Human Abilities: Cognition James Landay John Kelleher
Plain English campaign! 2
Outline l l Human visual system Guidelines for design Models of human performance (MHP) Memory 3
Models of the User l Model Human Processor l Our Model Cognitive System Perceptual System Motor System Memory I/O CPU 4
Human I/O Channels l Input via the senses l l l Sight Hearing Touch Taste Smell l Output via motor control l l Limbs (feet? ) Fingers Eyes Head Voice 5
Why Model Human Performance? l l To test understanding To predict influence of new technology 6
The Model Human Processor l Developed by Card, Moran, & Newell (’ 83) l based on empirical data Long-term Memory Working Memory sensory buffers Eyes Ears Visual Image Store Perceptual Processor Auditory Image Store Motor Processor Fingers, etc. Cognitive Processor 7
Model Human Processor (Card, Moran & Newell) 8
Model Human Processor Components l l Cycle times Decay Rates Storage Capacities Coding/Representation Schemes 9
Perceptual (Sensory) Memories l l l Area of memory that deals with information from the senses Perceptual memories are highly volatile information stores Information flows from perceptual (sensory) memories into Working Memory Perceptual memories decay almost immediately and are replaced by new, incoming information Selection of stimuli governed by level of arousal 10
What is This? 11
Parameters of Perceptual Memories l Visual (Iconic) Memory l l Coding Scheme – Physical analogs Capacity – ~17 letters Decay Rate – ~200 ms Auditory (Echoic) Memory l l Coding Scheme – Physical analogs Capacity – ~5 letters Decay Rate – ~1500 ms Buffers for stimuli 12
Perceptual Processor The speed of the perceptual processor is about ~100 ms per cycle l l l Light blinks appearing within 100 ms look like a single brighter light l e. g. frames of a film seen as continuous fluid scene reflects some processing by sensory processor. Light blinks in two locations within 100 ms look like motion of a single light Auditory clicks occurring within 100 ms sound like one louder tone l Also, sensory memory for hearing more durable than others Multiple taps occurring within 100 ms feel like one tap of greater pressure 13
Working Memory (Short-term Memory) l l l Working Memory is a temporary information store. Working Memory receives information from Perceptual Memories and LTM Working Memory can influence LTM Information in Working Memory is often recoded l e. g. Visual information is rehearsed as auditory l Chunking is already happening People have some control over Working Memory l Rehearsal l Attention Sensory Attention memory Working Rehearsal memory Long-term memory 14
Memory l Working memory (short term) l small capacity (7 ± 2 “chunks”) l 6174591765 vs. (617) 459 -1765 l AIBIBMEMC vs. AIB IBM EMC l Chunking l Grouping together information into sections that make sense to the individual and seen as entities by that individual § E. g. master chess players (but only for legal positions) l rapid access (~ 70 ms) & decay (~200 ms) l pass to LTM after a few seconds 15
Chunking 1 chunk Recoded to image 3 chunks Recoded to words BIG OLD MAN B I G O L xxx xxx D M A N 9 chunks Recoded to letters Perceptual memory 16
Primacy and Recency Effects List 1 List 2 List 3 Barrier Babies File Firearms Sofa Heart Scarf Lobby Scarecrow Newspaper Clock Stylus Sea-shell Polish Maggot Tomato Lintels Rug Apologies Dog Flea Table Dolls-house Ball-pen Plant Oasis Jamboree Chemist Festival Neptune Identity Gnat Magnum Percolator Curtains Paper-clip Saucer Income Typist Tiles Precinct Subway Directory Argument Accident 17
Primacy & Recency Effect l Free recall l l Primacy effect l l Recalling lists in any order Tendency for first words on list to be commonly recalled Recency effect l Tendency for last words on list to be commonly recalled 18
Fun with Working Memory 25439762608 456 295 1413 HEC ATR ANU PTH ETR EET 19
Interference affects recency List 1 List 2 Then do… Aubergine Rescue 1 Chickenpox Gravestone +6 Elephant Flower -4 Telephone Fountain +9 Pendant Statue -2 Egg Fool +8 Melancholy Aphid -9 Cheese Surprise -6 Mug Printer -2 Nymph Cenotaph Dinghy Dog basket Tray Magnet Mole Lawn Tram Pram Macabre Sandwich 20
Memory l l Extent to which new material can be remembered depends on its meaningfulness Levels (Depth) of processing theory (Craik & Lockhart, 1972) l Info processed at different levels l l l e. g. , processing physical features of a word such as its sound Deep, semantical analysis Depth of processing determines how well remembered Elaborative (effortful process) vs. maintenance rehearsal Closure l Feeling of ‘relief’ when task successfully completed l l E. g. successful logon Important to permit processes to be chunked in memory l E. g. avoid traversing between windows within an application 21
Depth of Processing (shallow) l l l l l flame patch sonic bless avarice pears spade bliss forth l l l l l peels speed avoid freak pints rare blush slow pluck 22
Depth of Processing (deep) l l l l l spoon shares glass ports spray boots goose prize steam l l l l l runner grass pint stink bride green queen story brown 23
Memory l Factors that determine meaningfulness l Familiarity of an item, and the frequency with which a word occurs in everyday language l l l It’s associated imagery, the ability with which the word can elicit images in one’s mind l l l Familiar: Door, read, stop Unfamiliar: compile, substitute, scan Ride, sleep, eat Begin, increase, evaluate Information best recalled if in situ l E. g. Diver education (Baddeley) 24
Improving memory l Method of Loci l l Peg-word method l l Associate items with rhyming words & numbers Creating a narrative l l Visualising familiar route & associated items at particular locations long route Create story or song linking concepts together Creating acronyms l e. g. A. B. C. of First Aid 25
Implications for UI design l Items that need to be remembered at the interface should be as meaningful as possible l l l Problems with command line interfaces e. g. , command names and icons should be selected according to meaningfulness l cp vs. copy Words that represent visible objects easiest to recall Memory best facilitated by relaxed user Ask only relevant material l l Or provide user with reason for action Asking for non-sensible information 26
Peg Word Memory Aid l l l l l 1 2 3 4 5 6 7 8 9 10 bun shoe tree door hive sticks heaven gate wine hen 27
28
Memory Factors Total time hypothesis l Distribution of Practice Effect l Listen l The engines roared above the noise of the crowd. Even in the blistering heat people rose to their feet and waved their hands in excitement. The flag fell and they were off. Within seconds the car had pulled away from the pack and was careering round the bend at a desperate pace. Its wheels momentarily left the ground as it cornered. Coming down the straight the sun glinted on its shimmering paint. The driver gripped the wheel with fierce concentration. Sweat lay in fine drops on its brow. l People prone to embellishment and ‘localisation’ of facts. 29
Long-Term Memory (LTM) Long-term memory stores everything that we “know” -- facts, experience, knowledge, procedural rules of behavior. LTM has huge capacity. LTM has a relatively slow access compared to short-term memory. l l ‘activation’ is process of recall to WM l Only encode the important information l Forgetting also occurs slowly. Causes forgetting l 1) Never stored; encoding failed 2) Gone from storage; storage failed (? ? ) 3) Can’t get out of storage; retrieval failed l Interference model of forgetting l l l one item inhibits the retrieval of another proactive interference (3) retroactive interference (3 & 2) 30
Pennies Example 31
Long-term Memory l Long-term memory works by semantics and by association. breathes moves ANIMAL Is a barks eats Four legs has DOG Is a HOUND has tail Is a SHEEPDOG 32
Parameters of LTM Semantic network (encoded in terms of meaning and relationships). l l l Related associations, images, and past experiences How knowledge is encoded makes a difference in how knowledge is recalled Recognition is much easier than recall l l (DOS prompt vs. Mac user interface) Coding Scheme – Semantic Capacity – Unlimited Decay Rate – None l l However, recall from LTM is affected by encoding specificity and retrieval cues 33
Encoding Specificity l l “Specific encoding operations performed on what is perceived determine what is stored, and what is stored determines what retrieval cues are effective in providing access to what is stored. ” -- Card, Moran, & Newell (1983) This is a fancy way of saying that the encoding context matters 34
Discrimination Principle l “The difficulty of memory retrieval is determined by the candidates that exist in the memory relative to the retrieval cues. ” -- Card, Moran, & Newell (1983) 35
Cognitive Processor l l l “The recognize-act cycle…is the basic quantum of cognitive processing. ” (CMN, 1983) In each cycle, the contents of Working Memory activate something in LTM which in turn modifies the contents of Working Memory Cycle time is ~70 ms 36
Motor Processor l l Draw parallel lines (approx. 4 cm apart) For a duration of 5 seconds… l l Draw a zig-zag line back and forth between the lines working left to right The basic motor cycle time is ~70 ms l l l Move pen back and forth between two lines ~71 reversals in 5 sec, or ~70 ms/reversal forgetting anything? 37
Putting It All Together l True reaction time l l l 1 perceptual cycle + 1 cognitive cycle + 1 motor cycle 100 ms+70 ms = 240 ms Some studies include additional cognitive step l Raises total by 70 ms to 340 ms 38
Principles of Operation (cont. ) l Fitts’ Law l moving hand is a series of microcorrections l l time Tpos to move the hand to target size S which is distance D away is given by: l l correction takes Tp + Tc + Tm = 240 msec Tpos = a + b log 2 (D/S + 1) summary l time to move the hand depends only on the relative precision required 39
Fitts’ Law Example Pop-up Linear Menu Pop-up Pie Menu Today Sunday Monday Tuesday Wednesday Thursday Friday Saturday l Which will be faster on average? l pie menu (bigger targets & less distance) 40
Perception l Stimuli that occur within one PP cycle fuse into a single concept l frame rate needed for movies to look real? l l time for 1 frame < Tp (100 msec) -> 10 frame/sec. Perceptual causality l l two distinct stimuli can fuse if the first event appears to cause the other events must occur in the same cycle 41
Perceptual Causality l How soon must red ball move after cue ball collides with it? l must move in < Tp (100 msec) 42
What is missing from MHP? l Haptic memory l l Moving from sensory memory to WM l l for touch attention filters stimuli & passes to WM Moving from WM to LTM l elaboration 43
Cognitive Processes l Controlled l limited capacity; require attention and conscious control easier to change Automatic l Activities we carry out that have become automated l Reading, writing, speaking in native language… (others? ) l We don’t have to attend to (think about) what we are doing. 44
Automatic Processing l l The more we practice, the more our performance improves to the point that we become skilled, and performance is automatic Characteristics l l l fast, demanding minimal attention, therefore doesn’t interfere with other activities unavailable to consciousness hard to change once learned 45
Effect on UI design decisions l Interactions that have become automatic are difficult to unlearn l l Microsoft’s approach to Word. Perfect domination Consistency across versions, tools can help avoid this problem l Microsoft Office critical mass of usage stiffles Star. Office 46
Stroop Effect l l l Example of automatic behaviour Volunteer Start saying colors you see in list of words l l when slide comes up as fast as you can Say “done” when finished Everyone else time it… 47
Say the colour of these words Paper Home Back Schedule Page Change 48
Simple Experiment l l Do it again Say “done” when finished 49
Now do it again… Blue Red Black White Green Yellow 50
Importance of Context l l Bottom-up perception uses features of stimulus Top-down perception uses context (and prior knowledge) l l l Temporal (for hearing) – what we heard before or after stimulus Spatial (for visual) – what’s around the stimulus (as below) draws on long-term memory 51
Gestalt Laws of Grouping l l l German Psychologists Primary purpose of visual system is recognition of objects from basic visual elements Objects seen as more than a sum of the parts l l When elements are arranged in groups that define an object, we tend to see the object and not the elements. e. g. ascii art 52
Gestalt Principles Similarity Proximity Similarity Closure Continuity Symmetry 53
Law of Proximity l Things that are relatively close to one another tend to be grouped together. 54
Laws of Similarity l Items that look similar will be seen as parts of the same form 55
Law of good continuation l Objects arranged in either a straight line or a smooth curve tend to be seen as a unit. 56
Law of Closure l Innate tendency to perceive incomplete objects as complete and to close or fill gaps and to perceive asymmetric stimuli as symmetric 57
Law of common fate The law of common fate leads us to group together objects that move in the same direction. In the following illustration, imagine that three of the balls are moving in one direction, and two of the balls are moving in the opposite direction. If you saw these in actual motion, you would mentally group the balls that moved in the same direction. Because of this principle, we often see flocks of birds or schools of fish as one unit. 58
Attention “Everyone knows what attention is. It is the taking possession of mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought … It requires withdrawal from some things in order to deal effectively with others. ” W. James, 1890 59
“cocktail party phenomenon” l Ability to focus on one activity, while tuning out others l can be distracted from one task if attention called to another 60
Attention l Can design interfaces to help users find information they need l l l Can structure interface so it is easy to navigate Not too much information, nor too little Rather than arbitrarily presenting information: l use groupings l order in meaningful way l See Gestalt laws of perceptual grouping 61
Attention l Manner in which we deploy our attention has a tremendous bearing on how effectively we can interact with a system l l Focused attention l Ability to attend to one event from what amounts to a mass of competing stimuli in the environment Divided attention l Ability to attend to more than one stimuli at a time § l Cocktail party phenomenon Voluntary or involuntary 62
Implications for Design l Users are prone to distraction l On returning to suspended activity: l May forget where they left off l May forget whether they completed the task or not § l Forgetting to put salt French fries, or doing it twice Answer: cognitive aids l Reminders or external representations intended to gain attention at a time relevant to the task that needs to be performed. l E. g. , status area indicating task status, coffee cup on flaps 63
Attention l Other techniques for presenting information to guide attention l Spatial and temporal cues l l l Example Color Alerting techniques l l Flashing and reverse video Audio warnings 64
Attention l Windows are a useful way to partition the screen into discrete or overlapping sections l enable different types of information to be separated, provides meaningful groupings l e. g. , word processor § § § Text area Footnote area Command area 65
Implications for Design l Info which needs immediate attention should always be displayed in a prominent place (error and warning messages) 66
Implications for Design l Less urgent info should be allocated to a less prominent but specific areas of the screen so that the user will know where to look when this information is required. 67
Implications for Design l Information that is not needed very often should not be displayed but should be made available on request. 68
Goals of Representation Aiding l l Turn a cognitive task into a perceptual task. Offload human working memory onto an external representation. Map relevant constraints in the domain onto relevant representational properties. Encourage people to develop a “correct” mental model 69
Implications for Design l l l You have to understand the person’s task!! You have to understand the domain! People have cognitive constraints and abilities The domain imposes constraints Map those together into a design that represents domain constraints in a way that people can best perceive/understand. 70
Ventilator Management Example l l Practitioners: Intensive Care Unit specialists Task: To evaluate whether a patient is recovering his/her own breathing over time l Ventilator vs. Patient: Rate of breathing, depth of breathing 71
Today’s displays l l l Typical process control displays with tables and tables of “Label: Value” parameters Every variable that’s measured by the patient is displayed on the screen, (single sensor single indicator) e. g. : Ventilator Patient l l l Rate of breathing: 10 Depth of breathing: 5 Rate of breathing: 2 Depth of breathing: 6 Difficult to get “status at a glance”, to judge whether patient is improving or not. 72
Novel Display (Volume Rectangles) Ventilator Patient rate volume (Cole and Stewart, 1994) 73
A series of volume rectangles Ventilator Patient at time t 1 Ventilator Patient at time t 21 The patient is clearly doing more of his/her own breathing over time. 74
Evaluation of Novel Display l l Physicians had perfect performance in judging whether patients were getting better or worse over time. Furthermore, the physicians’ judgments were significantly faster using the novel volume rectangle display than using the familiar table of numbers currently used in practice. 75
Recognition over Recall l l l info reproduced from memory e. g. , command name & semantics Recognition l l l presentation of info provides knowledge that info has been seen before e. g. , command in menu reminds you of semantics easier because of cues to retrieval l cue can be anything related to item or situation where it was learned l example: giving hints l other examples in software? § icons, labels, menu names, etc. 76
Knowledge in the Head and in the World l l l Knowledge in the world l is the information in the environment Knowledge in the head l is the information that is stored in memory Most of the time we need to combine the two types knowledge to operate things. 77
Knowledge in the World/Head 78
Because. . . l Not all of the knowledge required for precise behavior has to be in the head l l partly in the head partly in the world partly in the constraints E. g. clipboard and ‘spike’ 79
Also because. . . l We can recognize material far more easily than we can recall it. l Knowledge in the world lets people recognize facts or things. § l E. g. road signs Knowledge in the head requires recall. 80
AIB 24 -Hour Online Banking l User logs on to: l l Check balances Move funds Cancel cheques AIB offers Transact. Online to: l Provide one-time credit card numbers tied to user’s account 81
Opening Logon Screen 82
Further Validation Screen 83
Transact. Online Sign. On 84
85
86
87
88
Further Reading Vision and Cognition l Books l l The Psychology Of Human-Computer Interaction, by Card, Moran, & Newell, Erlbaum, 1983 Human-Computer Interaction, by Dix, Finlay, Abowd, and Beale, 1998. Perception, Irvin Rock, 1995. Articles l “Using Color Effectively (or Peacocks Can't Fly)” by Lawrence J. Najjar, IBM TR 52. 0018, January, 1990, http: //mime 1. marc. gatech. edu/mime/papers/color. TR. html 89
Extra Slides 90
Vision (1/3) l l Two aspects: physical receptor & subsequent perception processing Photoreceptors: l l l Ganglion cells l l Rods (120 m. , light sensitive, can be saturated, concentrated on edges of retina, poor visual acuity). Cones (6 m. , less light sensitive, colour perceptors, concentrated on fovea, blind spot). specialised nerve cells X-cells (fovea centred, pattern detection) Y-cells (distributed on retina, movement detection) Size & Depth Perception l l l Visual angle (larger angle at same distance implies larger object) Visual acuity (fine detail perception) Law of size constancy l relies on cues - overlapping objects, size and height of object, familiarity with object. 91
Vision (2/3) l Brightness l l Subjective quantity; affected by luminance; contrast visual system adjusts to perceive in differing lighting; rods/cones visual acuity increases with luminance as does ‘flicker’ Colour l l 3 components (hue, intensity, saturation) Hue determined by wavelength l l l Blues short, greens medium and reds long wavelength. Intensity is brightness of colour Saturation is amount of whiteness in the colour (‘washed-out’ affect) 3 types of cones sensitive to RGB (fewest cones for blue) colour blindness Visual Processing l l Movement of retina & changes in luminance are perceived as constant Ability to interpret and anticipate images is vital - easily fooled, however. l Muller-Lyer illusion, Ponzo illusion. 92
Touch l l Secondary source of information Crucial to people with disabilities Touch is not localised 3 Types of sensory receptor l l l Thermoreceptors - heat and cold Nociceptors - intense pressure, heat and pain Mechanoreceptors - pressure l l rapidly adapting slowly adapting two-point threshold test Kinesthesis l l awareness of position of body and limbs three types l l l rapidly adapting (moving of limb) slowly adapting (movement and static position) positional receptors (static position only) 93
Engineering Models of Human Performance l l Predictive Quantitative l l l time to perform time to learn number and type of errors time to recover from errors Learnable and usable by systems designers Usefully approximate 94
LTM Processes • Remembering or Storing • • • Forgetting (2 Theories) • • • Repeat rehearsal or exposure to information aids remembering Ebbinghaus’s Total time hypothesis Baddeley’s Distribution of practice effect. Factors boosting memorability: familiarity, concrete images, meaningfulness, structure. Decay • Ebbinghaus - information decays logarithmically • Jost’s Law - Older memories more durable Interference Losses • Retroactive interference (newer knowledge inhibits older) • Proactive inhibition (older knowledge reappears) Non-emotive words more durable than emotive words (exhibited in nostalgia of ‘good old days’) Difficulty in proving forgetfulness. Associations need to be exercised. Retrieval • • Recall and Recognition Categorisation, vivid imagery and familiarity aid retrieval. 95
Reasoning • Deductive Reasoning • • • Inductive Reasoning • • Derives the logically necessary conclusion from the given premises. Some people are babies, some babies cry. Some people cry? Truth and validity clash. Bring world knowledge into reasoning process to facilitate shortcuts. Generalising from cases we have seen to infer information about cases we have not seen Every elephant we have seen has a trunk; therefore we infer all elephants have trunks • Unreliable inference • Cannot be proved; only disproved by producing a ‘trunkless’ elephant. Wason’s Cards. Need to disprove statement not add more proof. Abductive Reasoning • • • Reasoning from a fact to the action or state that caused it. Used to derive explanations from the events we observe. Often unreliable; though we hold such explanations until they can be disproven. 96
More Slides from… Washington State University School of EECS Cpt. S 443 - Human Computer Interaction
Long Term Memory l The Human World-Wide Web l Two types l l l episodic - events, organized temporally semantic - facts, organized associatively Representations l l l semantic nets frames scripts 98
Semantic Network university is a UI UW WSU vandal husky cougar is a animal is a 99
Frames l Extends semantic nets to include structured hierarchical information University WSU Fixed: type of school Fixed: type of University Default: has colleges Default: public Variable: public/private Variable: campus 100
Scripts l Stereotypical information l l l Entry conditions: need job, have money Result: educated, less money Props: books, schedule, new car Roles: instructor talks, students listen Scenes: classroom, dorm Tracks: internships, apprenticeships 101
Processes l How does information get from short term memory into long term memory? l l l Total time hypothesis - hit the books Distribution of practice effect - don’t cram Meaning - concrete better than abstract l l l faith age cold tenet quiet logic idea value past boat tree cat child rug plate gun flame head Structure, familiarity and concreteness 102
How We Forget l Decay l l l Logarithmically - forget most early Jost’s Law - if two equally strong memories at a given time, then the older is more durable. Interference l l l retroactive interference - old phone number proactive inhibition - driving to the old house emotion - good old days, forget the mundane 103
Information Retrieval l How do we recall details? l l Categorization Visualization 1 2 3 4 5 bun shoe tree door hive 6 sticks 7 heaven 8 gate 9 wine 10 hen 104
Real Intelligence l How is information processed and manipulated? l l Animals - receive and store info, but do not process it as well as humans Computers - receive and store info better then humans, but do not process it as well as humans 105
Human Intelligence l Humans use information to l l l Reason & solve problems Even if the info is partially missing or completely absent! Human thought is l l conscious & self-aware capable of imagination 106
Reasoning l Inferring missing information l l l Deductive - conclusions Inductive - generalizations Abductive - suppositions 107
Deductive Reasoning l If A then B l l l A. Therefore B not B, therefore not A. The phone rings when I’m in the shower l l l If I’m in the shower, then the phone rings When the phone rings, take a shower No shower? Phone doesn’t ring. 108
Inductive Reasoning l Specific A has property B then all A is B l l l Elephants have trunks Computers are slow Classes are exciting Students hand homework in on time WHETS is fun Geeks are rich 109
Wason Cards l If a card has a vowel on one side it has an even number on the other. True or False? 4 E 7 K 110
Abductive Reasoning l From fact to the action that caused it l l l Totalled car Black eye 4. 0 GPA Smile/frown Core dump Phone ring 111
Problem Solving l Using knowledge to find a solution l l l Gestalt theory Problem space theory Analogy 112
Gestalt Theory l Finding new solutions l Reproductive problem solving l l Productive problem solving l l Learned behavior, trial and error Behavioralist Fixation Invention, innovation, insight Pendulum problem 113
Problem Space Theory l Mapping out a solution step by step l l Problem states, goal state, current state Legal state transition operators Heuristics, e. g. means-ends analysis Examples l l l Games: 15 puzzle, chess Tasks: Setting the VCR clock Life (emphasis on “legal”) 114
Analogy l Applying one solution to a different problem l l Analogical mapping Purely productive reasoning is hard (10%) Drawing analogies is easier (80%) Existing solution “semantically close” to problem domain 115
Skill Acquisition l Solving problems that are not completely new l e. g. Chess l l Same goal (different goal states) Same transitions Different “skills” Problem groups l l novices group problems superficially experts group problems conceptually 116
ACT Skill Acquisition Model l How is skill acquired? General rules Proceduralization Specific rules Generalization Tuned rules 117
Errors l How do we make mistakes? l l Slips - change in context of skill Mental models - incorrect interpretation of the evidence 118
Design l How do we use what we know about humans to make better user interfaces? l l l Guidelines Models Evaluation 119
Скачать презентацию Human Abilities Cognition James Landay John Kelleher
a27eaa5af8c26a898989ff697d227e72.ppt