📘 Lecture 9 — Cognitive Process - Part I
📖 Overview: This lecture introduces the first two cognitive processes from the Extended Human Processing Model: Attention and Memory. It explains how attention allows us to select and focus on relevant information, and details the structure and function of human memory, including sensory, short-term, and long-term memory stores. Understanding these processes is crucial for designing interfaces that align with human cognitive capabilities and limitations.
🗂️ Topics Covered
This lecture begins by listing the cognitive processes (Attention, Memory, Perception, Learning, Reading/Speaking/Listening, Problem-solving). It then provides a detailed explanation of Attention, including its types (focused, divided, voluntary, involuntary) and how to focus attention at the interface. The second half covers Memory, explaining the multi-store model with sensory, short-term, and long-term memory, including the recency and primacy effects, and concludes with a revised memory model.
📝 Lecture Summary
9.1 Attention
Attention is the process of selecting a specific thing to concentrate on from all the available possibilities at a given moment. As psychologist Williams James described, it involves "taking possession of the mind by one out of several simultaneously possible objects or trains of thought," requiring withdrawal from some things to deal effectively with others. Attention can involve our auditory senses (e.g., waiting for your name to be called in a dentist's office, based on pitch, timber, and intensity) or visual senses (e.g., scanning a newspaper for football results, based on color and location). The ease of attending depends on whether we have clear goals and whether the needed information is salient in the environment.
Our goals guide our attention. If we know exactly what we want (e.g., who won the World Cup), we actively search for matching information. If our goal is vague (e.g., deciding what to eat at a restaurant), we browse, allowing our attention to be drawn to salient items. Information presentation also greatly influences attention. Structuring information meaningfully makes it easier to find. For example, a screen with information grouped into vertical categories with spaces between columns was found to be searchable in an average of 3.2 seconds, while a screen with the same information bunched together took an average of 5.5 seconds to search.
🔑 Definition — Attention: The process of selecting things to concentrate on, at a point in time, from the range of possibilities available. 📐 Formula/Model: Focused Attention (attending to one event from competing stimuli) vs. Divided Attention (attending to more than one thing at a time). Attention can be Voluntary (conscious effort) or Involuntary (grabbed by salient stimuli). 📌 Example - Focused vs. Divided: Having a conversation with one person is focused attention. Having a conversation while intermittently watching someone else is divided attention. An example of involuntary attention is being distracted from your work by music from the next room.
Focusing attention at the interface
For HCI, understanding attention is key to designing effective interfaces. Designers must consider how to focus users' attention on the relevant information for a given task, how to get their attention when needed, and how to avoid unnecessary distractions.
🔑 Definition — Structuring Information: Designing the interface so information is easy to navigate by presenting the right amount of information, grouped and ordered into meaningful parts using perceptual laws of grouping. 💡 Why this matters: Proper structuring helps the user attend to their task, not the interface itself. 📌 Example - Design Considerations: Help users stay focused by avoiding unnecessary distractions. Only create distraction or use alerts when they are truly appropriate. Make information salient when it needs to be attended to by using techniques like color, ordering, spacing, underlining, sequencing, and animation. Avoid cluttering the interface; follow the example of a crisp, simple design like Google.com.
9.2 Memory
Memory is essential for everyday activities, storing factual knowledge, actions/procedures, and our sense of identity. The multi-store model of memory describes three types of memory or memory functions.
Memory Model
The multi-store model of memory consists of three types of stores:
- Sensory store: modality-specific, holds information for a very brief period (a few tenths of a second).
- Short-term memory store: holds limited information for a short period (a few seconds).
- Permanent long-term memory store: holds information indefinitely.
Sensory memory acts as a buffer for stimuli received through the senses. A sensory memory exists for each sensory channel:
- Iconic memory: for visual stimuli. Information remains for about 0.5 seconds (e.g., the persistent image of a moving sparkler).
- Echoic memory: for aural stimuli. Allows brief playback of information (e.g., realizing you knew what someone asked even after you asked them to repeat it).
- Haptic memory: for touch.
Information is passed from sensory memory to short-term memory (STM) by attention, which filters stimuli to only those of interest. This selective attention is necessary due to our limited processing capacity and explains the cocktail party phenomenon—being able to focus on one conversation but switch attention if your name is mentioned elsewhere.
🔑 Definition — Short-term memory (STM) or Working memory: Acts as a "scratch pad" for temporary recall of information, used for fleetingly needed information (e.g., intermediate steps in mental arithmetic, holding the beginning of a sentence in mind while reading the rest). 📐 Capacity and Duration: STM has a rapid access time (~70ms) but decays rapidly (~200ms). Its capacity is limited to 7±2 chunks of information (Miller's Law). This can be increased by chunking information (e.g., remembering "22 55 36 8998 30" is easier than "54988319814237"). 📌 Example - Recency Effect: In a free recall task, items at the end of the list are recalled better because they are still active in STM. The primacy effect is the better recall for items at the beginning of the list, as they have been rehearsed more and placed into LTM.
Long-term memory (LTM) is our main resource, storing all our factual and experiential knowledge. It differs from STM in having a huge, if not unlimited, capacity, a slower access time (~0.1 seconds), and very slow forgetting.
Long-term memory structure
LTM is divided into two main types:
- Episodic memory: Represents our memory of events and experiences in a serial form. It allows us to reconstruct the actual events that took place at a given period of our lives.
- Semantic memory: A structured record of facts, concepts, and skills we have acquired. Information is derived from episodic memory. Semantic memory is thought to be structured as a semantic network, where items are associated in classes and can inherit attributes from parent classes (e.g., inferring that a specific sheepdog named Shadow has four legs and a tail). Other structures like frames and scripts organize information into data structures with slots for attribute values.
🔑 Definition — Episodic memory: The memory of events and experiences in a serial form. 🔑 Definition — Semantic memory: A structured record of facts, concepts, and skills.
9.3 Revised Memory Model
According to the revised memory model, working memory is a subset of LTM. In this view, items in working memory are activated chunks from long-term memory. Activation is supplied from other linked chunks and from sensory input. This model views human performance as a series of processing stages involving perceptual, motor, and cognitive systems, each with its own memory and processes, organized in a particular way.
⭐ Key Takeaways
The most critical concepts from this lecture are the two types of cognitive processes: attention and memory. For attention, you must understand how goals and information presentation (structuring) affect what we focus on, and the difference between focused, divided, voluntary, and involuntary attention. For memory, you must know the three stores of the multi-store model: sensory memory (iconic, echoic, haptic), short-term memory (with its 7±2 capacity and rapid decay), and long-term memory (episodic and semantic, including semantic networks). Key effects like the recency and primacy effects, and the concept of chunking, are essential for understanding how information is processed and recalled, directly informing interface design.
🧠 Quick Revision Questions
- What are the two main factors that determine how easy or difficult it is to pay attention to relevant information?
- Describe the difference between focused and divided attention, and provide an example of each from the lecture.
- What is the multi-store model of memory? Name the three stores and one key characteristic of each.
- Explain Miller's Law (the 7±2 rule) and provide an example of what "chunking" means in this context.
- What is the difference between episodic memory and semantic memory in long-term memory?
📘 Lecture 10 — Cognitive Processes - Part II
📖 Overview: This lecture continues the exploration of cognitive processes in Human Computer Interaction, focusing on learning and thinking. It covers how people learn through procedural and declarative methods, the psychology of reading/speaking/listening, and the higher-order cognitive processes of problem-solving, planning, reasoning, and decision-making. Understanding these cognitive mechanisms is crucial for designing interfaces that align with natural human learning and thinking patterns.
🗂️ Topics Covered
The lecture examines learning through procedural and declarative approaches, including practical learning strategies like the "training wheels" method and dynalinking. It then covers the three modes of language processing—reading, speaking, and listening—with their design implications. The final major section explores reflective cognition through problem-solving theories (Gestalt and Problem Space Theory), three types of reasoning (deductive, inductive, abductive), analogical mapping, and skill acquisition differences between novices and experts.
📝 Lecture Summary
10.1 Learning
Learning is divided into two categories: procedural learning, which focuses on "how to" use something (e.g., how to use a computer application), and declarative learning, which focuses on facts about something (e.g., using an application to understand a topic). Jack Carroll's research shows that people find it very hard to learn by following sets of instructions in a manual. When people encounter a computer for the first time, their common reaction is fear and trepidation, contrasting sharply with the motivation people feel when learning to drive a car. This discrepancy arises because driving is taught through actual doing, while computer learning often involves overwhelming manuals.
Experienced users are also reluctant to learn new methods from manuals, preferring to continue using familiar procedures even if less effective. People prefer learning through doing, and GUI and direct manipulation interfaces support this by allowing exploratory interaction and the ability to 'undo' actions. Carroll suggested the 'training wheels' approach, which restricts novice users to basic functions and extends them as the user gains experience, making initial learning more tractable.
Dynalinking is the process of linking and manipulating multimedia representations at the interface. It helps learners understand complex relationships by connecting concrete simulations with abstract representations. For example, a multimedia simulation of a pond ecosystem showed organisms swimming, and clicking on an organism revealed what it was and what it ate. This simulation was dynalinked with abstract food web diagrams, allowing users to see how changes in one representation affected the other. Dynalinking is especially powerful for learning complex, multi-dimensional information.
🔑 Definition — Procedural Learning: Learning focused on how to do something or how to use an object. 🔑 Definition — Declarative Learning: Learning focused on facts about something. 🔑 Definition — Training Wheels Approach: Restricting novice users to basic functions and extending them as they gain experience. 🔑 Definition — Dynalinking: The process of linking and manipulating multimedia representations at the interface to show relationships. 💡 Why this matters: Understanding learning preferences helps designers create interfaces that support exploratory learning, undo capabilities, and gradual skill development rather than forcing users through linear manual instruction.
10.2 Reading, Speaking and Listening
These three forms of language processing share the property that sentence meaning remains the same regardless of the mode of conveyance. However, the ease of reading, listening, or speaking differs based on person, task, and context. Written language is permanent while listening is transient—rereading is possible but rebroadcasting spoken information is not. Reading can be quicker than speaking or listening because text can be rapidly scanned. Listening requires less cognitive effort than reading or speaking, especially for children. Written language tends to be grammatical while spoken language is often ungrammatical. Differences in ability exist, with dyslexics having difficulty understanding written words, and people who are hard of hearing or seeing being restricted in language processing.
Design implications include keeping speech-based menus and instructions short (people struggle with more than three or four options), accentuating intonation in artificial speech voices, and providing opportunities to enlarge text without affecting formatting for those who find small text difficult to read.
🔑 Definition — Dyslexics: Individuals who have difficulties understanding and recognizing written words, making it hard to write grammatical sentences and spell correctly.
10.3 Problem Solving, Planning, Reasoning and Decision-making
These cognitive processes involve reflective cognition, which includes thinking about what to do, what the options are, and what the consequences might be. They involve conscious processing, discussion with others, and use of artifacts like maps and books. The extent to which people engage in reflective cognition depends on their level of experience. Novices often act by trial and error, making mistakes and acting inefficiently, while experts have more knowledge and can select optimal strategies and think ahead about consequences.
Reasoning uses knowledge to draw conclusions or infer new information about a domain. There are three types:
- Deductive reasoning derives the logically necessary conclusion from premises. For example: "If it is Friday then she will go to work. It is Friday. Therefore she will go to work." Validity does not require truth—a valid deduction can conflict with real-world knowledge.
- Inductive reasoning generalizes from cases we have seen to infer information about unseen cases. For example, if every elephant observed has a trunk, we infer all elephants have trunks. This inference is unreliable—it can only be proved false, never proved true.
- Abductive reasoning reasons from a fact to the action or state that caused it. For example, if Sam always drives fast after drinking, and we see Sam driving fast, we may infer she has been drinking. This is also unreliable but people hold onto explanations until evidence supports an alternative.
Problem solving is the process of finding a solution to an unfamiliar task using available knowledge. The Gestalt theory views problem solving as both reproductive (drawing on previous experience) and productive (involving insight and restructuring). Reproductive problem solving can be a hindrance if a person 'fixates' on known aspects and cannot see novel interpretations. However, Gestalt theory lacks sufficient evidence about when restructuring occurs or what insight is.
The Problem Space Theory (Newell and Simon, 1970s) proposes that problem solving centers on the problem space, which comprises problem states. People use legal state transition operators to move from an initial state to a goal state. Means-ends analysis is a heuristic where the initial state is compared with the goal state, and an operator is chosen to reduce the difference. For example, moving a desk requires recognizing the location difference and creating sub-goals like making the desk light enough to carry. Problem space searching is limited by short-term memory capacity and retrieval speed.
Analogy in problem solving uses analogical mapping—mapping knowledge from a similar known domain to a new problem. In Gick and Holyoak's experiment, only 10% of subjects solved the tumor problem (needing converging low-intensity rays) without help, but 80% succeeded when given the analogous fortress story about converging small groups of soldiers.
Skill acquisition shows differences between novices and experts. Chess experiments demonstrated that masters remembered board configurations and associated good moves using larger 'chunks' than less experienced players. Masters considered no more alternatives than novices but took less time and produced better moves. Novices group problems by superficial characteristics (objects or features), while experts group by underlying conceptual similarities.
🔑 Definition — Reflective Cognition: Cognitive processes involving thinking about what to do, options, and consequences, often requiring conscious processing. 🔑 Definition — Deductive Reasoning: Deriving the logically necessary conclusion from given premises. 🔑 Definition — Inductive Reasoning: Generalizing from observed cases to infer information about unobserved cases. 🔑 Definition — Abductive Reasoning: Reasoning from a fact to the action or state that caused it. 🔑 Definition — Problem Space Theory: The theory that problem solving centers on problem states and uses legal operators to move from initial to goal states. 🔑 Definition — Means-Ends Analysis: A heuristic where the initial state is compared with the goal state and an operator is chosen to reduce the difference. 🔑 Definition — Analogical Mapping: Mapping knowledge from a similar known domain to a new problem. 📐 Formula: Deductive Reasoning: If P then Q; P is true → Therefore Q is true. 📐 Formula: Inductive Reasoning: All observed cases of X have property Y → Therefore all X have property Y (unreliable). 📐 Formula: Abductive Reasoning: Fact F is observed; Known rule: Action A causes F → Therefore A occurred (unreliable). 📌 Example (Deductive): If it is raining then the ground is dry (premise). It is raining (premise). Therefore the ground is dry—valid deduction even though it conflicts with real-world knowledge. 📌 Example (Inductive): Every elephant ever seen has a trunk → Infer all elephants have trunks. Cannot be proved true (next elephant might be trunkless). 📌 Example (Abductive): Sam drives too fast when she has been drinking (known rule). Sam is driving too fast (observed fact) → Infer Sam has been drinking (possibly wrong—she may have an emergency). 📌 Example (Problem Space - Means-Ends): Initial state: desk at north wall. Goal state: desk by window. Difference: location. Desk is heavy (constraint). Sub-goal: make desk light (remove drawers). Operator: carry desk.
⭐ Key Takeaways
Students must understand the fundamental distinction between procedural and declarative learning and why people prefer learning through doing rather than reading manuals. The "training wheels" approach and dynalinking are critical design strategies for facilitating learning at interfaces. The three modes of language processing have distinct properties that inform design choices, particularly the limitations of speech-based interfaces. The three types of reasoning (deductive, inductive, abductive) are essential for understanding how users draw conclusions from system behavior, with abduction being particularly important for interface design because users infer causation from correlation. Finally, the Problem Space Theory with means-ends analysis, analogical mapping, and the differences between novice and expert problem-solving (chunking, superficial vs. conceptual grouping) are foundational concepts for designing systems that support users at different skill levels.
🧠 Quick Revision Questions
- What are the two types of learning, and how does the "training wheels" approach support each type?
- What is dynalinking, and how did the pond ecosystem experiment demonstrate its effectiveness for learning complex concepts?
- What are the three types of reasoning? Give a real-world example of each type and explain why abductive reasoning is particularly relevant to interface design.
- According to Problem Space Theory, what is means-ends analysis, and how does it help users solve problems within the constraints of human processing limitations?
- What differences were observed between chess masters and less experienced players in terms of board memory and problem grouping strategies?
📘 Lecture 11 — The Psychology of Actions
📖 Overview: This lecture explores how people form mental models to understand and interact with systems, and how these models can lead to errors. It examines the psychology of action through the seven stages of action cycle and classifies different types of human errors, explaining why self-blame and learned helplessness occur in human-computer interaction.
🗂️ Topics Covered
The lecture covers mental models and their characteristics, the difference between images and mental models, erroneous mental models with examples like thermostats and elevator buttons, self-blame and learned helplessness, taught helplessness, the nature of human thought and explanation with case studies of Three Mile Island and Lockheed L-1011 accidents, the action cycle including execution and evaluation stages, and a classification of errors into slips and mistakes.
📝 Lecture Summary
11.1 Mental model
The concept of mental model was first developed in the early 1640s by Kenneth Craik, who proposed that thinking involves "models, or parallels reality." Donald Norman defines mental models as "the model people have of themselves, others, the environment, and the things with which they interact. People form mental models through experience, training and instruction."
Mental models are built to make predictions about external events before carrying out an action. An important observation is that they are invariably incomplete, unstable, easily confusable, and often based on superstition rather than scientific fact.
Within cognitive psychology, Johnson-Laird (1983, 1988) explicated mental models regarding their structure and function in human reasoning. He argues that mental models are either analogical representations or a combination of analogical and propositional representations. A mental model represents the relative position of objects in an analogical manner that parallels the structure of the state of objects in the world.
🔑 Definition — Mental Model: "The model people have of themselves, others, the environment, and the things with which they interact. People form mental models through experience, training and instruction." (Donald Norman)
An important difference exists between images and mental models regarding their function. Mental models are constructed when we need to make an inference or prediction, and a conscious mental simulation may be "run" from which conclusions can be deduced. An image, conversely, is a one-off representation. A simplified analogy: an image is like a frame in a movie while a mental model is more like a short snippet of a movie.
While learning and using a system, people develop knowledge of how to use the system and, to a lesser extent, how the system works. These two kinds of knowledge are often referred to as a user's mental model. The more someone learns about a system, the more their mental model develops. TV engineers have a deep mental model of how TVs work, while an average citizen has a shallow mental model of operation.
📌 Example — Thermostat Mental Models: When arriving home to a cold house with a baby, most people set the thermostat as high as possible, believing this increases the rate of warming. This is incorrect. Two commonly held folk theories exist:
- Timer theory: The thermostat controls the proportion of time the device stays on (set higher = more time on)
- Valve theory: The thermostat controls how much heat comes out (set higher = more heat)
In reality, thermostats work by switching the heat on at full power until the desired temperature is reached, then turning off completely. Setting the thermostat at one extreme cannot affect how long it takes to reach the desired temperature. The design gives absolutely no hint as to the correct answer, so people form theories (mental models) to explain what they observe.
Why do people use erroneous mental models? They are running a mental model based on general valve theory, assuming the principle of "more is more": the more you turn or push something, the more it causes the desired effect. This holds for taps and radio controls but not for thermostats, which function based on an on-off switch principle.
Other common examples of erroneous mental models include:
- People pressing an elevator button multiple times, believing it will make it arrive faster
- Bashing keys when the cursor freezes on a computer screen
- Hitting the top of a TV when it starts acting up
💡 Why this matters: Not having appropriate mental models causes people to become very frustrated, often resulting in stereotypical "venting" behavior. If people could develop better mental models, they would be better equipped to carry out tasks efficiently and know what to do when systems act up.
Self-blaming: When users cannot use an everyday thing, they often blame themselves. If they believe others can use the device and that it is not very complex, they conclude difficulties must be their own fault. This creates a conspiracy of silence, maintaining feelings of guilt and helplessness. Interestingly, people normally attribute their own problems to the environment and other people's problems to their personalities—but with everyday objects, they reverse this tendency.
Reason for self-blaming:
- Learned helplessness: People experience failure at a task numerous times, decide the task cannot be done (at least not by them), and stop trying. A few bad experiences can create this feeling, potentially leading to depression.
- Taught helplessness: The design of everyday things seems almost guaranteed to cause this phenomenon. With badly designed objects, faulty mental models, and poor feedback, people feel guilty. The vicious cycle: fail at something → think it's your fault → think you can't do it → don't try → can't do it → self-fulfilling prophecy.
The nature of human thought and explanation: It isn't always easy to tell where blame should be placed. Dramatic accidents have occurred from false assessment of blame. When instruments indicate something is wrong, operators must consider whether the instruments themselves are wrong.
📌 Example — Three Mile Island Nuclear Power Plant: Operators pushed a button to close a valve that had been properly opened. The valve was deficient and didn't close, but a light indicated the valve position was closed. The light monitored only the electrical signal, not the actual valve. Operators saw high temperature in the pipe leading from the valve, indicating fluid was still flowing, but knew the valve had been leaky and assumed the leak wouldn't affect main operation. They were wrong—escaping water added significantly to the nuclear disaster.
📌 Example — Lockheed L-1011: Flying from Miami to Nassau, the low oil pressure light for one engine went on. The crew shut it down and turned back. Eight minutes later, low-pressure lights for the other two engines went on. The crew didn't believe it—simultaneous oil exhaustion in all three engines seemed "one in millions." The second and third engines failed for lack of oil. Three missing O-rings, one missing from each oil plug, allowed all oil to seep out. The National Transportation Safety Board declared the crew's analysis was "logical."
💡 Why this matters: Once we have an explanation—correct or incorrect—for discrepant events, there is no more puzzle. We become complacent. Our explanations are based on analogy with past experience that may not apply in the current situation.
How people do things: To get something done, you need:
- The goal—what is to be achieved
- Action to move yourself or manipulate something
- Check to see that your goal was met
A goal must be transformed into specific statements called intentions. An intention is the specific action taken to get to the goal. Intentions must be made into specific action sequences that control muscles.
📌 Example: Sitting in an armchair reading as light dims. Goal: get more light. Intention: push the switch button on the lamp. Action sequence: how to move the body, stretch to reach the switch, extend the finger to push the button.
Action Cycle: Human action has two aspects—execution and evaluation. Execution involves doing something. Evaluation is the comparison of what happened in the world with what we wanted to happen.
Stages of Execution:
- Start with the goal (state to be achieved)
- Goal translated into an intention to do some action
- Intention translated into a set of internal commands, an action sequence
- Action sequence is executed, performed upon the world
Stages of Evaluation:
- Perception of the world
- Perception is interpreted according to expectations
- Interpretation is compared with respect to both intentions and goals
Seven stages of action: The stages of execution (intentions, action sequence, and execution) are coupled with the stages of evaluation (perception, interpretation, and evaluation), with goals common to both stages.
11.2 Errors
Human capability for interpreting and manipulating information is impressive, but we do make mistakes. When learning to use a computer system, learners are often frightened of making errors because, as well as making them feel stupid, they think it can result in catastrophe. The anticipation of making an error can hinder a user's interaction with a system.
Some errors result from changes in the context of skilled behavior. If a pattern of behavior has become automatic and we change some aspect, the more familiar pattern may break through. Example: intending to stop at the shop on the way home but driving past—the more familiar activity of driving home overrides the less familiar intention.
Other errors result from an incorrect understanding or model of a situation or system—these are mental models. Characteristics: partial, unstable, subject to change, internally inconsistent, often unscientific, based on superstition or incorrect interpretation of evidence.
A classification of errors: Norman categorized errors into two main types:
🔑 Definition — Mistakes: Occur through conscious deliberation. An incorrect action is taken based on an incorrect decision. Example: trying to throw the icon of the hard disk into the wastebasket as a way of removing all existing files from the disk is a mistake. A menu option to erase the disk is the appropriate action.
🔑 Definition — Slips: Are unintentional. They happen by accident, such as making typos by pressing the wrong key or selecting the wrong menu item by overshooting. The most frequent errors are slips, especially in well-learned behavior.
⭐ Key Takeaways
Mental models are incomplete, unstable representations that people form through experience and instruction to predict and understand system behavior. Erroneous mental models, like the "more is more" philosophy applied to thermostats and elevator buttons, are surprisingly common and lead to user frustration and self-blame. The phenomenon of learned helplessness and taught helplessness explains why repeated failure with poorly designed objects creates self-fulfilling prophecies of inability. The seven stages of action (goal, intention, action sequence, execution, perception, interpretation, evaluation) provide a framework for understanding how people interact with systems. Finally, errors are classified into mistakes (conscious incorrect decisions) and slips (unintentional accidents), with slips being the most frequent in well-learned behavior.
🧠 Quick Revision Questions
- What are the key characteristics of mental models according to Norman and Johnson-Laird?
- Why do people set thermostats to maximum when trying to heat a room quickly, and what is the actual correct functioning of a thermostat?
- What is the difference between learned helplessness and taught helplessness in the context of human-computer interaction?
- What are the seven stages of action in the action cycle, and how do execution and evaluation differ?
- What is the difference between a mistake and a slip, and provide an example of each?
📘 Lecture 12 — Design Principles
📖 Overview: This lecture focuses on conceptual models in HCI and key design principles that guide the creation of usable interfaces. It explains how designers can bridge the gap between user mental models and system implementation through represented models, and introduces six fundamental design principles that make interfaces intuitive and easy to use.
🗂️ Topics Covered
The lecture begins by defining conceptual models and their importance in design, then explores the three models in digital systems (user mental model, designer's model, and implementation model). It discusses the Gulfs of Execution and Evaluation, introduces the Seven Stages of Action as a design aid, and concludes with a detailed examination of six core design principles: Visibility, Affordance, Constraints, Mapping, Consistency, and Feedback.
📝 Lecture Summary
Conceptual Model
The lecture opens with a quote from David Liddle: "The most important thing to design is the user's conceptual model." A conceptual model is defined as a description of the proposed system in terms of a set of integrated ideas and concepts about what it should do, behave, and look like, that will be understandable by the users in the manner intended. Developing a conceptual model involves envisioning the proposed product based on user needs and requires iterative testing. A key aspect is deciding what the user will be doing—searching, creating documents, communicating, or recording events. The interaction mode (browsing vs. direct questioning) must be considered separately from the interaction style (menu-based, speech inputs, commands), with the former being at a higher level of abstraction. Once possible ways of interacting are identified, the conceptual model is "fleshed out" by working out interface behavior, interaction style, and "look and feel." Another approach is using interface metaphors, like the desktop, which provide a basic structure using familiar knowledge.
The lecture then introduces three models in the digital world:
- Implementation model — what is actually happening inside the computer
- Designer's represented model — the way the designer chooses to represent a program's functioning to the user
- User's mental model — the user's understanding of how the system works
Donald Norman refers to the represented model simply as the designer's model. In software, a program's represented model can be quite different from the actual processing structure. For example, an operating system can make a network file server look like a local disk even though the physical disk drive may be miles away. The closer the represented model comes to the user's mental model, the easier the program is to use. Offering a represented model that follows the implementation model too closely reduces the user's ability to learn and use the program. We form mental models that are simpler than reality, so creating represented models that are simpler than the implementation model helps users achieve better understanding—like imagining pressing a brake pedal pushes a lever against wheels, when the actual mechanism involves hydraulic cylinders and perforated disks.
🔑 Definition — Conceptual model: A description of the proposed system in terms of a set of integrated ideas and concepts about what it should do, behave, and look like, that will be understandable by the users in the manner intended.
🔑 Definition — Interface metaphor: A basic structure for the conceptual model that is couched in knowledge users are familiar with, such as the desktop metaphor.
🔑 Definition — Designer's represented model: The way the designer chooses to represent a program's functioning to the user, which can differ from the actual processing structure.
The Gulf of Execution and Gulf of Evaluation
There are several gulfs that separate mental states from physical ones, each reflecting the distance between the person's mental representation and the physical components of the environment. These present major problems for users.
The Gulf of Execution — Does the system provide actions that correspond to the person's intentions? The difference between intentions and allowable actions is the gulf of execution. One measure is how well the system allows the person to do intended actions directly, without extra effort.
The Gulf of Evaluation — Does the system provide a physical representation that can be directly perceived and interpreted in terms of the person's intentions and expectations? The gulf of evaluation reflects the effort required to interpret the physical state of the system and determine how well expectations have been met. The gulf is small when the system provides information that is easy to get, easy to interpret, and matches how the person thinks of the system.
The Seven Stages of Action serves as a valuable design aid, providing a basic checklist to ensure the gulfs are bridged. Each stage of action requires its own design strategies and provides its own opportunity for disaster. The questions for each stage boil down to the principles of good design.
🔑 Definition — Gulf of Execution: The difference between the person's intentions and the allowable actions provided by the system.
🔑 Definition — Gulf of Evaluation: The amount of effort the person must exert to interpret the physical state of the system and determine how well expectations and intentions have been met.
12.1 Design Principles
The lecture describes six common design principles: Visibility, Affordance, Constraints, Mapping, Consistency, and Feedback.
Visibility
The more visible functions are, the more likely users will be able to know what to do next. When functions are "out of sight," they become more difficult to find and know how to use. Norman describes car controls as an example—indicator, headlights, horn, and hazard warning lights are clearly visible and positioned to make it easy for the driver to find the appropriate control. The lecture provides a personal example of encountering a problem in word processing software where the "Properties" option was hidden under an arrow at the bottom of the File menu, requiring the user to click the arrow to see it—demonstrating poor visibility.
Affordance
Affordance refers to an attribute of an object that allows people to know how to use it. For example, a mouse button invites pushing by the way it is physically constrained in its plastic shell. At a simple level, to afford means "to give a clue." When affordances are perceptually obvious, it is easy to know how to interact—a door handle affords pulling, a cup handle affords grasping, a mouse button affords pushing.
There are two kinds of affordance:
- Real affordances — Physical objects have real affordances like grasping that are perceptually obvious and do not have to be learned.
- Perceived affordances — Screen-based interfaces are better conceptualized as perceived affordances, which are essentially learned conventions. Physical devices like control consoles benefit from real affordances (pulling, pressing), while virtual interfaces rely on perceived affordances such as icons that afford clicking and scroll bars that afford moving up and down.
💡 Why this matters: Understanding the difference between real and perceived affordances helps designers create interfaces where users intuitively know what to do without instruction.
Constraints
Constraints refer to determining ways of restricting the kind of user interaction that can take place at a given moment. A common design practice is deactivating certain menu options by shading them, restricting users to only permissible actions and reducing the chances of mistakes. Different kinds of graphical representations can also constrain interpretation—for example, flow chart diagrams show which objects are related to which.
Norman classified constraints into three categories:
-
Physical constraints — The way physical objects restrict movement. For example, a disk can be inserted into a disk drive in only one way due to its shape and size. Keys on a pad can usually be pressed in only one way.
-
Logical constraints — Rely on people's understanding of how the world works and common-sense reasoning about actions and consequences. For example, picking up a physical marble and placing it elsewhere would be expected to trigger something. Disabling menu options when inappropriate provides logical constraining.
-
Cultural constraints — Rely on learned conventions, like using red for warning, certain signals for danger, or the smiley face for happy emotions. Most cultural constraints are arbitrary—their relationship with what is being represented is abstract and could have evolved differently (e.g., yellow instead of red for warning). Once learned and accepted by a cultural group, they become universally accepted conventions. Two universally accepted interface conventions are windowing for displaying information and icons on the desktop.
🔑 Definition — Constraints: Ways of restricting the kind of user interaction that can take place at a given moment.
Mapping
Mapping refers to the relationship between controls and their effects in the world. Nearly all artifacts need some kind of mapping between controls and effects. An example of good mapping is the up and down arrows on a computer keyboard representing up and down cursor movement. The mapping of relative position is also important—consider music players where play is in the middle, rewind on the left, and fast-forward on the right, mapping directly onto the directionality of actions. The lecture illustrates this with two figures: Figure (a) shows poor mapping (which would be difficult to use) and Figure (b) shows good mapping following the conventional sequence.
🔑 Definition — Mapping: The relationship between controls and their effects in the world.
Consistency
Consistency refers to designing interfaces to have similar operations and use similar elements for achieving similar tasks. A consistent interface follows rules, such as using the same operation (left mouse button click) to select all objects. Inconsistent interfaces allow exceptions to a rule—for example, where certain graphical objects are highlighted using the right mouse button while all others use the left button. This makes it difficult for users to remember and prone to mistakes.
Benefits of consistent interfaces include being easier to learn and use—users learn only a single mode of operation applicable to all objects. This works well for simple interfaces with limited operations (like a mini CD player). However, consistency is more problematic for complex interfaces with hundreds of operations, where creating categories of commands mapped into subsets of operations is more effective (like categorizing word-processing operations into different menus).
Another problem is determining what to make consistent with what else. There are many choices:
- External consistency — Designing to be consistent with how people do things in the outside world
- Internal consistency — Designing to be consistent with the existing system
The problem facing designers is knowing which approach to use, as there are many different ways of doing things both physically and electronically.
🔑 Definition — Consistency: Designing interfaces to have similar operations and use similar elements for achieving similar tasks.
Feedback
Feedback is related to visibility and involves sending back information about what action has been done and what has been accomplished, allowing the person to continue with the activity. Without feedback, actions would have no effect—imagine playing a guitar, slicing bread, or writing with a pen if none of the actions produced any effect for several seconds. Various kinds of feedback are available for interaction design: audio, tactile, verbal, visual, and combinations of these. Deciding which combinations are appropriate for different activities is central. Using feedback in the right way also provides the necessary visibility for user interaction.
🔑 Definition — Feedback: Sending back information about what action has been done and what has been accomplished, allowing the person to continue with the activity.
⭐ Key Takeaways
The most critical concept from this lecture is that a well-designed conceptual model bridges the gap between the user's mental model and the system's implementation through an appropriate represented model. The six design principles—Visibility, Affordance, Constraints, Mapping, Consistency, and Feedback—serve as practical guidelines for creating interfaces that are intuitive and easy to use. Understanding the Gulfs of Execution and Evaluation helps designers identify where users struggle, while the Seven Stages of Action provides a checklist for bridging these gulfs. Students must remember the three types of constraints (physical, logical, cultural) and the difference between real and perceived affordances, as these are fundamental to designing interfaces that prevent errors and guide user behavior naturally.
🧠 Quick Revision Questions
- What is the difference between the user's mental model, the designer's represented model, and the implementation model?
- How do the Gulf of Execution and the Gulf of Evaluation differ, and what does it mean when each gulf is "small"?
- What are the three categories of constraints according to Norman, and give one example of each from user interfaces?
- Why can consistency be problematic when designing complex interfaces with hundreds of operations?
- What is the difference between real affordances and perceived affordances, and which type is more relevant for screen-based interfaces?