HCI 3.0 Key Summary - From Usability to Interface Design

I've condensed and summarized the core concepts of HCI 3.0 based on Introduction to HCI (Jinwoo Kim, Yonsei Univ. HCI Lab & HAII). Designed for quick scanning, focusing on comparison tables, names, and year matchings.

Chapter 1: Understanding HCI 3.0

HCI is the creation and evaluation of Digital Artifacts that people can use usefully, conveniently, and enjoyably. The ultimate goal is to develop high-quality digital products, services, and content that deliver an optimal experience. Herbert A. Simon's quote from The Sciences of the Artificial was also cited: "The proper study of mankind is the observation of how people actually make and use things."

Category	HCI 1.0	HCI 2.0	HCI 3.0
Focus	Fragmented single user–computer interaction	UX of multiple users & social media	Human–AI Interaction = HAII
Interface	Text-based (without GUI)	GUI and mouse	Natural interaction (NLP, voice, gestures)
Features & Scope	Functional aspects, command input, high learning curve, one-to-one	Behavioral & psychological elements like scrolling/empathy (likes), multiple users	Collaboration & value coexistence with AI, individuals, groups, AI agents

Perception–Reasoning–Execution (Sensing–Reasoning–Actuating)

AI-embedded systems evolve from passive to active, capable of learning and acting on their own. As direct user intervention becomes unnecessary, processes move at a faster pace.

Stage	Function	Interaction Point	Example
Sensing (Perception)	Collecting data needed by AI	Accurate data collection + providing specific descriptions to assist AI	Mind Garden: Heart rate variability (HRV) via camera sensors, etc.
Reasoning	Understanding context from data & deriving optimal solutions	Interactive Machine Learning (IML), in an easily understandable format	Interpreting data to provide patterns
Actuating (Execution)	Predicting context to guide user behavior	Impossible to predict all situations → Supplemented by human intervention (e.g., experts)	AlzGuard: Agent 'Sammy', brain training

Digital Health

Composition = Digital Biomarkers (DBM) + Digital Therapeutics (DTx). The key point is that these two cannot exist without artificial intelligence.

Digital Biomarkers: Objective indicators measured and analyzed from physiological and behavioral data obtained through devices like smartphones and wearables, used for diagnosis, monitoring, and personalized treatment.
Digital Therapeutics: Evidence-based software programs designed to prevent, manage, or treat medical conditions or disorders.

3 Types of DTx: Standalone (delivers cognitive behavioral therapy independently), Augmentary (used alongside existing treatments to enhance efficacy), Complementary (complements existing treatments, supporting self-management and lifestyle improvement). Examples include NightWare (learns sleep habits) and Anzeilax (deep learning data refinement).

Memorization: The ultimate goal is the Real Experience - Dewey & Martin, 1934. The 3 fundamental elements are usability, usefulness, and trustworthiness. UX design consists of 4 stages: Analysis – Planning – Design – Evaluation.

Chapter 3: HCI 3.0 and Usefulness

Usefulness: The principle of helping effectively achieve the purpose of use. Focuses on accomplishing the goal even if the process is difficult.
Usability: That the process of use must be efficient - "Even if it's somewhat inconvenient, it's for infection prevention!"

Even the same portal differs depending on the purpose: Google = Searching, Naver = Browsing. By domain: Mobility = Safety, Finance = Security, Healthcare = Safety & Efficacy.

Problem Space ↔ Solution Space

Problem solving = Searching from an initial state to a goal state using operators within path constraints. The problem space is the set of all solvable states, and the solution space is the set of all alternative solutions that can solve the problem.

Problem Space Element	Corresponding Analysis
Initial State (User's current state)	Stakeholder Analysis
Goal State (State to reach)	Task Analysis
Constraints (Environmental impact)	Context Analysis
Operators (Available functions)	Context Analysis

Usefulness = Consistency between the problem space representation and the solution space mental model. A conceptual model is the result of abstracting structure and operation (e.g., a company organizational chart), done for the sake of simplification - as defined by Liddle (1996).

4 Types of Mental Models

Mental Model = A user's Conceptual Model of the system inside their head. A mental knowledge structure that simply abstracts the system's structure and dynamic operation.

Mental Model	Content	Keywords
Value Model	Thoughts on what value it will provide	User Purpose
Structural Model	Describes what detailed elements compose it	Information Architecture
Functional Model	How to use it	User Flow
Representation Model	Visualized through visual information	User Interface

In HCI 3.0, AI is regarded in a "human-like manner" and is easily anthropomorphized. A representative illusion is that "everything should be done automatically on its own" - it is crucial to distinguish between values that can and cannot be provided.

Need = A requirement to improve dissatisfaction ("Eating bread resolves hunger"), Want = A desire for satisfaction beyond the minimum, which is a concept similar to value.

Value	Concept	Example
Utilitarian (Instrumental)	Tool for practical purposes/needs	Pria (Medication guidance robot)
Hedonistic (Experiential)	Emotional satisfaction & enjoyment	EndeavorRx by Akili (Gamified ADHD therapeutic)

Efficacy vs. Environmental Suitability (Suitability)

Category	Efficacy	Suitability
Environment	Controlled experimental environment	Real-world practice (RWP)
Proof	Health benefits proven numerically, safety (risk) proven	Meeting requirements proven via accuracy, specificity, reliability, and scope of use data

Since the two have a trade-off relationship, it is difficult to enhance both simultaneously. Balancing them is the role of an HCI expert.

SaMD = Software as a Medical Device; standalone software used for diagnosis, treatment, or monitoring for independent medical purposes without being dependent on hardware (utilizes standalone/general-purpose platforms). Efficacy classification (KHIDI, 2019): Therapeutic function / Risk detection through monitoring / Convenient chronic disease management / Indirect support for diagnosis and screening. Example: ForMe is a voice-based tablet service for enhancing executive function in children with ADHD.

Chapter 6: Stakeholder Analysis

Stakeholder = A person with a direct or indirect interest in a product or service (clients, users, consumers, project managers, etc.). Analysis is the first step in understanding the 'user's problem space' - the analysis of the initial state. Kirby (1991): In a broad sense, prioritize and analyze at an appropriate level. Procedure: ① Preparation → ② Socio-Technical Model → ③ Persona Model.

Preparation 3 STEPS: ① Identification (patients, doctors, insurance companies, government, etc.) ② Characteristics (primary users = direct interaction, secondary users = those affecting or affected) ③ 4 Types of Impact (clinical, cost, workflow/operational, social value). Cost Analysis: CMA (minimization), CEA (effectiveness), CUA (utility), CBA (benefit). Priorities are Essential / Important / Consider / Reference.

Socio-Technical Model

Since technology is created within an organizational environment, this model considers social and technical factors simultaneously. Methodologies: USTM, CUSTOM, OSTA, ETHICS, SSM.

Group Investigation - Interviews (In-depth / Focus group / Contextual) + Observations (Photo diary, town watching, cultural probes, shadowing)
Root Definition - Beneficiaries, ecosystem changes, acceptance, usage environment
Conceptual Model - Solid line = strong connection, Dotted line = weak connection
Evaluation - Stakeholder verification + Ecosystem verification (the main reason for creation)

Persona

The etymology comes from the mask in Greek theater, and Carl Jung defined it as the 'outer personality'. Definition: A fictional character representing various types within the target user group. Pruitt and Adlin (2006) defined it as a personality with a human face. "Design for only one person" - the broader the target, the further off the center of the target it lands.

What a persona is NOT: A statistical average (derived from qualitative research), a real person, a target market (purchasing ≠ using), a job description. Components (Olsen, 2004): Person's background, relationship with the design target, purpose of use, needs, attitudes, specific knowledge/skill level, context of use.

Creation 7 STEPS: ① Identify categories (goal: life/experience/use / role / market division) ② Classify clues & name ③ Build sub-categories & structural framework ④ Prioritize (frequency of use / market size / purchasing power / strategic importance) ⑤ Draft (basic whitepaper → story, eliminate stereotypes) ⑥ Evaluate (completeness, redundancy, validity) ⑦ User profiles.

4 Types of Structural Frameworks: Primary, Secondary, Served (sub-users), Negative (types the primary user might dislike). 3 Types of Profiles (1–10 scale): Usage value (utilitarian, hedonistic, social, personal), Usability (efficiency, accuracy, consistency, flexibility, meaningfulness), Reliability (safety, security, stability).

AI Persona: The procedure is the same, but available information is scarce, so gaps are filled with the provider's intent. Stakeholder persona = AS-IS, AI persona = TO-BE.

Chapter 9: Conceptual Model Planning

Conceptual Model Planning = Establishing a conceptual model of structure, function, and expression that remains relatively stable - the blueprint of a service. Approach: Concept-driven paradigm - designing generalizable concepts and values such as stories and creativity rather than focusing solely on technology and performance.

System Conflict = Conflict between system components and the surrounding environment during main function usage (e.g., pedometer step count error). 3 TRIZ solution methods:

Method	Key Concept	Example
Eliminate or Replace	Ideality strategy, replace/eliminate templates	Ultrasonic sensor height measuring device, IBM 'Doctor Watson'
Separate	Separate physical contradictions by space, time, or whole vs. part	Removable battery for electric scooters
Add	Auxiliary tools, component templates	GPS-embedded smartwatch, VR headset + robotic arm remote surgery

Concept Selection consists of two steps: Screening (filtering: +, -, 0 → total score ranking) + Scoring (1–5 points relative to a reference product, weighted W-Score). Note: Screening criteria markings in materials vary between 'Usefulness, Usability, Affect' and 'Usefulness, Usability, Trustworthiness' - remember both.

Metaphor = A rhetorical device comparing two objects to express one indirectly or implicitly. Connects a familiar source domain with an unfamiliar target domain to enhance user understanding. Caution: ① Focus selection on System Features - metaphors are a means, not an end ② The same metaphor may have different meanings for users vs. developers.

5 STEPS: ① Feature list ② Identify candidates (10+, contextual questions "Who (person) in what situation (state) performing what action (activity)" - Moran and Anderson, 1990) ③ Degree of match (Maximize S+M+, Minimize S-M+) ④ Select primary/secondary metaphors ⑤ Component diagram (1 primary + 3 secondary).

The primary metaphor covers overall functionality, information, and expression; secondary metaphors complement what primary ones cannot cover. Mind Inspection: Primary = 'Inspection', Secondary = Questionnaire, Mental Health Care, Psychological Counseling.

Chapter 11: Architecture Design

Architecture Design = Designing the static structure of a digital system. HCI perspective: Structure = Architecture, Function (Action) = Interaction, Expression = Interface. 4 Stages: Collect feature data → Classify categories → Establish relationships between categories → Diagramming.

8 Types of Feature Data

Type	Focus	Example
Fact	Understood without explanation	News content (5Ws and 1H)
Procedure	Sequential actions; usage stops if misunderstood	SammyTalk training guide
Principle	Operating principles; tailored to layperson level	AlzGuard card news
Rule / Policy	Required compliance (direct) / Success cases (indirect)	Medication "3 times a day, 30 min post-meal", EndeavorRx 25 min/day, 5 days/week
Story	Real or fictional narrative experiences	AlzGuard card news
Description	Stimulates emotion; preference drives value	TRIPP - VR Metaverse Wellness
Prediction	Anticipates trends; considers ethics	Sleep Cycle - Wake-up time prediction
Metadata	Data describing other data	Tags

When constructing with Generative AI: Beware of Hallucination limitations + design so the data collection process does not cause discomfort or suspicion.

Classification: Objective (mutually exclusive, structured - based on text/time/numbers) vs. Subjective (ambiguous, unstructured - based on topic: Rahman 2019 symptom identification / based on user: open/closed card sorting, KGC 'CareNow' is closed). Because information searching is inherently subjective, subjective classification is widely used. Deep learning classification faces black-box issues.

Relationship	Key Concept
Sequential	Simplest and most familiar; suitable for narrative or chronological flow
Grid	Matrix style; independent multiple criteria
Hierarchical	Top-down hierarchy; foundational structure for most services
Network	Unstructured connections; high interactivity
Hybrid	Hierarchical base + partial mixing - adopted by most recent systems
AI-Generated	Metadata-centric learning, flexible / requires accurate tagging

ER (Entity-Relationship) Diagram = Standardized notation for data structure and DB design; a tool to communicate feature data types, classifications, and relationships to developers. Entity = Rectangle, Attribute = Oval (e.g., Student ID), Relationship = Diamond (one of the main purposes of ERD). 3 Steps to create: Define data → Connect data to categories → Display relationships between categories. Example: 'Reborn' (sarcopenia) evaluates exercise level via 'Monitoring' and provides difficulty-adjusted 'Exercise'.

Chapter 13: Interface Design

Interface Design = Concrete representation of concepts, business models, architecture, and interactions in a form perceptible to users - key is accurate representation. A pretty interface ≠ a good interface. Requirements: Clear representation of value, function, and structural models; rapid representation of information and functions; faithful representation of aesthetic impression and individuality.

11 GUI Principles (Michael D. Byrne, 2001): Balance (weight distribution), Symmetry (axial symmetry), Continuity (gaze alignment), Rhythm (pattern variation), Contrast, Proportion, Unity, Simplicity, Density (information quantity), Regularity, Stability (center alignment). For complex services, Consistency and Diversity are added.

Effects are divided into Cognitive (visual hierarchy + legibility) and Affective (aesthetic impression + 3 dimensions of individuality: strength, formality, openness). 4 Layers of Hierarchy: Foreground (first information noticed), Midground (actual meaningful information), Background (highlights foreground), Off-screen/Out-of-frame (outside screen but connected). The same system may show different screens depending on the user - SammyTalk patient vs. expert view.

Voice User Interface (VUI) = Human-machine conversation via 'speech'. Non-verbal messages - Auditory Icons are natural sounds, Earcons are artificial signals indicating events, status changes, or errors.

Conversational User Interface (CUI) = Based on real human conversation models, natural language communication. 3 Elements of Good Conversation: Mutual context awareness, memory of past interactions, appropriate exchange of questions. Cooperative Principle (Grice, 1975): Quality (be truthful), Quantity (as much as needed), Relation (stay relevant), Manner (be clear and concise). Examples: Woebot, SammyTalk.

Haptic Interface = Uses tactile organs for easier and precise interaction - Neofect Smart Glove, vibrations for the hearing impaired.

Design Guidelines: Visual = Apple 10 principles (interactivity, legibility, readability, etc.), Voice = Google (providing info, error handling, keeping context, etc.), Conversational = 10 principles (visible feedback, consistency, error prevention, recognition over recall, etc.), Haptic = Apple 6 principles (consistency, complementarity, avoiding overuse, etc.).

Wireframe = Static or dynamic Low Fidelity reproduction of layout. Used to explain structure (page composition), content (displayed material), and function (operation method).

Paper + Pen - Numbering, omit icons/colors, indicate navigation with arrows
Digital Tools - Artboards, focus on achromatic colors & rectangles
Metaphor-based Tone & Manner - Mind Garden: Primary metaphor 'Garden', Chatbot 'Gardener'
Color, Icons, Fonts, Illustrations - Foreground = Titles, Midground = Descriptions, Background = Sun, rivers, mountains, fields
Various Modes - Tables, graphs, animations
Eliminate Unnecessary Elements - Prevention over error handling (e.g., removing record icon)
Measure Effects - Heuristic evaluation, usability testing, SUS, uMARS
Organize Results - Finalize standard interface, document cover & history

Easily Confused Concepts Comparison

Pair	Distinguishing Point
Usefulness vs. Usability	Achieving purpose vs. Efficiency of the usage process
Efficacy vs. Suitability	Controlled experiment vs. Real-world usage environment (Trade-off)
Utilitarian vs. Hedonistic Value	Practical tool (Pria) vs. Emotional enjoyment (EndeavorRx)
Socio-Technical Model vs. Persona Model	Ecosystem/Macro vs. Product focus/Micro
Auditory Icons vs. Earcons	Natural sounds vs. Artificial signal tones
Primary vs. Secondary Metaphor	Overall function/expression vs. Partial complement
Conceptual Model vs. Mental Model	Abstraction on system side vs. Knowledge structure in user's mind
Primary User vs. Secondary User	Direct interaction vs. Influencing/influenced by usage

Quick Scan: Names, Years & Acronyms

Keyword	Connection
Herbert A. Simon	Quote on observing artifacts
Dewey & Martin (1934)	Real Experience
Carl Jung	Persona = Outer personality
Pruitt & Adlin (2006)	Persona = Personality with a human face
Moran & Anderson (1990)	Contextual questions: person / state / activity
Grice (1975)	Cooperative Principle: Quality, Quantity, Relation, Manner
Michael D. Byrne (2001)	11 GUI Principles
KHIDI (2019)	SaMD efficacy classification

Acronyms: HRV, IML, DTx, SaMD, CMA/CEA/CUA/CBA, USTM, CUSTOM, OSTA, ETHICS, SSM, GUI/VUI/CUI, ERD, SUS, uMARS, W-Score.

Final Checklist

HCI 1.0/2.0/3.0 comparison table (focus, interface, features)?
Sensing–Reasoning–Actuating example matching, Digital Health = DBM+DTx, 3 DTx types?
Stakeholder definition, Kirby (1991), 3-step procedure (Preparation → Socio-Technical → Persona)?
Socio-Technical Model 4 STEPS → Persona 7 STEPS, 4 types of impacts, 4 structural frameworks, 3 profile types?
Problem space 4 elements ↔ 4 analyses, 4 types of mental models, Efficacy vs. Suitability?
3 TRIZ methods, Metaphor 5 STEPS?
8 types of feature data, 6 relationship types ('Recent = Hybrid'), ERD symbols?
11 GUI principles, Foreground/Midground/Background/Off-screen, Grice's 4 maxims, Wireframe 8 steps?

Reviewing up to this point completes the concept matching. A final glance at the comparison tables and the Names/Years table will wrap it up.

★ PINION ★