Topic 5_Human information processing_Fall 2024_Final Updated.pdf

Topic 5: Human Information Processing (CH6 & CH7) Instructor: Dr. Shuping Xiong Visiting Professor, School of Industrial Engineering Purdue University Fall, 2024 CONTENTS Introduction to Information & Info. Quantification Human Informa...

Topic 5: Human Information Processing (CH6 & CH7) Instructor: Dr. Shuping Xiong Visiting Professor, School of Industrial Engineering Purdue University Fall, 2024 CONTENTS Introduction to Information & Info. Quantification Human Information Processing (HIP) Model Four Stages of HIP Model Mental Workload and Evaluation Methods Summary 2 INFORMATION 3  TYPES OF INFORMATION  Quantitative (e.g., 100% charged, 63% charged)  Qualitative (e.g., fully charged, partially charged)  Status(normal, abnormal)  Warning (abnormal -potentially dangerous)  Representational (e.g., pictures, diagrams)  Identification (e.g., labels)  …… 4 WHAT IS INFORMATION?  Information is defined as the reduction of uncertainty. Q: Which event (if it happened) conveys more information to drivers?  The occurrence of highly certain events do not convey much information  ‘Fasten seat belt sign’ conveys less info. because it is expected  Although it is an important message, the importance is not directly considered in the information definition.  The occurrence of highly unlikely events convey more information 5  ‘High temperature of car engine warning’ in a car HOW TO QUANTIFY INFORMATION?  Informationis measured in bits (symbolised by H). A bit is the amount of information required to decide between two equally likely alternatives. H = log2 N bits N = number of equiprobable alternatives 6  (1) EQUIPROBABLE CASE  Formula H = log2 N bits; N = number of equiprobable alternatives  Note log2 X @ 3.32 log10 X  Examples  Tossing a single coin to see which way up it falls: There are 2, equally likely alternatives (heads & tails), i.e. n = 2; H = log2n=1. That is, the amount of information ‘resolved’ by tossing a single, unbiased coin is 1 bit  N = 8  H = log2 8 = 3 bits 7  INFORMATION QUANTIFICATION-RATIONALE Number of binary questions you have to ask to find which of the 8 it is. Q1 1 2 3 4 5 6 7 8 Q2 5 6 7 8 Q3 5 6 6  3 bits 8 The minimum number of binary questions needed to find the answer provides the measure of information. (2) NON-EQUIPROBABLE CASE  To deal with unequal probabilities, we need first to modify the basic equation by replacing n with 1/p, where p = the probability of each event.  Mathematically, this is OK because they have the same value when probabilities are all equal: p = 1/n, or n = 1/p.  Making this substitution, H = log n can be written as: H = 2 log 1/p 2  Probabilities of different events are unequal, the equation is expanded so that it has a separate term for each possible event: pi ×log2 1/pi  The total amount of information (from all possible events): 9 (2) NON-EQUIPROBABLE CASE-EXAMPLES  When tossing a very biased coin the possible outcomes might be: Heads: pheads = 0.2 (i.e. it comes down heads only once in five tosses); Tails: ptails = 0.8 (i.e. it comes down tails four out of five times)  Q: What is the total amount of information? 10 = 0.72 bit < 1 bit (2) NON-EQUIPROBABLE CASE-EXAMPLES  When tossing a coin (could be fair coin or biased), the possible outcomes: pheads; ptails  Q: When the total amount of info. is maximized?  H=-[Pheads×LN(Pheads)+ (1-Pheads)×LN(1-Pheads)]  Pheads=x  f’(x)=0→LN(x)+x*1/x+(-LN(1-x)-(1-x)/(1-x)]=0 →f’(x)=LN(x)-LN(1-x)=0 →x =1-x → x=1/2  Pheads=Ptails=1/2  The maximum possible information occurs when the two alternatives have equal probability. 11 The maximum possible information occurs when all the alternatives have equal probability. Information decreases from maximum as the difference between the probabilities of the events is increased. The reduction in information from the maximum due to unequal probabilities of events is called REDUNDANCY. Percentage of redundancy is calculated according to the following formula: 12 (2) NON-EQUIPROBABLE CASE: IN-CLASS EXERCISE  Message probabilities  p1 = 0.25  p2 = 0.25  p3 = 0.45  p4 = 0.05  What is total information H? H = [ 0.25(2.0) + 0.25(2.0) + 0.45(1.15) + 0.05(4.32)] = 1.73 bits  In this way, we can quantify differences in difficulty (in terms of the amount of information processed) between different tasks.  In general, reaction time is a linear function of information in bits (Hick’s law-will be introduced soon). 13 CONTENTS Introduction to Information & Info. Quantification Human Information Processing (HIP) Model Four Stages of HIP Model Mental Workload and Evaluation Methods Summary 14 HUMAN INFORMATION PROCESSING: AN EXAMPLE 13 HUMAN INFORMATION PROCESSING-HIP MODEL 1. Sensory stimuli entering short-term sensory store where they are transformed into a form that the perceptual processes within the brain can understand 2. Once perceived, processed stimuli are transferred to working memory (WM) 3. WM interacts with long-term memory for active manipulation of information 4. Based on the information in memory, decisions are made, in order to determine responses 5. To execute the responses 6. Feedback loop Wickens' Model of Human Information Processing 16 Wickens, C.D., Engineering Psychology and Human Performance, Harper Collins, New York, 1992. CONTENTS Introduction to Information & Info. Quantification Human Information Processing (HIP) Model Four Stages of HIP Model Mental Workload and Evaluation Methods Summary 17  HIP MODEL (STAGE 1) COVERED IN TOPICS 3/4 Wickens' Model of Human Information Processing 18 Wickens, C.D., Engineering Psychology and Human Performance, Harper Collins, New York, 1992.  HIP MODEL (STAGE 2) Wickens' Model of Human Information Processing 19 Wickens, C.D., Engineering Psychology and Human Performance, Harper Collins, New York, 1992.  WHAT IS MEMORY?  Human ability to encode, store, retain and subsequently recall information and past experiences in the human brain.  It can be thought of in general terms as the use of past experience to affect or influence current behaviour.  The memory can divide into three major types: sensory, short-term, and long-term memory 20  TYPES OF MEMORY Memory Short Long Sensory Term Term Retain impressions of sensory info. Even after original stimulus ceases Capacity for holding a small amount of info. Readily available for a short period of time Memory that can last as little as a 21 few days or as long as decades. TYPES OF MEMORY (Cont.) 7±2 Miller’s Magic Number (1956) 22 WORKING MEMORY (SHORT TERM MEMORY)  Definition  store for information being actively processed  Examples of WM/STM use  telephone number to be dialed 8 4 6 7 3 5 9 0 2  observed stimulus and standard stimuli Red Blue ? Compare with Green Yellow 23 SHORT TERM MEMORY (Cont.)  Memory span  The longest list of items that a person can repeat back immediately after presentation in correct order on 50% of trials Memory span challenge: a volunteer is needed! Link: http://brainscale.net/memory- span/training  Miller observed this span to be approx. 7±2 (Miller’s Magic Number) for adults  Memory span not limited in terms of bits but rather in 24 terms of chunks  SHORT TERM MEMORY (Cont.) o Chunk: The largest meaningful unit in the presented material that the person recognizes (depends on what you already know)  digits (0, 1, 2,...)  B M W R C A A O L I B M F B I→ BMW RCA AOL IBM FBI  Numbers like 1945, 911 can be associated with important events;  names (“Bill”, “Sue”, “Nan”, etc.)  digit sequences (737-, 752-, 745-, 754-,...): 3-4 digit chunking is ideal for encoding unrelated digits *RCA: Radio Corporation of America; AOL: Application Object Library  Human memory capacity: 7±2 chunks ; Very significant human limitation→ has design implications 25  MEMORY AS ART !  Memory not a static entity. It can be honed by practice.  Mnemotechnics: Used to organize memory impressions, improve recall, and assist in the combination of ideas.  Techniques involve Architectural Association (Method of Loci), Graphical Mnemonic, Textual Mnemonic etc.  Improving memory from a Student’s perspective:  Rephrase and explain  Be emotionally involved  Schedule and read in chunks  Use visual aids/word associations  …… 26  HIP MODEL (STAGE 3) Stage 4 will be covered in the next topic CH7 Wickens' Model of Human Information Processing 27 Wickens, C.D., Engineering Psychology and Human Performance, Harper Collins, New York, 1992. DECISION MAKING  Characteristics of a decision making situation  select one from several choices  some amount of information available  uncertainty  relatively long time frame 28  CLASSICAL DECISION THEORY  Normative Decision Models (Central Concept of Utility- Overall value of a choice): What a rational decision maker should do  Expected Value Theory (EVT)  Probability of outcome, given decision  Value of outcome, given decision  Choose the action with max. weighted value (the highest expected value)—Focus on objective values  An Example Gambling: 29 (1) get $12 when a die comes up with a 6 and 0$ otherwise; vs. (2) get $12 when a die comes up with a 6 and pay 3$ otherwise  CLASSICAL DECISION THEORY (CONT.)  Subjective Utility Theory (SUT)  People do not always make decisions based purely on rational, mathematical calculation. Instead, decisions are influenced by factors like personal experience, emotions, and risk tolerance  People's preferences, perceptions, and attitudes toward risk and outcomes are subjective  Individuals assign their own subjective probabilities and utilities to outcomes rather than relying on objective values  Could be biased by recent experience…  An example 30  DESCRIPTIVE DECISION MODELS  Background information based on experiments  Humans violate key assumptions of the normative models  Don’t explicitly evaluate all hypotheses: People rely on simpler and less-complete means of selecting choices  Early descriptive model: Simon’s Satisficing (Simon, 1957)  Instead of assuming that individuals always seek to maximize utility, people often aim for a "good enough" solution  Why?  Bounded Rationality: Limited information-processing capabilities→ Impossible to evaluate all possible options.  Satisficing: Instead of optimizing, individuals search through available options and choose one that meets a certain threshold of acceptability-"satisficing” 31  NATURALISTIC DECISION MAKING: RASMUSSEN’S SRK MODEL (SKILL, RULE, KNOWLEDGE) ⚫ Skill-based ⚫ Knowledge-based experts, very experienced decisions makers  novices or novel/complex problems automatic, unconscious  knowledge-intensive  analytical processing requires less attention  high attentional demand errors: misallocation of attention  errors: limited WM, biases ⚫ Rule-based 32 experienced decision makers if-then rules errors: wrong rule FACTORS AFFECTING DECISION MAKING  Amount/quality of cue information in WM  WM capacity limitations  Available time  Limits to attentional resources  Amount and quality of knowledge available  Ability to retrieve relevant knowledge 33 RESPONSE SELECTION: REACTION TIME (RT) Definition: time it takes for a human to respond to a stimulus 1. Simple RT: 1 stimulus, 1 response 2. Choice RT: 1-to-1 match (n stimuli--n responses) 3. Choice RT: 1-to-n match (1 response for n stimuli) 34 RESPONSE: SELECTION  Phenomenon  Response time proportional to stimulus uncertainty  Example  J. Merkel discovered that the response time is longer when a stimulus belongs to a larger set of stimuli Source Figure. Mean choice-reaction times from Hick and Merkel plotted against ln(n+1). 10.3390/e12040720 35  Explanation  Hick Hyman Law (Hick’s Law) HICK-HYMAN LAW Response time is proportional to stimulus uncertainty OR, equivalently stimulus information content. 36  HICK’S LAW (HICK-HYMAN LAW)  RT is determined by H – the total amount of information transmitted – regardless of the factors causing it to vary  Expressed formally, the Hick-Hyman Law is: RT = a + b[T(S,R)]  where T(S,R) is the information transmitted;  a is a constant representing sensory and motor factors, and is equal to simple RT (no choice);  b is the time to transmit 1 bit of information (varies with individual & system factors) 37 HICK’S LAW (HICK-HYMAN LAW)  Other factors can also exert a strong influence on RT and on overall rate of information  Stimulus type  Strength of the sensory evidence (d’)  Speed-accuracy tradeoff  Practice  …… 38  RESPONSE: STIMULUS TYPE Data from Neuroscience & Medicine 1(01):30-32  Phenomenon  Simple RT to auditory stimuli faster than visual  Males have faster reaction times when compared to females  Explanation (Auditory stimuli has: )  Fast processing time in the auditory cortex  The fastest conduction time to the motor cortex 39  Therefore faster reaction time & quick muscle contraction Jose Shelton and Gideon Praveen Kumar, 2010. Comparison between Auditory and Visual Simple Reaction Times. Neuroscience & Medicine 1(01):30-32 RESPONSE: STIMULUS INTENSITY  Phenomenon  Simple RT inversely proportional to stimulus intensity  Example  Cockpit master warning  Explanation  Salience (being noticeable)  Countermeasures  Control stimulus intensity 40  RESPONSE: PRACTICE  Phenomenon  Response time inversely proportional to practice  Example  Trained radar operator faster at detecting and identifying targets  Explanation  Automaticity of responses  Countermeasures 41  Provide training  RESPONSE: SPEED-ACCURACY TRADEOFF  Phenomenon  Response time inversely proportional to required accuracy  Explanation  Speed-accuracy tradeoff  Countermeasures  Reduce accuracy requirements 42  Enhance operator accuracy through training & other means RESPONSE: OTHER FACTORS  Stimulus complexity  Stimulus-response compatibility  Temporal uncertainty  Workload  Repeated stimuli  Task interference/workload  Motivation  Fatigue  Environmental variables  etc. 43 CONTENTS Introduction to Information & Info. Quantification Human Information Processing (HIP) Model Four Stages of HIP Model Mental Workload and Evaluation Methods Summary 44  MENTAL WORKLOAD (COGNITIVE WORKLOAD) Changing Nature of Work  Definition  “amount” of mental resources required by a set of concurrent tasks and the mental resources actually available  Examples  Low: driving on a straight rural road  High: driving in heavy traffic: on wet, slippery road surface  reading map  dialing cell phone  talking with passenger  worrying about fuel quantity 45  Significance  high workload  stressful, and/or poor task performance BASIC APPROACHES TO EVALUATE MENTAL WORKLOAD  Subjective rating measures  Uni-dimensional scale (Overall workload scale)  Modified Cooper-Harper (MCH) rating scale  NASA Task Load Index (TLX) Multi-  SWAT (Subjective workload assessment technique) dimensional  Task performance measures  Primary task method: e.g., driving task  Secondary task method: e.g., driving task plus mental arithmetic  Psychophysiological measures  Electroencephalography (EEG)  Eye movement (pupil diameter, blink rate etc)  Heart rate 46 NASA-TLX (TASK LOAD INDEX)  There are SIX Sources of Workload TASK LOAD FACTORS OUTCOMES, & PERSONAL COSTS 47 NASA-TLX (TASK LOAD INDEX)  Rating scale(0-100) 48 NASA-TLX (TASK LOAD INDEX)  Implementation steps 1) Each source of workload is Compared Pairwise against the others to give a Rank Order (0-5) 2) Subjects rate each EVENT by giving a 0-100 score for each source 3) These values are multiplied by the RANK and the total is divided by 15 to get the Workload Score on a 0-100 Scale 49 CASE STUDY: APPLICATION OF NASA-TLX Step 1: 50 PD: Physical demand; MD: Mental demand; TD: Temporal demand; EF: Effort; OP: Performance; FR: Frustration level  CASE STUDY: APPLICATION OF NASA-TLX Step 2: 51  CASE STUDY: APPLICATION OF NASA-TLX Step 3: Rating calculation and normalization Task 1 (ISI=500ms) Task 2 (ISI=300ms) 90 0 300 40 90 120 640 15 42 52 PRIMARY TASK METHOD  Use operator behavior to determine workload  usually accuracy and reaction time, tracking performance as the performance measures  Assumption: as task load increases, the additional demands on mental capacities result in a degradation in performance How well are you How well are you Driving? Flying? 53 Notice that its not so easy to specify ‘well’ in complicated performance environments. SECONDARY TASK METHOD  Multiple tasks (primary task+ secondary task: auxiliary and non-instrusive) at the same time  Subjects are instructed to maintain consistent performance on the primary task regardless of the difficulty of the overall task  Secondary tasks include, but are no means limited to:  Rhythmic tapping / Verbal shadowing;  Spatial reasoning / Critical instability tracking task;  Compensatory or pursuit tracking tasks/ …. 54 Psychophysiological Reflections of Mental Workload Electroencephalography (EEG) Pupillometry/ Eyetracking Eyeblink & Reflex Electrocardiography Modification (ECG) 55 CONTENTS Introduction to Information & Info. Quantification Human Information Processing (HIP) Model Four Stages of HIP Model Summary 56 SUMMARY Equation to quantify information Equal probability case Non-equal probability case HIP model with 4 stages Sensation and perception, memory, decision, response Hick-Hyman’s law Mental Workload and Evaluation Methods 57 Shuping Xiong (PhD, Professor) Human Factors and Ergonomics Lab Department of Industrial & Systems Engineering Korea Advanced Institute of Science and Technology Lab Homepage: http://hfel.kaist.ac.kr/ Email: [email protected] 58