● Case 1 · Student

User-Centered Learning with myKE

A college student facing a difficult topic uses myKE to build a personalized understanding on their own terms — choosing what to explore, at what depth, and in what order.

👩‍💻
Sofia Ramos
2nd-Year Psychology Major · State University
Research Methods Statistics Anxiety Visual Learner Works Part-Time
"I can memorise definitions the night before an exam, but two weeks later I can't explain why any of it matters. I need to actually understand this, not just pass it."
1
The problem: a topic that won't stick

Sofia's Research Methods course has just covered Null Hypothesis Significance Testing (NHST). She passed the quiz by memorising the steps, but her professor's feedback keeps using words like "p-value interpretation," "Type I error trade-offs," and "statistical power" — and she can't connect them.

She has a major paper due in three weeks that requires her to choose and justify a statistical approach. Memorised steps won't get her there.

2
Generating her personal knowledge map

Sofia opens ke.skoonline.org and types: Domain → Psychology, Subdomain → Research Methods, Concept → Null Hypothesis Significance Testing.

Within seconds a map appears. She can immediately see that NHST sits at the centre, with prerequisite nodes like Probability Distributions and Sampling Theory, and enables nodes like Effect Size, Statistical Power, and Confidence Intervals.

🗺️
Concept Map generation — the map makes the structure of the domain visible at a glance. Sofia immediately sees the gaps in her knowledge: she never properly learned Sampling Theory, and that explains why NHST felt arbitrary.
3
Following her own curiosity through the map

She clicks the Sampling Theory prerequisite node. The details panel opens with a plain-language definition and three action buttons. She clicks "Why it matters" — the chat explains how NHST assumes random sampling and what breaks down without it.

She then clicks the requires edge between Statistical Power and Sample Size. myKE explains the relationship: increasing sample size increases power, and why ignoring this leads to underpowered studies that produce misleading results.

🔗
Edge explanations — understanding relationships between concepts (not just the concepts themselves) is what moves Sofia from remembering to understanding on Bloom's scale.
"I always treated p-value and power as two separate things I had to memorise. I didn't realise they were in tension with each other. This changes how I read every paper."
4
Expanding into a sub-map

The node Effect Size keeps appearing in her professor's slides. She right-clicks it and selects Explore in New Tab. A new map generates centred on Effect Size, revealing its relationship to Cohen's d, power analysis, and practical vs. statistical significance.

She now has two tabs open — her original NHST map and the Effect Size deep-dive — and navigates between them as she reads her textbook.

🔭
Node expansion + multi-tab — Sofia controls the depth of her exploration. She isn't following a fixed chapter order; she's following the conceptual threads that matter for her paper.
5
Q2L — practising the skill of asking better questions

Two days before her paper deadline, Sofia uses the Q2L (Question to Learn) mode on the NHST map. myKE generates a scenario: "A clinical trial reports p = 0.048 with a sample of n = 40 and Cohen's d = 0.18. The authors conclude their intervention is effective. A reviewer disputes this. Why might the reviewer be sceptical?"

Sofia selects Statistical Power and Effect Size as her focus concepts, then submits her question: "What does it mean when a result is statistically significant but the effect size is small?"

myKE evaluates it as deep and enters discussion mode — asking her follow-up questions rather than giving the answer directly. By the end of the exchange she has articulated, in her own words, the difference between statistical and practical significance.

🤔
Q2L evaluation — the AI doesn't reward Sofia for asking "What is effect size?" It pushes her toward questions that reveal her own thinking, then guides that thinking without replacing it.
"My question was marked deep on the first try. That felt different from getting a multiple-choice answer right. I actually had to think."
6
Wonderment — confronting what she doesn't know she doesn't know

The next day Sofia activates Wonderment on the Statistical Power node. Three AI agents simultaneously surface what she hadn't considered: the Empiricist shows a paradox about underpowered studies that correctly detected real effects by accident; the Connector links statistical power to signal-to-noise ratios in engineering; the Provocateur inverts the question — "What would research look like if we designed for maximum power rather than minimum sample size?"

She selects the paradox observation, picks a "reframe" genius question, and composes an answer. When the critics panel arrives, she reads it twice — it challenges her answer directly. The synthesis reframes her entire understanding of why sample size is a research design decision, not a statistical afterthought.

Wonderment — reaches Bloom's Evaluate and Create levels through deliberate epistemic discomfort. The critics panel dwell time is tracked: long dwell = genuine integration of challenge; short = seeking validation.
"I thought I understood power. The paradox example shook that. I had to rewrite two sentences of my paper."
7
When Things Break Down — diagnosing a real research failure

The morning before submission, Sofia tries When Things Break Down on her NHST map. myKE generates a scenario: "A psychology researcher ran 40 subjects, got p = 0.049, published the result. Two years later, three labs failed to replicate it. Why did the original study likely fail?"

Sofia expands three evidence clues (sample demographics, effect size, study design) and asks two investigation questions about what conditions produce false positives. She submits her diagnosis: "The study was underpowered — p < 0.05 was achieved at an n where the true effect size would require n=120. The result was a false positive from sampling noise." The AI scores it as correct with deep reasoning — she identified the mechanism, not just the symptom.

🔧
When Things Break Down — Sofia is no longer describing NHST; she is reasoning about its failure modes. This is the Apply and Analyze Bloom's levels in action: exactly what her paper requires.
"That scenario was my paper. I'd been trying to write the critique section for two days. Ten minutes of troubleshooting gave me the structure."
8
Reviewing her own learning quality

Before submitting her paper, Sofia opens "How Am I Doing?". The dashboard shows her interaction pattern across three sessions: she spent most of her time on nodes in the Apply and Analyze Bloom levels, asked two deep Q2L questions, and her AI Collaboration score is high on Iteration Depth — she followed up multiple times rather than accepting the first response.

It also flags that she never engaged with the Confidence Intervals node despite it appearing in her map. She spends fifteen minutes there before closing the laptop.

📊
Learning analytics dashboard — Sofia can see not just what she studied but how she studied it. The system identifies the gap she would have missed, giving her a chance to fill it before the paper is graded.

✦ What myKE delivered for Sofia

🗺️
She saw the structure of the domain — not just isolated definitions — and immediately identified the foundational gap (Sampling Theory) that had made everything feel arbitrary.
🔍
She explored at her own pace and in her own order, following the thread from NHST → Effect Size → Power Analysis as her paper required, not as the textbook chapter was laid out.
🧠
The Q2L session moved her from Bloom's Understand to Evaluate — she could now critique a published result, not just describe the method used.
📋
The analytics dashboard caught a blind spot (Confidence Intervals) before submission — functioning as a self-assessment tool rather than a grade.
Wonderment forced her to confront a paradox in her understanding of statistical power — prompting a revision of her paper's framing that she wouldn't have found through standard study.
🔧
The WTB diagnostic scenario gave her the failure-mode reasoning she needed to write the critique section of her paper — by reasoning about why NHST fails, not just what it does.

● Case 2 · Faculty

Supporting Self-Regulated Learning with myKE

A faculty member uses myKE's teacher tools to shift pre-class preparation from passive reading to active concept exploration — and uses the resulting analytics to teach more responsively.

👨‍🏫
Prof. David Osei
Associate Professor · Cognitive Science Department
Intro to Cognitive Science 35 Students Mixed Background Flipped Classroom
"Students arrive to class having 'done the reading' but unable to explain the first concept on the slide. I need them to arrive having actually thought — not just scrolled."
1
The problem: shallow pre-class preparation

Prof. Osei runs a flipped classroom for Introduction to Cognitive Science. Students are supposed to engage with material before class so sessions can focus on application and discussion. In practice, most arrive having skimmed — they can name concepts but cannot connect them.

Week 4 topic: Working Memory and Cognitive Load. Last year, 60% of students failed the follow-up quiz on this topic. He wants a different approach this time.

2
Building the curriculum on the Teacher Dashboard

Prof. Osei logs in as Teacher and opens the Curriculum Builder. He seeds it with three core concepts: Working Memory, Cognitive Load Theory, and Dual Coding. He clicks AI Expand.

myKE generates a full concept structure: it adds Baddeley's Model, Phonological Loop, Visuospatial Sketchpad, Central Executive, Intrinsic/Extraneous/Germane Load, and their relationships. Prof. Osei reviews the map, removes two nodes he considers out of scope for Week 4, and rearranges the suggested learning sequence to match his course arc.

🏗️
AI-assisted curriculum expansion — AI provides comprehensive coverage of the domain; the professor exercises pedagogical judgement over scope and sequence. Neither could produce the result alone.
3
Assigning the map to the class

He creates a class called COGS 101 — Week 4 and generates an enrollment code. He posts the code and a two-sentence instruction on the course LMS:

"Before Thursday's class, spend 30–40 minutes exploring the Working Memory map on ke.skoonline.org. Your goal is not to read every node — it is to find the two or three connections you find most surprising and be ready to explain why."

He deliberately does not prescribe a path. That ambiguity is the instruction.

🎓
Class enrollment + structured freedom — students receive a shared starting map but each explores it differently. This surfaces the diversity of prior knowledge that lecture-then-quiz preparation hides.
"I'm not replacing the reading. I'm replacing the passive scroll with something that requires them to make choices — and every choice is a decision about what they don't yet understand."
4
Students explore independently before class

Over the 48 hours before Thursday, 31 of 35 students log in. Their exploration patterns diverge immediately:

  • Students with Psychology backgrounds go deep into Baddeley's Model and its clinical applications.
  • Students from Computer Science gravitate toward the parallel between Working Memory capacity limits and cache architecture — a connection the curriculum didn't explicitly draw.
  • Several students use Q2L mode and generate questions about why Cognitive Load varies by expertise level.
  • Four students expand the Dual Coding node into its own sub-map, discovering connections to multimedia learning theory.
🧭
Self-regulated exploration — each student's path through the map reflects their existing schema, curiosity, and gaps. myKE captures all of this without requiring students to fill out a pre-class survey.
5
Prof. Osei reviews the class analytics before Thursday

Wednesday evening, Prof. Osei opens the Teacher Analytics Dashboard. He can see:

  • Most visited node: Central Executive — but low dwell time, suggesting students clicked it without engaging deeply.
  • Most skipped node: Germane Cognitive Load — only 6 of 31 students visited it.
  • Highest engagement: the edge between Intrinsic Load and Expertise Reversal Effect — 9 students asked the chat to explain this relationship.
  • Q2L activity: 11 students attempted Q2L; 7 achieved a "deep question" rating on the first or second try.

He rewrites the first 20 minutes of Thursday's class. Instead of introducing Germane Load from scratch, he opens with: "Most of you skipped Germane Load on the map. Let's talk about why that node is the most important one there."

📡
Stealth assessment analytics — Prof. Osei walks into class knowing exactly where collective understanding is strong, where it is shallow, and what connections students found surprising — without administering a single pre-test.
"For the first time, I entered class knowing what my students had actually thought about — not what they had claimed to read."
6
Class session: discussion, not delivery

Thursday's class runs differently. Because students arrive with genuine exposure — not just claimed reading — Prof. Osei can open with discussion questions rather than explanation:

"Someone in Q2L asked: 'Why does Cognitive Load decrease as expertise increases — isn't more knowledge more to manage?' Who can answer that using what you found on the map?"

The CS students draw on their cache analogy. The Psychology students connect it to schema automation. A student who had explored Dual Coding connects it to chunking. The class constructs the explanation collectively.

Prof. Osei spends the last 15 minutes on Germane Load — the one node most students skipped — knowing they now have the prerequisite understanding to grasp it.

💬
Analytics-informed teaching — the class session builds on real prior engagement rather than assumed reading. The professor's role shifts from deliverer of content to facilitator of connections students are already forming.
7
Ongoing: monitoring individual self-regulation

Over the following weeks, Prof. Osei uses the Learner Profile view to monitor individual students. He notices two patterns:

High self-regulators (about 40% of the class) revisit nodes after class, expand sub-maps independently, and show increasing Q2L depth scores over time — their questions move from factual to causal across weeks.

Passive consumers (about 25%) visit the map, read the top-level description of each node, and close it. They never use Q2L. Their interaction pattern looks like skimming the textbook, just faster.

He emails the passive-consumer group personally — not to penalise, but to invite office hours. He uses their specific map paths as the conversation starter: "I see you spent time on Working Memory but didn't click the edge to Cognitive Load. What made that connection unclear?"

🔎
Individual learner profiles — the system surfaces which students are actively self-regulating and which are performing the motions of engagement. This gives Prof. Osei specific, evidence-based entry points for intervention — not just attendance records.

✦ What myKE delivered for Prof. Osei

📊
He arrived to class with real data about student thinking — which concepts they explored, which they skipped, and what questions they were forming — without administering a quiz.
🗣️
Class time shifted from content delivery to guided discussion. Students arrived having made choices about what to explore, which meant they arrived with something to say.
🎯
He could identify the specific gap (Germane Load) that needed in-class attention, rather than re-explaining everything from the reading.
🤝
He could reach out to passive-consumer students individually with specific, evidence-based conversation starters — not generic "attend office hours" messages.