Files
gh-lyndonkl-claude/skills/cognitive-design/resources/frameworks.md
2025-11-30 08:38:26 +08:00

18 KiB

Design Frameworks

This resource provides three systematic frameworks for structuring cognitive design thinking and decision-making.

Frameworks covered:

  1. Cognitive Design Pyramid (hierarchical quality assessment)
  2. Design Feedback Loop (interaction cycles)
  3. Three-Layer Visualization Model (data communication fidelity)

Why Use Frameworks

WHY This Matters

Frameworks provide:

  • Systematic structure for design thinking (not ad-hoc)
  • Complete coverage across multiple dimensions
  • Prioritization guidance (what to fix first)
  • Shared vocabulary for team communication
  • Repeatable process applicable across projects

Mental model: Like architectural blueprints - frameworks give structure to design decisions so nothing critical is forgotten.

Without frameworks: inconsistent application of principles, missed dimensions, unclear priorities, reinventing approach each time.


##What You'll Learn

Three complementary frameworks, each suited for different contexts:

Cognitive Design Pyramid: When you need comprehensive multi-dimensional quality assessment

Design Feedback Loop: When you're designing interactive interfaces and need to support user perception-action cycles

Three-Layer Visualization Model: When you're creating data visualizations and need to ensure fidelity from data through interpretation


Why Cognitive Design Pyramid

WHY This Matters

Core insight: Design effectiveness depends on satisfying multiple needs in hierarchical sequence - perceptual clarity is foundation, enabling cognitive coherence, which enables emotional engagement, which enables behavioral outcomes.

Key principle: Lower levels are prerequisites for higher levels:

  • If users can't perceive elements clearly → coherence impossible
  • If design isn't coherent → engagement suffers
  • If not engaging → behavior change fails

Mental model: Like Maslow's hierarchy of needs - must satisfy foundational needs before higher-level needs matter.

Use when:

  • Comprehensive design quality check needed
  • Diagnosing what level is failing
  • Prioritizing improvements (fix foundation first)
  • Evaluating entire user experience holistically

WHAT to Apply

The Pyramid (4 Tiers)

        ┌─────────────────────────┐
        │ BEHAVIORAL ALIGNMENT    │ ← Peak: Guides actual user behavior
        ├─────────────────────────┤
        │ EMOTIONAL ENGAGEMENT    │ ← Higher: Motivates and doesn't frustrate
        ├─────────────────────────┤
        │ COGNITIVE COHERENCE     │ ← Middle: Makes logical sense
        ├─────────────────────────┤
        │ PERCEPTUAL EFFICIENCY   │ ← Base: Seen and registered correctly
        └─────────────────────────┘

Tier 1: Perceptual Efficiency (Foundation)

Goal: Users can see and register all necessary elements clearly.

Checkpoints:

  • Sufficient contrast for all text and key elements (WCAG AA minimum: 4.5:1 for body text, 3:1 for large text)
  • Visual hierarchy obvious (squint test - primary elements still visible when blurred)
  • No overwhelming clutter or visual noise (data-ink ratio high)
  • Typography legible (appropriate size, line height, line length)
  • Color distinguishable (not relying solely on hue for colorblind users)
  • Grouping perceivable (Gestalt principles applied)

Common failures:

  • Low contrast text (gray on light gray)
  • Everything same visual weight (no hierarchy)
  • Chartjunk obscuring data
  • Text too small or lines too long

Fix priority: HIGHEST - without perception, nothing else works

Example: All metrics gray 12px → ✓ Primary KPIs 36px bold, secondary 18px gray


Tier 2: Cognitive Coherence (Comprehension)

Goal: Information is organized to align with how users think and expect.

Checkpoints:

  • Layout matches mental models and familiar patterns (standard UI conventions)
  • Terminology consistent throughout (same words for same concepts)
  • Navigation/flow logical and predictable (no mystery meat navigation)
  • Memory aids present (breadcrumbs, state indicators, progress bars)
  • Chunking within working memory capacity (≤7 items per group, ideally 4±1)
  • Recognition over recall (show options, don't require memorization)

Common failures:

  • Inconsistent terminology confusing users
  • Non-standard UI patterns requiring re-learning
  • Hidden navigation or state
  • Unchunked long lists overwhelming memory

Fix priority: HIGH - enables users to understand and navigate

Example: 30 ungrouped fields, inconsistent labels → ✓ 4-step wizard, grouped fields, consistent terms, progress visible


Tier 3: Emotional Engagement (Motivation)

Goal: Design feels pleasant and motivating, not frustrating or anxiety-inducing.

Checkpoints:

  • Aesthetic quality appropriate for context (professional/friendly/serious)
  • Design feels pleasant to use (not frustrating)
  • Anxiety reduced (progress shown, undo available, clear next steps)
  • Tone matches user emotional state (encouraging for learning, calm for high-stress tasks)
  • Accomplishment visible (checkmarks, confirmations, completed states)

Common failures:

  • Ugly or amateurish appearance reducing trust
  • Frustrating interactions causing stress
  • No reassurance in multi-step processes
  • Inappropriate tone (playful for serious tasks)

Fix priority: MEDIUM - affects engagement and trust, but only after foundation solid

Why it matters: Emotional state influences cognitive performance - pleasant affect improves problem-solving, stress narrows attention

Example: Dense text, no progress, no encouragement → ✓ Friendly tone, progress bar, checkmarks, "You're almost done!"


Tier 4: Behavioral Alignment (Action)

Goal: Design guides users toward desired behaviors and outcomes.

Checkpoints:

  • Calls-to-action clear and prominent (primary action obvious)
  • Visual emphasis on actionable items (buttons stand out)
  • Key takeaways highlighted (not buried in text)
  • Ethical nudges toward good decisions (defaults favor user)
  • Success paths more accessible than failure paths

Common failures:

  • Primary CTA not visually prominent
  • Actionable insights buried in data
  • Destructive actions too easy (no confirmation)
  • Defaults favor business over user

Fix priority: MEDIUM-LOW - optimize after foundation, coherence, engagement work

Example: Insights in footnotes, equal button sizes → ✓ Key insight top (large), "Take Action" prominent green, export secondary gray


Applying the Pyramid

Process:

  1. Assess bottom-up: Evaluate T1 (see clearly?) → T2 (makes sense?) → T3 (pleasant?) → T4 (guides action?)
  2. Identify weakest tier: Where are most failures? Which blocks success?
  3. Prioritize foundation-up: Fix T1 first (enables everything), then T2, then T3/T4
  4. Re-evaluate: Check fixes don't break higher tiers, iterate until all pass

Why Design Feedback Loop

WHY This Matters

Core insight: Users don't passively consume interfaces - they actively engage in continuous perception-action-learning cycles.

Loop stages: Perceive → Interpret → Decide → Act → Learn → (loop repeats)

Key principle: Design must support every stage; break anywhere causes confusion or failure.

Mental model: Like a conversation - user asks (via perception), interface answers (via display), user responds (via action), interface confirms (via feedback), understanding updates (learning).

Use when:

  • Designing interactive interfaces (apps, dashboards, tools)
  • Ensuring each screen answers user's questions
  • Providing appropriate feedback for actions
  • Diagnosing where interaction breaks down

WHAT to Apply

The Loop (5 Stages)

┌──────────┐
│ PERCEIVE │ → "What am I seeing?"
└────┬─────┘
     ↓
┌──────────┐
│INTERPRET │ → "What does this mean?"
└────┬─────┘
     ↓
┌──────────┐
│  DECIDE  │ → "What can I do next?"
└────┬─────┘
     ↓
┌──────────┐
│   ACT    │ → "How do I do it?"
└────┬─────┘
     ↓
┌──────────┐
│  LEARN   │ → "What happened? Was it successful?"
└────┬─────┘
     ↓
 (repeat)

Stage 1: Perceive

User question: "What am I seeing?"

Design must provide:

  • Visibility of system status (current page, active state, what's happening)
  • Clear visual hierarchy (where to look first)
  • Salient critical elements (preattentive cues for important items)

Common failures:

  • Hidden state (user doesn't know where they are)
  • No visual hierarchy (don't know what's important)
  • Loading without indicator (appears broken)

Example: No active filter indication → ✓ Active filters shown as visible chips at top


Stage 2: Interpret

User question: "What does this mean?"

Design must provide:

  • Clear labels and context (explain what user is seeing)
  • Familiar terminology (or definitions for specialized terms)
  • Adequate context (why am I here, what are these options)
  • Visual encoding that matches meaning (charts appropriate for data type)

Common failures:

  • Jargon or abbreviations without explanation
  • Missing context (chart without title/axes labels)
  • Unclear purpose of page/screen

Example: No title, axes "X"/"Y", no units → ✓ "Q4 Sales by Region (thousands USD)", labeled axes, annotated events


Stage 3: Decide

User question: "What can I do next?"

Design must provide:

  • Available actions obvious (clear CTAs)
  • Choices not overwhelming (Hick's Law - limit options)
  • Recommended/default option suggested when appropriate
  • Consequences of choices clear (especially for destructive actions)

Common failures:

  • Too many options causing paralysis
  • No guidance on what to do next
  • Unclear consequences ("Are you sure?" without explaining what happens)

Example: 10 unclear buttons → ✓ Primary "Continue" prominent, secondary "Save Draft" gray, "Cancel" text link


Stage 4: Act

User question: "How do I do it?"

Design must provide:

  • Affordances clear (buttons look pressable, sliders look draggable)
  • Controls accessible (appropriate size/position per Fitts's Law)
  • Keyboard shortcuts available for power users
  • Constraints prevent invalid actions (disabled states, input masking)

Common failures:

  • Unclear what's clickable (flat design with no affordances)
  • Tiny touch targets on mobile
  • No keyboard access
  • Allowing invalid inputs

Example: Flat text, no interactive cue → ✓ Raised appearance, hover state, cursor→pointer, focus ring


Stage 5: Learn

User question: "What happened? Was it successful?"

Design must provide:

  • Immediate feedback for every action (< 100ms for responsiveness perception)
  • Confirmations for successful actions (checkmarks, "Saved" message, state change)
  • Clear error messages in context (next to problem, plain language)
  • Updated system state visible (if filter applied, data updates)

Common failures:

  • No feedback (user clicks button, nothing visible happens)
  • Delayed feedback (loading without indication)
  • Generic errors ("Error occurred" without explanation)
  • Feedback hidden or dismissible before user sees it

Example: Click → nothing visible → page maybe changes → ✓ Click → spinner → success message → confirmed transition


Applying the Feedback Loop

For each screen/interaction, ask:

  1. Perceive: Can user see current state and what's important?
  2. Interpret: Will user understand what they're seeing?
  3. Decide: Are next actions clear and not overwhelming?
  4. Act: Are controls obvious and accessible?
  5. Learn: Will user get immediate, clear feedback?

Example: Login form

  • Perceive: Fields visible, labels clear
  • Interpret: "Log in to your account" heading
  • Decide: "Log in" button obvious, "Forgot password?" link
  • Act: Focus states, clickable button, Enter submits
  • Learn: Spinner on submit, success→redirect, error→message next to field with red border

Why Three-Layer Visualization Model

WHY This Matters

Core insight: Effective data visualization requires success at three distinct layers - accurate data, appropriate visual encoding, and correct user interpretation. Failure at any layer breaks the entire communication chain.

Layers: Data → Visual Encoding → Cognitive Interpretation

Key principle: You can have perfect data with wrong chart type (encoding failure) or perfect chart with user misunderstanding (interpretation failure). All three must succeed.

Mental model: Like a telephone game - message (data) must be transmitted (encoding) and received (interpretation) accurately or communication fails.

Use when:

  • Creating any data visualization
  • Diagnosing why visualization isn't working
  • Choosing chart types
  • Validating user understanding

WHAT to Apply

Layer 1: Data

Question: Is the underlying data accurate, complete, and relevant?

Checkpoints:

  • Data quality verified (no errors, outliers investigated)
  • Complete dataset (not cherry-picked subset that misleads)
  • Relevant to question being answered (not tangential data)
  • Appropriate aggregation/granularity (not hiding or overwhelming with detail)
  • Time period representative (not artificially truncated to show desired trend)

Common failures:

  • Garbage data → garbage visualization
  • Cherry-picked dates hiding broader context
  • Outliers distorting scale
  • Wrong metric for question

Fix: Validate data quality before designing visualization

Example: "Are sales improving?" Only last 3 months (hides 2-year decline) → ✓ 2-year trend with annotation: "Recent uptick after sustained decline"


Layer 2: Visual Encoding

Question: Are visualization choices appropriate for the data type, user task, and perceptual capabilities?

Checkpoints:

  • Chart type matches task (compare → bar, trend → line, distribution → histogram)
  • Encoding matches perceptual hierarchy (position > angle > area)
  • Axes scaled appropriately (start at zero for bars, or clearly note truncation)
  • Color usage correct (hue for categories, lightness for quantities)
  • Labels clear and sufficient (title, axes, units, legend if needed)

Common failures:

  • Pie chart when bar chart would be clearer
  • Truncated axis exaggerating differences
  • Rainbow color scale for quantitative data
  • Missing units or context

Fix: Match encoding to task using Cleveland & McGill hierarchy

Example: Compare 6 regional sales. Pie chart (poor angle/area) → ✓ Bar chart (precise position/length)


Layer 3: Cognitive Interpretation

Question: Will users correctly understand the message and draw valid conclusions?

Checkpoints:

  • Main insight obvious or annotated (don't require users to discover it)
  • Context provided (baselines, comparisons, historical trends)
  • Audience knowledge level accommodated (annotations for novices)
  • Potential misinterpretations prevented (annotations clarifying what NOT to conclude)
  • Self-contained (doesn't require remembering distant information)

Common failures:

  • Heap of data without guidance to key insight
  • Missing context (percentage without denominator, comparison without baseline)
  • Assumes expert knowledge novices lack
  • Spurious correlation without clarification

Fix: Add titles, annotations, context; test with target users

Example: Ice cream sales vs drowning deaths. No annotation (viewers assume causation) → ✓ "Both increase in summer (common cause), not causally related"


Applying the Three-Layer Model

Process:

  1. Validate data: Check quality, completeness, relevance, outliers, time period, aggregation
  2. Choose encoding: Identify task → select chart type matching task + perceptual hierarchy → design axes, colors, labels → maximize data-ink
  3. Support interpretation: Add title conveying message → annotate insights → provide context → test with users → clarify misinterpretations
  4. Iterate: Fix weak layers, re-validate

Example: Sales dashboard

  • Data: Complete 2-year sales, verified quality, identified 2 outlier months (note in viz)
  • Encoding: Task = trend + regional comparison. Line chart, distinct hue per region (limit 5), y-axis at zero "Sales (thousands USD)", time on x-axis
  • Interpretation: Title "Regional Sales Trends 2023-2024: Overall Growth with West Leading", annotate outliers ("Holiday promo Nov 2023", "Launch June 2024"), show previous year dotted baseline, tested with sales team
  • Result: Accurate data + appropriate encoding + correct interpretation = insight

Choosing the Right Framework

Use Cognitive Design Pyramid when:

  • Comprehensive multi-dimensional quality assessment needed
  • Diagnosing which aspect of design is failing
  • Prioritizing fixes (foundation → higher tiers)
  • Evaluating entire user experience

Use Design Feedback Loop when:

  • Designing interactive interfaces
  • Ensuring each screen supports user questions
  • Providing appropriate feedback
  • Diagnosing where interaction breaks down

Use Three-Layer Model when:

  • Creating data visualizations
  • Choosing chart types
  • Validating data quality through interpretation
  • Diagnosing visualization failures

Use multiple frameworks together for complete coverage