Human Factor

Course Description

This course provides an introduction to human factors research and applications with emphasis on mature areas such as sensation and perception and manual control. Each class will introduce some concrete human factors problem and explore theory and application relevant to solving it. The term long conceptual design assignment is intended to help maintain focus on applications to design.

Week1

Learning Objectives:

Theoretical understanding: Develop an understanding of the historical context, disciplines, and schools of thought that led to the development of the current field of human factors engineering/psychology; Develop an acquaintance with basic elements of scientific method and decision making including tests of hypotheses, dependent and independent variables, and inferential and descriptive use of statistics.
Apply theory and skills: Use scientific methods including selection of measures and experimental tasks to evaluate the utility of a variety of pointing devices
Proficiency in information related skills: Learn how to conduct a critical incident-based evaluation of an interactive system and to perform a standard task analysis for such a system

Research Methods

IV & DV
Cause vs. Chance
Descriptive vs. Inferential Statistics
Null hypothesis(H0) & Experimentalhypothesis(H1)
Hypothesis Testing
Statistical Significance & Practical Significance
Graph & Interpetation

Goals of System Evaluation

Functionality
- Can it do what it is supposed to do?
Usability
- Does it make the task easier?
Diagnosticity
- Does it pinpoint what is wrong?

User Population
- Who are they?
- What are their goals?
- What do they already know how to do?
- Ask, don’t assume!
Research Objectives
- Generalizability
- Precision
- Realism
Usability Testing
Design an Experiment

Two Models for Human Factors

System component (computer)
Embedded organism (cybernetic)

Task Analyses

为了研究如何使系统更好地和人的能力相匹配所进行的一种描述人机交互的方法 - Sequential 任务的顺序和不同任务在时间序列上的关系 - Procedural - Therp - Hierarchical 描述一个大的任务如何由子任务组成以及这些任务又是如何联系起来体现其功能的 - HTA - GOMS - Cognitive

Evaluation

Types of Evaluation
- Laboratory Studies
- Field Studies
- Participatory Design/Rapid Prototyping
- Brainstorming
- StoryBoarding/Wizard of Oz
- Workshops/Role Playing
- Walkthrough/Talkthrough
Observation Evaluation

Function Allocation

每个功能由人来实现还是系统来实现 - Mandatory - Relative value - Cost based - Cognitive/affective

Week2 Sensation

Learning Objectives:

Psychophysical methods & the problem of separating bias from measurement
Signal Detection: Perform specific calculations
- Conceptual model (associating hits, FA’s, misses, & CR with areas under the curves)
- Computational understanding (interpreting problems, using table, & finding d’)
- Vigilance - what the vigilance decrement is and how signal detection has been used to better understand it
DESIGN: understanding of the cognitive information capabilities of humans General knowledge- wide range of sensitivity, approximately logarithmic, etc.
Place encoding of frequency and its consequences for:
- Masking
- Threshold shift & frequency related loss, etc.
Measurement issues
- Loudness & equal loudness contours
- SPL meters
- Articulation index, SIL, and other speech & noise issues
- Noise & annoyance

Classical Errors

Error of Habituation - Keep on saying the same thing Error of Anticipation - Shift to new response

Signal Detection Model

Vigilance

警报的设计必须建立在对人类的听觉加工充分了解的基础上

Near threshold signals
Low rate of occurrence
Extended Watch
Inspection(definitetrial)
Free response time arbitrarily brokeninto intervals
Successive(noavailablestandard)
Simultaneous(signal to standard comparison)

Fatigue

Sustained attention leads to fatigue
Load on working memory to keep target in mind depletes resources

Expectancy

Remedies

Sound

Patterns of rare faction/compression of air

Physics
- Intensity = amplitude
- Frequency = cycles(hz)
Perception
- Loudness
- Pitch

强度决定响度
频率决定音调
位置决定听觉定位
品质由频率及其掩蔽来决定

DECIBEL

dB(a): dB weighted by threshold equal loudness curve
dB(c)/dB (spl): dB with no weighting
dB(d) weighted by equal annoyance curves

Encoding Pitch by Place

Auditory Masking

the presence of tone that inhibits the perception of another tone that occurs before, at the same time, or after it.

Week3 Vision and Color

Learning Objectives:

Even more than place perception of pitch vision is all about relative differences (contrasts) and adaptation
The relation between what we experience and what is physically out there isn’t direct
Good Human Factors engineering requires designing so our users don’t notice

Overview of vision - Peripheral/central processing - Dark Adaptation & Illusions & their relation to the structure of the eye & vision - Visual angle & spatial frequency - VDTs, visual fatigue, & ergonomic effects - Color measurement Munsell color wheel vs. CIE

Rods and Cones

位置 location
视敏度 acuity（解析细节的能力）
敏感性 sensitivity (即使光很少，Rods也能工作)
color sensitivity （rods 是色盲）
adaption （rods 受光刺激的影响大）
diffrrential wavelength sensitivity (cones对所有光敏感，rods对红光不敏感）

对比敏感度

c = (L-D) / (L+D) ### color sensation 对单色进行设计，然后将颜色作为冗余编码信息提供 ### dark sensation 当照明条件比较差时，所有空间评率的对比度都会降低

检测

在很多系统中，影响作业绩效的重要因素是密切相关的两个过程：视觉搜索和物体或时间的检测
在视觉搜索领域，一旦某个项目被确定可能是目标，就必须对它是否是真的目标进行确认。
d' 反映了一个操作者从噪音中分辨出信号的能力，它等于好结果的数目除以所有结果的总和

Peripheral Processing

Photo receptors are interconnected and can reciprocally inhibit one another
Can be tuned for featurere cognition

Adaptation

Adaptation is a major characteristic of sensation

Week4 Reaction time

Learning Objectives:

Subtractive&additive factors analyses of choice reaction time
Limits&values of psychological experimentation
Human bottleneck in choice responses
Power law of learning
Automaticvs.controlledresponding
Information theoretic interpretations of reaction time

Hick’s Law

Hick’s Law holds that choice reaction time is proportional to log2 of the number of alternatives.（反应时间是log2N的函数） RT = a + blog2N

Conventional Controls & Displays

conventional practice in design of controls
displays Acquire familiarity with human factors design principles and heuristics

Control

Controls are used by the human operator to communicate with the machine/device in the system. It’simportantthatcontrolsservetheirfunction. Based on: - Ease of operation (considering population, biomechanics, etc.). - Nature of the task (force, precision, etc.). - Arrangement.

Basic dimensions:

Discrete (e.g., light switch) vs. continuous (e.g., dimmer)
Linear vs. rotary
Unidimensional vs. multidimensional
Isometric vs. isotonic
Plus mass, shape, range of motion, resistance to movement.

Control Features

Control resistance
- Elastic resistance: Spring loaded.
  - Resistance increases as control gets farther from neutral.
  - Gives proprioceptive feedback about control position.
  - Returns to neutral when released (deadman switch).
- Frictional resistance
  - Static friction for resting state, decreases when pushed.
  - Sliding friction not influenced by velocity or position.
- Viscous resistance
  - Increases as a function of velocity.
  - Gives proprioceptive feedback about speed.
  - Promotes smooth movement.
- Inertial resistance
  - Hard to start and stop.
  - Users tend to overshoot (revolving doors).
- Performance and resistance
  - For frictional and inertial, the JND is 10%-20% of resting state.
  - Lighter controls preferred to heavy.
  - Viscous preferred to frictional.
  - For continuous, inertial hurts performance;
  - Elastic is the best.
  - For all of these, hard-and-fast rules are not available. It’s a function of the system.
Control-display ratio:
- Ratio of magnitude of control adjustment to magnitude of change in display.
Gain:
- Responsiveness of control.

Control Panels

Location coding
- Need to be able to reliably distinguish locations.
- Vertical localization is easier than horizontal.
- Overuse of location coding is still a factor in some aircraft designs.
Labels
- Not recommended as the sole code
General rules for labels
- Locate labels systematically with respect to controls (all above, etc.).
- Make labels brief.
- Avoid abstract symbols; use standards.
- Attend to fonts.
- Position labels so they can be seen while the control is in use.
Coding of controls
- Color coding
- Shape coding
- Size coding
- Texture coding. – Coding by type of operation. – Redundant coding: Multiple dimensions
Control arrangements
- Grouping is important.
- Population stereotypes for control arrangements can be device specific
- Attend to the reach envelope
Preventing accidental operation
Specific Controls
- Hand operated controls
- Foot operated controls
- Specialized controls

Visual Displays

Display - anything that conveys information

Requirements:
- Compatibility to senses
- Language compatibility
- Right info at the right time
Types of information to display:
- Instructional
- Command - direct orders
- Advisory
- Historical/predictive
- Answers
Functions of Dynamic Visual Displays
- Continuous System Control
- System Status Monitoring
- Briefing
- Search and Identification
- Decision Making
Visual Display Technology
- Mechanical
- Electronic
- Optical Projection
General Display Principles
- Color
- Shape
- Coding
- Approximation
  - Get attention with one display – Present detailed info with another
- Integration

Principles of Display Design

Perceptual Principles
- Avoid absolute judgments
- Top-down processing （信号的显示方式尽量与人的经验相符合）
- Redundancy gain
- Discriminability
Mental Model Principles （显示方式与操作员的心理模型一致，有助于提高正确操作）
- Pictorial realism （形如其表）
- Principle of the moving part （运动一致）
Principles Based on Attention
- Minimizing information access costs （将访问信息的消耗降到最低）
- Proximity compatibility principle （接近相容原则）
- Principle of multiple resources （要同时对多种信息进行加工时，可以将信息的呈现方式区分开）
Memory Principles
- Principle of predictive aiding （预测辅助原则）
- Principle of knowledge in the world （利用知识降低记忆负荷）
- Principle of consistency （一致性原则）
Two-Valued Info
Quantitative Information
Qualitative readings
Check Reading
Situation awareness

Three Heuristics

Minimize Information
Promote Good C-D mappings
Provide Feedback

Spatial & Integrative Displays

Learning Objectives:

6 DOF & moving & orienting in 3 space
Problems with viewpoint & situation awareness
Applications VR, games, & robotics – Attitude & pose
Navigation & search
Cues to depth & distance
Use of emergent features and perceptual salience to integrate displays
TMI and need to provide context to events

6 Degrees of Freedom (6DOF)

– Position (X,Y, Z) – Orientation (Yaw, Pitch, Roll)

3d display on 2d panels

Depth is often poorly represented & less discernable than other dimensions

Configural Displays

Low level data: usually individual sensor data
High level relation: a more global and general display of what the data means
Emergent property or emergent feature: a pattern or shape that is created from the low level data, is recognisable and has meaning

Separable

Show each variable as a single output

Separable vs Configural vs Integral

Separable generally makes it easier to extract low level information
Integral Show high level information but not low level information
Configural Arrange low level data into a meaningful form,whole is greater than the sum of the parts
Configural makes it harder to extract low level information

Tracking (manual control)

Learning Objectives:

Fitts’ Law Ability to apply theory and skills: Design Use Fitts’ Law to evaluate/predict pointing performance
Theoretical understanding: Order of control

Open versus Closed Loop Systems

Tracking Terms

Control movement
Controlled element
Target
Forcing function- disturbances to target

Pursuit and Compensatory

补偿追踪与尾随追踪 - Pursuit
- Target moved - Usually more accurate - Compensatory - Target fixed - Target & control movements confounded

Fitts Law

MT(movement time) = a + blog2(2A/W)

Tracking vs. pointing

Pointing as expressed by Fitts law is a very special case of tracking. In pointing: - Stationary target - No lag - Gain is only control system parameter

Gain & C/D ratio

Gain describes the change in the controlled element (display) corresponding to a movement of the control: gain = y/x..
C/D ratio describes the movement of a control needed for a given change in the display: C/D = x/y

Order of Control

0 order: Position
- A 0 order system has no integrations between input and output
1 order: Velocity
- A 1 order system has one integrations between input and output
2 order Acceleration
- A 2 order system has two integrations between input and output

Week 8 HIP & Workload

Learning Objectives:

What is mental workload? – Subjective, performance, & physiological measures
HIP & human factors – Working memory, absolute judgment, & other aspects of the bottleneck
Mental representation & difficulty

Basic approaches to measuring mental workload

Analytic – Task difficulty
- Number of simultaneous tasks
Task performance
- Primary task
- Secondary task
Physiological (arousal/effort)
- heart rate
- evoked response amplitude
- ......
Subjective assessment
- Cooper-Harris
- SWAT
- NASA

Selective Attention

通道的选择性注意主要受下面因素的影响

Three aspects of perception

Sensory based
- Bottom-up feature analysis
  - clear stimuli/minimize sensory similarities
Memory based
- Unitization – perceive grouped features as a whole (Gestalt)
- Top-down [correct guesses & fill-ins]

working memory

working memory
long term memory

90’s model (nods to Baddeley & Schneider)

1. Central executive - Coordinate multiple tasks (OS) - Hold & manipulate info from LTM (RAM) - Control retrieval strategies from LTM (data access) - Attend selectively to stimuli (time share) - 协调两个存储子系统 2. Visual sketchpad - 以模拟的，空间的形式保持正在使用的信息 3. Phonological store - 存储以声音的形式存在的信息

Human Error and Reliability

Understanding Mechanisms underlying human error
What types of errors can be predicted?
Proficiency in information-related skills: Analysis Perform THERP analyses to predict errors for a design/task

Error

Any act with adverse consequences
An act resulting from an inappropriate intention
Keeping the pressurizer level under control

Action Schema

1. Intention - mode errors - description errors 2. Activation - capture errors - data-driven - associative activation - loss of activation - sequence error 3. Faulty Triggering - Out of sequence and “mangled” execution..

Reliability Engineering

The Key to reliability is Redundancy

Components in Series

### Components in Parallel ### Tradeoffs (Redundant or not)

How redundancy works

For the mathematics to work out the probabilities of failure for redundant components or subsystems must be completely independent

What Reliability Engineers do

The primary task of a reliability engineer is to defend redundancy against unexpected violations of independence.
In a well designed system only the human operator bridges these islands of independence

THERP

Quality control method for estimating errors
Model
- Errors: such as reading or omitting an instructional step, or choosing the wrong switch, are presumed to occur at constant rates
- If tasks can be broken down into subtasks for which errors can be predicted, then the probability of the successful completion of the overall task can be predicted
- The probability of successfully completing the task (if its something like warhead assembly) is then simply the joint probability that everything is done correctly

Points on THERP analyses

Tree is not sacrosanct but a convenient way to organize independent tasks
Probabilities for errors and recoveries should be entered into trees at level of aggregation at which independence holds
Method is ultimately simply a way to make our commonsense about the likelihood of failing more explicit

Human-Computer interaction

Understand basic assumptions and mechanics of constructing GOMS keystroke level model
Contrast the HIP vs. Ecological vision of problems in HCI
Standard visualizations and the problem(s) they solve- finding context for local views

GOMS Models

用于设计的用户绩效模型 - Goals - Operators - Methods - Selection rules

用户可以通过方法和选择形成他们要达到的目标和子目标。方法是一系列知觉的、认知的或行为操作的步骤。

步骤

列出目标和子目标
明确达到问题的方法
写出选择关系
揭示问题

Norman’s 7 Stages & design

用户导向界面设计的七阶段理论

实施的鸿沟：用户的目的和软件所支持的行为之间的错误匹配（通过好的人因学方案解决，input tracking position）
评价的鸿沟：用户期望与系统状态的不匹配 (好的说明性显示)

Mplications of working memory & absolute judgment limitations

Recognition is MUCH easier than Recall :Make the objects of working memory available to perception..

The Power Law of Practice

Improvement in performance is logarithmic in the N of trials

Mental Models

跨越实施和评价的鸿沟依赖于心理模型，好的心理模型可以帮助房主错误和改进绩效 - Allows people to make predictions about how things will work - Mental models are often wrong

Conceptual Models

使用户看不见的部分变为可见比如"房间"

State Transition Models of Devices

Speech

Theoretical Understanding: Problems in building applications using speech recognition Design
Strategies for getting good performance despite poor recognition
Theoretical Understanding: Computer supported cooperative work
Theoretical Understanding: Design Strategies for using sensors to augment human inputs and improve interaction

Basic Speech Parameters

Speaker Dependent/Independent
Size/Type of Vocabulary
Isolated word vs. Continuous Speech
Grammar/constraint
Environment/noise tolerance
Noise canceling unidirectional microphones – Quiet environments

Rec System maximizes chance of getting things right by

Restricting vocabulary
Specifying order & transitions for recognition
Associating actions/meaning with partial recognition of phrase
Use of context particularly within dialog to adjust constraints

Lombard Effect

During noise, speakers have an automatic normalization response that causes systematic speech modifications, including increased volume, reduced speaking rate, and changes in articulation and pitch.

Issues in Ubicomp

Issues in Ubicomp - Context - Uneven conditioning - Inferring user intent - System interoperation

Decision Making & Diagnosis

Learning Objectives - Theoretical Understanding: Normative vs. Behavioral theories of decision making - Models of decision making in diagnosis Name & illustrate standard fallacies in decision making

Rational Decision Making

A rational decision maker is one who chooses the alternative which maximizes his expected utility.
A rational decision maker is presumed to maximize her Subjective Utility which is likely to be some function of objective

Prospect theory

Loss hurts more than Gain helps （抛硬币，正面赢20，反面输 10，大多数人选择不玩）

Base Rate Fallacy

Undervalue base rates!!

Behavioral Decision Making

Treate extreme values as more moderate （感知不到极端数值，就比如考试明明只剩三天，但还觉得时间很充裕不好好看hf）
Imperfections in memory （记忆缺陷，不能收集到所有过去的信息帮助做决策，比如期中复习hf就十分紧张期末还是这样）
Inability to do complex math in our heads （做不了复杂算数）

Gambler’s Fallacy

Error: treating independent events as though they were dependent

Availability heuristic可得性偏差

人们做决策总会基于 avaliability & imaginability

Imaginability

Confirmation bias

确定偏差:简而言之就是听不进新的观点，无论怎样论证都是认为自己原本认为的是对的，本质还是overconfidence 例如：给一些本身对于某件事有观点的人接受正反两面信息，人们通常都只注意到支持自己观点的理论，暗中反驳不符合自己观点的理论

Representativeness bias 代表性偏差

人类在对事件做出判断的时候，过度关注于这个事件的某个特征，而忽略了这个事件发生的大环境概率和样本大小。例如，你看到一家公司连续3年利润都翻番，然后立即对它的股票做出判断——买！错在代表性偏差。连续3年利润翻番，是一个好公司的代表性特征。但这并不意味着这家公司真的就是一家好公司，这家公司还有好多信息都被你忽略掉了。比如说，业绩可能是有意调整出来的；再比如说，这家公司未来的盈利机会消失，业绩不能持续。

Anchoring锚定效应

人们在对某人某事做出判断时，易受第一印象影响从而先入为主