever a frontline team invents a workaround, and whenever a stakeholder cannot explain a screen without referring to the org chart. Any of those three signals means the model and the service have started diverging again, and the cheaper moment to catch it is always the earlier one.