Exam 3

Decision making

Select one option from many available options, has some information about all the options, has relatively long time frame, there is some uncertainty about which is the optimal choice > pop machine

Problem solving

An ill defined problem that may include complex and dynamic information, requires the generation of possible solutions, may include interaction with other time stress and high risk, multiple uncertain outcome with different costs and benefits, implementing solution monitor and correct courses as needed

Macrocognition

Higher level mental processes that require information processing

Include situational awareness, decision making, problem solving, & metacognition (knowing what you know)

Normative decision making: expected value model

EV = value * probability

Assumes decision makers are rational and compute “value“ of competing options

Problems with expected value model

It doesn’t predict actual decisions, monetary value doesn’t match perceived value, rarely are both values and probabilities well-defined - most decisions don’t have those clear options, or clear probabilities

Expected utility model

Replaces “value” with “utility” which is subjective value, is the subjective utility of the option multiplied by its subjective probability, better captures human decision making, but not entirely. Replaces probability with user’s subjective probability

> $10 is relatively fixed value & CD of journey’s greatest hit is not

Normative decision making: Multi - attribute utility theory

Recognizes each option has multiple outcomes (some costs and some benefits), Utility of each option is sum of (probability value) of each benefit + (probability * value) of each cost.

Lots of subjective estimation

Descriptive decision making

How people actually make decisions (not how they should)

Heuristics

Theses are shortcuts and rule-of-thumb, that decision makers use to simplify the decision making process. These are typically effective but leads to predictable errors

Most of the time they work

Biases

Byproducts of our cognitive systems, not necessarily unrelated to heuristics but are less effective and more about managing our effort and emotions

Heuristics & biases

While acquiring & integrating cues, cue primacy, cue salience, both related to limited cognitive resources

Cue primacy

Anchoring, first cues are weighted heavily, first thing you notice

Cue salience

Easier to detect cues are weighted heavily, what jumps out to you first

Interpreting & assessing cures: Availability heuristic

Estimate the likelihood of an even through evaluating our memories, we use our memories to judge the frequently of something happening. Assumes availability in memory predicts probability in world. If we never seen it happen then we assume it will never happen in that moment

Hypotheses that are more available in memory are deemed more likely

Interpreting & assessing cures: Representative heuristic

Assumes how well a cue represents an event predicts how likely it is that event, likelihood of a situation is estimated by how representative (on the surface) it appears of that category

> Miriam story, which is most likely

Interpreting & assessing cures: Confirmation bias

Search for confirming evidence you are looking at the best cue, overconfidence in a cue, cognitive tunneling

We tend to search for evidence that we are correct

Overconfidence

Finding evidence that only supports your point of view

Cognitive tunneling

Refusing to consider other options and continuously only picking one choice even when you have evidence that its wrong, they still pick that option

Simplicity seeking: minimizing possible losses/pain

At the stage of planning & choosing. Eliminate any option that might lead to unacceptable losses

Simplicity seeking: Satisfying

At the stage of planning & choosing. Accept the first option that exceeds/meets minimum needs. Decision making the first option that’s good enough

Choice aversion: default heuristic

Accept the default settings, people don’t usually go out of their way to change things in the settings they just accept how things are, so designers should have good/default things

Choice aversion: Delay

Waiting for more information to make a decision. Depends on the situation if that a good choice or not but as a leadership decision that’s a bad one

Generate too few options

We generate too small/ small number of action plans to choose from. We chose options based upon recency of last use, frequency of use, if-then rules (if this happens then that…)

Framing bias

At the stage of planning & choosing. decisions framed as gains makes use risk-averse. Decisions framed as loses makes us risk seeking.

Sunk-cost effect

At the stage of planning & choosing, may be related to risk seeking. Weight past losses in future outcomes. When choosing future actions, we weight past losses. Being worried about pasted decisions that can’t changed when thinking about future decision. I don’t want to wast all the money/time I spent on this so I better use this!

Planning bias

At the stage of planning & choosing. We assume best case scenario and don’t consider that things will go wrong. Ass 20% of anytime estimate because there will be issues that will come up

> give yourself more time

Expertise & problem solving: Rassmussen’s SRK model

Skill, Rule-based, Knowledge-based

Knowledge-based decision making (SRK)

Typical of novice decision makers, effortful/ slow analytic process of available information.Can quickly become overwhelming for decision maker. Too much information that even an expert to weigh. Limited cognitive resources.

Rule-based decision making (SRK)

Typical of people with some familiarity but not true expertise, they have used knowledge and experience to develop decision rules, learned some decision rules. Heuristics are a common example of this level

> Following a recipe having some knowledge but not typical of a true expert

Skilled-based decision making (SRK)

Relies on procedural memory, appears intuitive to observer, experts recognize patterns of cues and respond. Is fast and largely automated as needed. Experts are able to revert to knowledge-based analysis when the situation is unusual, when there is time available, and that there is well-defined cues to evaluate

Experts differ from novices because they know which cues are most important and incorporate them into their responses

Task (re)design

Improve “choice architecture” for decision makers, make information more accessable & easy to see > Limit the number of options available, present data in a linear & comparable formate, present choices and consequences of each “real“ and concrete

Task (re)design: Improve display

Better able to understand what's going on & if something is wrong, have the display give more information

Make them integrated with one another (to reduce their numbers)

Proceduralization: Task (re)design

Include decision support tools > automation when appropriate, exert systems or other aids = use of checklists, decision tables/trees.

Use standard procedures and formats to guide decisions

Improve situational awareness

User better understands the “real“ problem, the display should help users notice change > especially when there is a dangerous change

Displays should aid in prediction of future state of the system

Organize displays & information according to the users goals

Keep the operator “in the loop” & engaged

Don’t relay too much on automation, make the system variables easy to understand, display all of the relevant system variables > no just okay / not okay

Improving diagnosis & problem solving

Better displays of system states, encourage evaluation of multiple alternatives > what else could be the cause of the problem, display evidence of non-events that are meaningful, simulations may possible solution

Controls

Involves response selection & execution

Response Selection

The goal of design is to aid in ease & speed of choosing the correct control action

Unexpectedness causes delays in responding

Controls that match expectation capitalize on top-down processing & results in quicker responses

Controls relates to “labels” from the previous chapter

Response execution

Don’t just define the choices

Helps define the “correct” choice

Response execution: Discrete controls

These have separate independent settings

Buttons, switches, levers

Response execution: Positioning controls

Point & select controls = moving something to hit the target

a computer mouse, involves moving a cursor to a target

Response execution: Tracking Controls

Involve continually operating a control to follow a moving target, like positioning but then its a continuous closed loop of a target

Like a Mario Kart

Principle of controls design- attention related (6)

Proximity Compatibility

Avoid Resource competition

Avoid absolute judgment limits

Exploit redundancy gain

Make controls discriminable

Make controls accessible

Proximity Compatibility

Control should be physically near the display/ what it controls, proximity can be shown in multiple ways > with location, color, boundaries

Avoid Resource competition

Divide overlapping responses into different cognitive/motor domains

Voice & manual responses

Hands & feet responses

Verbal & spatial tasks

Make controls accessible

Anthropometric concern > Measurement of humans/ Make the controls fit human body

Support “blind operation” > Make the controls so you can use them without looking at it you don’t have to see the controls to be able to use them

Make controls discriminable

Put different labels on the controls (the labels could rub off)

Making controls identifiable easily (support blind operation)

Tactual coding > Shape: shape of controls related to their purpose

Texture: having bumps/ ridges

Size: Varying sizes of controls can relate to function

Exploit redundancy gain

Make controls distinct in more then one way

Label them, add unique texture, different location/shapes

Avoid absolute judgment limits

Don’t differentiate multiple controls on a single dimension

Use of detents can create distinct controls positions = activation of controls van use detent (clicks) to create distinct positions

Principles of control design: memory related

Movement compatibility

Location compatibility

Knowledge of the world

Be consistent

Knowledge of the world

Labels to reveal the purpose of the control

Represent the status of the control (on/off, etc.)

Be consistent

Layout & movement of controls should be consistent

Consistent within a design & across similar products

Location compatibility

Controls should be near the entity they controls

Controls should be near the display of the variable they control

Movement compatibility

Control movement should match variable’s movement

Controls should match population stereotypes

Principles of control design- Response selection related

Avoid accidental activation

Hick-Hyman law

Decision complexity advantage

Fitt’s law

Provide feedback

Avoid accidental activation

Make them unlikely to be hit by accident by putting them in a safe location & orientation

Recessed/ shielded to physical protect them

Interlock &/ sequences operation: have two actions that must be done to activate it

Resist, delay, confirm: Requires more effort than an accident would so it forces the user to have to wait figure out if it should be activated

> Con is that is has a slower reaction time

Hick-Hyman Law

Relates to choice reaction time

Increasing the number of choices increases reaction time logarithmically

The more options there are for a single control, the slower its use

Morse code clicks are faster because there are few options

Decision complexity advantage

A small number of complex responses can be executed faster than a long string of simple responses

Hick-Hyman law losses its advantage when complexity of task requires lots of control operations

> Typing is more complex but faster than morse code

Fitt’s law

Related to time required to move an entity onto a target

Time increases by a constant when

the size of the target is halved or the distance from target is doubled

Provide feedback

Should be immediate > minimize delay between control & feedback

Make the feedback salient > shouldn’t be masked

Provide feedback: Visual

a sound to indicate a change in the system

Provide feedback: Auditory

A should indicate a change in the system

Provide feedback: Kinesthetic/tactile

Feel movement of control

Provide feedback: Current state feedback too

Can be important to indicate an ongoing state of the system > downloads

It can be visual/tactile rather then auditory

Discrete controls

Normally a switch that selects between separate states/settings

Dichotomous > on/off switches like lights

May have multiple settings > channels select switches

Design considerations for discrete controls: Feedback

Changing state: Tactile (click) & either auditory or visual is good

Continuous state: depressed buttons or switch position is good they are usually overlooked

Touchscreens are poor at giving feedback

Design considerations for discrete controls: Size

Smaller size for controls saves space & travel time

BUT it allows for blunder/touching something else on accident

Design considerations for discrete controls: Labeling & discriminability

Assume the user is a novice & label things unambiguously

Exploit redundancy in labeling to aid discriminability

Proximity > Labels should be proximal to controls & having labels on controls is good but it can wear off

Discrete controls - fixed vs moving pointers

Normally a switch that selects between separate states/setting

Moving pointer

Better representation of the system '“moving” to a new state, having a raised pointer supports blind operation

Fixed pointer

Can be used for less space

Keyboards > numerical data entry

specific collection of discrete control

Keyboards: linear display > 12345678

Fits neatly onto a typewriter keyboard, too much finger reach for extensive data entry

Keyboards: Telephone displace > 123 on the top line

Determined for bell telephone at MIT, best fit the population stereotype

Keyboards: calculator displace > 789 on the top line

Designed on the first consumer by Texas instruments

Keyboard linguistic data entry: QWERTY keyboard

Was originally designed to avoid mechanical constraints, became standard and has persisted until today

Keyboard linguistic data entry: Dvorak keyboard

Well known alternate design to maximize typing speed by 5-10%, hard to learn if know QWERTY because its uncommon

Keyboard linguistic data entry: chording keyboard

One key for each finger > no hand or finger movement required, much faster data entry less repetitive motion injuries

Different finger combinations represent different letter > requires extensive training & practice, no labels/ templates can be used easily

Positioning controls tasks

To move some entity cursor onto some target

Direct positioning controls

Moving the entity directly with operator’s hand/fingers

> Apple pen/touch screen

Indirect positioning control

Position of operator’s hand/fingers corresponds indirectly with the entity’s movement

> mouse/track ball

Speed-accuracy trade-off: design considerations

Movement time relates to both distance & target size

Fitt’s law: same increase in time when target is half the size or the distance is twice as far > reaction increases with larger distance/smaller target

Gain: design considerations

Output/Input: How much you get out vs how much you put in > ratio of change in output/ change in input

Low gain

accurate fine tuning but slow travel > low output, high input (turn steering wheel a lot to the tires)

High gain

Fast travel but instability when fine tuning

High output, low input > problematic for fine tuning

Continuous tracking controls - task

Moving an entity to match a moving target (continually/periodically)

Pursuit tracking

Trying to keep entity on moving target that you can see, often you have the ability to predict where the target is going, Moving vehicle on a road (road is moving)

> driving

Compensatory tracking

Simply responding to deviation from a (often fixed) target, often your process is moved away from the target & you react

> following a compass heading or sailing towards a fixed point

Closed feedback loop

> Operator notes an error (something comes up and have to react) > Force is applied to the control device > control device alters the system’s progress > system output is displayed for the operator to monitor and continues >

= When driving something comes up > user starts turning the steering wheel > the steering wheel turns the tires > the road is turning

Measurement of errors

At a moment in time, can measure distance of operator from target, across time these samples of error can be averaged together

> Measure how much time you are on the target, how much error deviation from the target

Closed loop instability

Overcorrections / pilot induced oscillation, particularly active in high gain systems

Closed loop instability causes: High gain

Small changes in control creates large changes in system, overcorrections are common

Closed loop instability causes: high bandwidth

Bandwith: is the number of target changes/sec (Hz), humans can respond about 1 cycle/sec, 0.5 is much better

> number of changes you have to make per sec, increase number of changes

Closed loop instability causes: system lag time

Excessive time between control and system response, due to complex/ otherwise sluggish system, slow feedback leads to oversteering

> system doesn’t react right away to control

Closed loop instability solutions: lower bandwidth

User can slow system, user can strategically ignore some changes & respond to a few number of them (aim for distant target)

> Slow down, react to fewer changes

Closed loop instability solutions: predictive aiding for sluggish systems

Tells you what will happen in the future

Closed loop instability solutions: open loop strategy

Stop responding to moment by moment deviation and aim for a final state

initiate controls that will lead to the desired final state without regard to fluctuation en route

Requires knowledge of what controls will lead to desired output

> Pick one target and get there (don’t react when you shouldn’t), don’t react & reuse, wait for system to react before you do