DECISION-MAKING

C h r i s t o p h e r Ta n
PSY 340 c h r i s t o p h e r. t a n @ h e l p . e d u . m y
OVERVIEW

• Decision-making approaches
o Normative vs. Descriptive
o Naturalistic

• Problems in decision-making
o Biases, anchoring, overconfidence

• Experts vs. Novices

• Improving decision-making
DECISION-MAKING

• Decision-Making (DM)
o Choosing between several alternatives with some uncertainty of which choice is best

• Human factors attempts to improve DM via changing environment

o E.g., aids, info displays, etc.
o Need to understand how decisions are made, info used, decision tasks that are
difficult, etc.

• Examples:
o 30% of all weather-related aviation accidents is due to VFR pilots flying into poor
weather conditions (VFR: pilot licensed to only fly under visual flight rules; clear
visibility)
o Driving? Firefighting? Medical officers?
DECISION-MAKING APPROACHES

• How do people make decisions? → defined by approaches

Decision-making approaches:
• Normative
o What a person should do
• Descriptive
o What a person actually does
• Naturalistic
o What people do in complex real-world settings; experienced decision-makers
DECISION-MAKING APPROACHES
Normative DM
Normative Decision-Making
• What should a person do given the facts?
• Rational and objective evaluation of evidence
• Empirical; “scientific method”
• Assumes rationality, absence of constraints (i.e. time pressure, stress, memory)

Process/Components of Normative DM:

1. Choice alternatives → 2. Possible events → 3. Possible outcomes → 4. Probability of outcomes

Alternatives Possible Events Possible Outcomes Probability of Outcomes

Option 1 Event 1 Outcome 1;1 % (1;1)?

Event 2 Outcome 1:2 % (1;2)?

Event 1 Outcome 2;1 % (2;1)?

Option 2 Event 2 Outcome 2;2 % (2;2)?
DECISION-MAKING APPROACHES
Normative DM
Example:
You are running late for your PSY 340 lecture. Worrying that you may miss out on an amazing lecture, you do
not realize that you are driving on the wrong lane. You suddenly realize that in order to make the correct turn-
off, you need to quickly cut across two lanes. However, you may not be able to see oncoming vehicles across
multiple lanes. If you cut across, you risk a collision. But if you don’t, you’ll be even later than you thought.

Alternatives Possible Events Possible Outcomes Probability of Outcomes

Oncoming vehicle Accident; Die Miss lecture %?

Cut across
No oncoming vehicle Safe; On time for lecture %?

Oncoming vehicle Safe; Late for lecture %?

Don’t cut across
No oncoming vehicle Safe; Late for lecture %?
DECISION-MAKING APPROACHES
Normative DM – Expected Value Theory
Expected Value Theory (EVT):
• Model that combines value (utility) & probabilities to identify best decision

Expected Value (EV) = [Value (Outc. 1) x Probability (Outc. 1)] + [Value (Outc. 2) x Probability (Outc. 2)]

Example: Gambling EV = [V(W) x P(W)] + [V(L) x P(L)]

Wheel of Fortune possibilities: 1 to 6 EV = (8 x 1/6) + (-2 x 5/6)
Pay $2 to play EV = 1.33 - 1.66
If pointer stops at 6 → Win $10 EV = -0.33 (average loss of $0.33)

BUT…
• Poor representation of human DM
o People are generally poor decision-makers
• Assumes objective quantifiability
o Many decisions involve subjective utility evaluations
DECISION-MAKING APPROACHES
Normative DM – Subjective Expected Utility
Subjective Expected Utility (SEU):
• Acknowledges that values of outcomes are complex (subjective; non-quantifiable)
• Comparing monetary gains/losses is objective
• Complex choice alternatives → value depends on individuals’ goals
• Similar to EVT, but subjective utility is multiplied by probability
DECISION-MAKING APPROACHES
Descriptive DM
Descriptive Decision-Making
• Explains discrepancy between normative DM and actual human behaviour
• Normative DM → not realistic
o In reality → limited time, information, cognitive resources, etc.
• Descriptive DM → decision strategies & heuristics used to reduce cognitive demands & speed up DM

Decision Strategies
• Strategy selection → trade-off between accuracy and effort Decision Strategies
• Payne et al. (1988)
o 2-3 alternatives → compensatory
o 6-12 alternatives → non-compensatory Compensatory Non-Compensatory

• Normative • Descriptive
• Evaluates all relevant info o Elimination by aspects
• Considers attribute trade-offs o Satisficing
• I.e., Positive attributes can
compensate for negative ones
DECISION-MAKING APPROACHES
Descriptive DM – Non-Compensatory Strategies
Elimination by Aspects (Tversky, 1972)
• Eliminate options that are least attractive based on attributes (aspects)
• Options are eliminated in stages according to attributes
o I.e., Each stage → eliminate options that do not meet minimum attribute requirement

Example:
Buying a laptop
• Desired attributes: Capable of gaming, costs < RM5k
• Importance of attribute determines when in the sequence it is evaluated
o More important attributes are evaluated in earlier stages
• E.g., A rich person would eliminate based on gaming specs first, then price; A cost-savvy person would
eliminate based on price first, then gaming specs

Buying a car
• Desired attributes: Family-oriented, costs < RM60k
• …
DECISION-MAKING APPROACHES
Descriptive DM – Non-Compensatory Strategies

Satisficing (Simon, 1957)

• Satisfying + Sufficing
• Usually when under time constraints or too many options
• Choosing an “acceptable alternative”; settling for “good enough”
• E.g., University applications (CGPA cut-offs); buying food when late for class
DECISION-MAKING APPROACHES
Descriptive DM – Heuristics
Heuristics
• Mental shortcuts; rules of thumb to support fast DM
• Often works well but may fail in unfamiliar circumstances

Availability Heuristic
• DM is influenced by info that is easily available/recalled
o Estimates likelihood of events to be higher if more ‘available’
• E.g., Pilots choose to fly in bad weather (recalls successful bad weather flights); choosing to travel by car
over plane, etc.

o Which is more likely to kill you? (A) falling airplane parts OR (B) sharks
o Which cause of death claims more lives? (A) diabetes OR (B) homicide
o “ “ (A) stomach cancer OR (B) car accidents

• Errors made when rare occurrences are memorable (e.g., TV reports, movies, viral news)
DECISION-MAKING APPROACHES
Descriptive DM – Heuristics
Recognition Heuristic (Goldstein & Gigerenzer, 2002)
• Recognized objects are judged to have higher value on a criterion
• Examples:
o Brand advertising
o Pilots’ selection of emergency destinations

Anchoring Heuristic (Wickens, 2003)

• Info processed earlier is often most influential; rely on first pieces of info
• First impressions; similar to primacy effect
• People tend to give initial cues more weight and underutilize subsequent cues
• Explains ‘cognitive tunnelling’ (why people get stuck in initial hypotheses)
• Examples:
o 1 x 2 x 3 x 4… x 8 vs. 8 x 7 x 6 x 5… x 1 (median estimates: 512 vs. 2,250)
o “Was Gandhi older or younger than ages 9 or 140 when he died?” (Strack & Mussweiler, 1997)
o Haggling prices, salary, sentences, etc.
o Deciding to fly in deteriorating weather conditions based on earlier favourable report
DECISION-MAKING APPROACHES
Naturalistic DM
Naturalistic Decision-Making
• Explains DM of experienced people in real-world conditions (traditional DM models inadequate)
o I.e., Time pressure, uncertain alternatives, dynamic & evolving events, competing & changing
goals

• Descriptive model for experienced decision-makers

o Firefighters, aviation, military, search & rescue, etc.

• E.g., Firefighters → make series of decisions based on several factors

o Size of fire
o Rate of fire spreading
o Risk to nearby structures, people, etc.
DECISION-MAKING APPROACHES
Naturalistic DM – Recognition Primed
Recognition Primed Decision Model (RPDM)
• Explains decision-making behaviour of experts in time-sensitive environments
• Presupposes that decision-maker → experienced professional
o Repeated exposure to same sets of correlated cues
o Recognize similarities between past encounters and current situation
o Identify appropriate response based on experiences
• Mental Simulation
o Used in more complex situations → evaluate alternate courses of action
o Infer outcomes of different decisions; predicts response of system/situation
o Evaluate ‘moves’ based on ‘countermoves’ (e.g., chess game)
• Fast; automatic; little conscious awareness → reduces cognitive demands
DECISION-MAKING
Expertise & Stress in DM

• Stressors (sudden, novel, intense) → physiological, cognitive, emotional responses

o Heart rate, blood pressure, respiratory changes
o Tension, anxiety, fear

• Stress effects on DM (under time pressure) (Janis, 1982; Janis & Mann, 1977):
o Consider fewer alternatives
o Evaluate less info
o Less systematic evaluation
o Less flexible (persisting with bad strategies)
DECISION-MAKING
Expertise & Stress in DM

• Positive effects of stress

o Motivator to act
o Increase attentional focus
o Reduce distraction from irrelevant stimuli

• Stress can be adaptive (with expertise)

o Simplifying strategies is adaptive given time constraints
o Compensatory methods not viable (time factors, insufficient & ambiguous info)
o Narrow field of attention & info; less demanding cognitive strategies
o Experts have a wider range of experiences → hence, adaptive despite less info being
evaluated
PROBLEMS IN DECISION-MAKING

• Descriptive DM methods (i.e., heuristics) are helpful under certain circumstances, but
may result in systematic errors/biases

• Decision biases include…

o Confirmation bias
o Hindsight bias
o Anchoring and adjustment
o Overconfidence
PROBLEMS IN DECISION-MAKING
Decision Biases
Confirmation Bias

• Favouring info/hypotheses that confirms or aligns with own beliefs

• Seek, recall, & interpret info that supports our beliefs
• Recall info that supports our hypotheses, but ignore other examples that disprove it
o “Never buy a used car from a Chinese man” – (Random Grab Driver, 2019)
• Examples:
o Political orientation; religious beliefs
o Rattling sound in car → interpreted as caused by different problems → influences
how we seek supporting info (i.e., symptoms)
PROBLEMS IN DECISION-MAKING
Decision Biases

Case Study: Aviation (2006), Lexington

• Dark conditions; took off from wrong runway
• Crashed into trees; 49 deaths (crew & passengers)
• Lack of runway lights confirmed belief that ‘lights were faulty’
• Other cues were missed/ignored by pilots; instead ‘talked themselves into believing’ they were
on the correct path

Case Study: Aviation (1989), Kegworth

• Engine 1 failure wrongly diagnosed as engine 2 failure
• Crew looked for supporting evidence
• Full power supplied to engine 1; shut down engine 2
• Engine 1 suffered further failure → full shut down & crash
PROBLEMS IN DECISION-MAKING
Decision Biases
Hindsight Bias
• “Knowing the outcome causes people to believe they could have predicted it, and therefore
would not have suffered the same consequence” – Stone (2017)
• ‘Knew-it-all-along’ phenomena; hindsight 20/20
o “I knew this would happen”; “they should have done this instead”
• Causes inappropriate blame to operators for seemingly obvious errors; problems then left
unaddressed
o Recall: why poor designs exist – designers don’t see them as problems; users at fault
• Application example: Medical 2nd opinions
o When seeking 2nd opinions, should doctors reveal their own diagnosis?
o Better to withhold own evaluations; seek independent opinions/feedback
PROBLEMS IN DECISION-MAKING
Decision Biases
Anchoring and Adjustment

• People tend to anchor on initial info and adjust from there to reach a plausible estimate

• As a result, decisions made are often close to anchor point

• Errors result when anchors are inaccurate/faulty to begin with

• Recall: Anchoring heuristic

PROBLEMS IN DECISION-MAKING
Decision Biases
Overconfidence

• Discrepancy between perceived and actual ability and skill

o How good are your driving skills? (rate 1-10)

• Overconfidence leads to poor decisions

Examples:
Doctors’ diagnostic skills (Meyer et al., 2013)
• Confidence rating for diagnosing easy vs. difficult cases – 7.2 vs. 6.4 (on 10-point scale)
• Doctors’ accuracy in diagnosing easy vs. difficult cases – 55.3% vs. 5.8% accuracy
• Higher confidence → decreased requests for additional diagnostic tests
PROBLEMS IN DECISION-MAKING
Decision Biases
Overconfidence

More examples:
• Approx. 90% of drivers think they are safer than the median driver (Svenson, 1981)
• Confidence on prediction accuracy of future events (e.g., elections, sports matches) is 90%;
actual accuracy is 60%
• Planning fallacy → overly optimistic in projection of required time/resources
o 84% students confident in handing in assignment on time; only 40% did (Buehler et al., 2002)
• Pilots with higher confidence in own situational awareness estimates were actually less accurate
(Sulistyawati et al., 2011)
PROBLEMS IN DECISION-MAKING
Decision Biases
Overconfidence

• Case Study: NASA Challenger space shuttle disaster

o NASA estimated the risk of accident to be 1/100,000 launches
o Later revised to be 1/100 – 1,000-fold increase in risk

• Overconfidence can be reduced through training and feedback

o Prompt feedback on accuracy of judgements
OVERTHINKING

• Descriptive DM (i.e., heuristics) → can lead to error

• But in time constrained situations, descriptive DM > than normative DM

• Normative DM → works best when there is little info to evaluate

• Descriptive DM → works best when there is more info to evaluate

• Normative DM results in too much thinking → impairs DM (poorer performance than

descriptive DM) because…
o Conscious processes have low capacity; can only evaluate a subset of relevant info
o Conscious thought → inappropriate weighting of cues
EXPERTS VS. NOVICES
Characteristics of Experts

Expert decision-makers

• Identify better, more efficient, and more alternative solutions more quickly

• More efficient coding of info in LTM (less affected by STM limitations)

o BUT memory skills are domain specific
o I.e., Chess players’ memory may be mediocre outside of their expertise

• Better at handling interruptions and multi-tasking

EXPERTS VS. NOVICES
Characteristics of Experts

Expert decision-makers

• Better at selecting only task-relevant info; less cognitive burden → better management
of resource demands

• Better understanding of utility of different cues

• Examples:
o Medical field (experts are better at identifying and recalling important info; students
recall more info in general) – Boshuizen & Schmidt (1982); Chase & Ericsson (1982)
o Aviation (experts more efficient at acquiring & evaluating weather info than novices)
– Rockwell & McCoy (1988)
EXPERTS VS. NOVICES
Characteristics of Experts

Expert decision-makers

• Metacognition
o Monitoring, controlling, & organizing own cognitive processes
o Monitor and evaluate goal progress & performance/accuracy
o Consider effective usage of time and mental effort in decision-making
o Rely on prior experiences to identify patterns & similarities to current situation
– Cohen et al. (2000); Halpern (1998)
o Metacognition → refines understanding and knowledge
EXPERTS VS. NOVICES
Novices

Novices – Unskilled and unaware of it

• Dunning-Kruger Effect (Kruger & Dunning, 1999)

o Cognitive bias; overestimation of ability
o Participants rated own reasoning ability
→ 68th percentile; actually 12th percentile
o Top quarter percentile participants
underestimated ability & performance

• Novices are less competent AND lack

awareness of their incompetence
EXPERTS VS. NOVICES
Limitations of Experts

• Experts have superior skills and knowledge to novices; but advantages are modest

• Experts usually (but not always) outperformed novices, and often outperformed by simple
statistical models
o Due to inconsistency & errors in weighting and combining info (Camerer & Johnson, 1991)

• Experts typically rely on less info & employ simplistic models

• ‘Too much’ info increases complexity of integrating info (i.e., difficult to accurately weight cues)
→ poorer DM

• Expert DM depends on nature and condition of tasks; performance is better when:

o Stimuli & cues are static
o Receive feedback on accuracy of judgements
EXPERTS VS. NOVICES
Limitations of Experts
Good Expert Performance Poor Expert Performance
Decisions about things Decisions regarding behaviour
Experts agree on stimuli Experts disagree on stimuli
Predictable problems Unpredictable problems
Task Properties
Tasks tend to be repetitive Tasks are unique
Feedback available Feedback unavailable
Objective analysis available Subjective analysis only
Weather forecasters Clinical psychologists
Accountants Polygraph administrators
Astronomers Personnel selectors
Industry/Field
Test pilots Student admissions personnel
Physicists Court judges
Chess masters Behavioural researchers
Source: Shanteau, J. (1992). Expertise and Decision Support. Plenum Press: New York
IMPROVING DECISION-MAKING
Methods
Training
• Extended practice in restricted domains
• Experience; repeated exposure to situations with correlated cues

Decision Aids
• “Tools that help deliberate upon > 2 options” (Bekker et al., 2003)
o Protocols
o List of options
o Flowcharts; decision trees
o Patient decision aids (involved decision-making; treatment selection)
o Automated (computer-based) decision aids (e.g., AI diagnostics; flight plan generation)
• Used by/in medical professionals, scientists, diagnostics, consumer purchasing, etc.
• Advantage → Not affected by biases & cognitive limitations
IMPROVING DECISION-MAKING
Bias in Use of Decision Aids
Bias in Use of Decision Aids

• Eliminates/reduces bias; but may lead to others

o Automated decision aids; computer-generated solutions

• Automation Bias
o Operators trained to trust system; overreliance → complacency
o Develop tendencies to disregard/fail to seek info contradictory to decision aid
o More likely to choose options provided by aids even when incorrect
IMPROVING DECISION-MAKING
Bias in Use of Decision Aids
Bias in Use of Decision Aids

Case Study: Operation Iraqi Freedom

• Missile system wrongly identified American F/A-18 & British Tornado aircrafts as foes
• Operators given 10s to veto identification
• Trusted software that was known perform poorly
• Aircrafts downed by missile; 3 crew members killed
RESOURCES

• Introduction to Human Factors: Applying Psychology to Design (Stone et al., 2017)