- Количество слайдов: 21
Decision Making Human Factors Psychology Dr. Steve
Historically Bad Decisions • • • Challenger explodes when launched in cold weather U. S. S. Vincennes shoots down Iranian civilian jetliner U. S. military is ambushed at Bay of Pigs Red Sox owner sells Babe Ruth to the Yankees Sony sticks with Betamax format for video recording U. S. supports Mujahadeen in war with Russia
Definitions n Decision making – selecting one choice from a number of choices involving some level of uncertainty. n Intuitive Decision Making – quick and relatively automatic responses to a problem. n n Ex: Response to yellow traffic light Analytic Decision Making – slow, deliberate, and controlled responses to a problem. n Ex: What stock to purchase
Rational Decision Making Research n Normative Decision Models – Assumes individuals act rationally in trying to find the best solution to optimize outcome. n n Utility – overall value or worth of a choice Expected Value – is the overall value of the choice determined by multiplying the utility of the choice times the probability of the outcome n Ex: The decision to buy a lottery ticket may be determined by how much the jackpot is worth times the probability of winning
Subjective Expected Utility Theory Alternative/Outcomes Probability Utility (0 to 1) Px. U (-10 to +10) Use Safety Device Alternative Expected Utility -1. 7 Accident: Injury or death is prevented . 10 +10. 0 +1. 0 No Accident: inefficient, restrictive, discomfort . 90 -3. 0 -2. 7 Not Use Safety Device +5. 3 Accident: Serious injury or death . 10 -10. 0 -1. 0 No Accident: more efficient and comfortable . 90 +7. 0 +6. 3 Model predicts workers will not use the safety device What if probability of an accident was. 50?
Descriptive Decision Models n Descriptive Decision Models – Assume humans do not act rationally in decision making (assumptions of normative models are violated) n Framing Effects – The way a problem is phrased affects the decision n n Ex: Choice to have surgery affected by whether told you have a 60% chance of living, or told you have a 40% chance of dying. Satisficing – making a decision that is just good enough without taking extra time and effort to do better n Ex: Decision as to how well I should define a term
Algorithms vs. Heuristics Solve the anagram: metssy n Algorithms are procedures that will always lead to correct answer n n Ex: Try every combination – metssy mtssye mssyet msyets …. Heuristics are shortcuts that are not guaranteed to lead to best answer, but are more efficient n Ex: use rules of thumb such as need a vowel to separate consonant sounds –mestys stymes system
Information Processing Model applied to Decision Making Cue Reception/Integration -cues from environment are placed in working memory (cues possess uncertainty) Hypothesis Generation - guesses about cues are made drawing form LTM Hypothesis Evaluation/Selection - collect additional cues to test the hypothesis Generating/Selecting Actions - alternative actions are generated by retrieving possibilities from LTM
Factors influencing Decision Making n Amount of cue info brought into Working Memory n n Time available for decision-making n n Function of how many tasks are occurring concurrently LTM retrieval ability n n Function of the time criticality of task Attentional resources n n Function of attentional demands Person may possess right info, but fail to retrieve it (inert knowledge) Working Memory Capacity n Can only hold so much info in WM at one time How do these factors affect your use of automated phone menu systems?
Criteria for “Good” Decisions n Outcome produces maximum value n Problem is that decision are often made to avoid worst outcome rather than maximize value n n Positive vs. Negative outcome n Problem is decision may be positive in the short term, but turn out to be a big mistake later n n Ex: Decision to buy the extended warranty on an appliance Ex: Japan’s decision to attack U. S. in 1941 Comparison to expert’s decisions n Problem is that experts don’t always make good decisions n Ex: Experts’ decision to launch Challenger
Heuristics Biases in using cues for DM n n n Attention to a limited number of cues – affected by the magical number 7 Cue Primacy – the first few cues are given greater importance Anchoring Heuristic – once an initial decision is made, later cues are often ignored Cue salience – cues that are easily noticed are most likely to be used Overweighting of unreliable cues – reliability of cues is often overlooked
Heuristics Biases in hypothesis generation Limited number of hypotheses generated – people consider only a small subset of possibilities n Availability heuristic – people make judgments based on how easily information is retrieved (e. g. , risk of airplane crash) n Representativeness heuristic – decision based on how closely info represents typical outcome n Overconfidence – individual’s belief that they are correct more often than they actually are n
Heuristics Hypothesis evaluation/selection n Cognitive fixation – identifying a hypothesis and sticking with it (mind set) n n (application to “Business in Bhopal”) Confirmation Bias (cognitive tunnel vision) – tendency to seek out only confirming information n Ex: car won’t start and battery dead, fail to check alternator 5 Dr. Steve Kass University of West Florida Pensacola, FL 32514 4 Dr. Steve Kass University of West Florida Pensacola, FL 32514 Hypothesis: If an envelope is sealed, then it has a 5 cent stamp on it. Turn over the minimum number of envelopes necessary to test this hypothesis
Heuristics Biases in action selection Retrieve a small number of actions – limited by how many action plans can be held in working memory n Availability heuristic for actions – easy to retrieve actions are most often chosen n Availability of possible outcomes – decisions will be made based on how memorable the outcome of that choice has been in the past n
Naturalistic Decision Making Naturalist Decision Making – research into the way people use their experience to make decisions in field settings Real-world decision making tasks typically include: n n n n Ill-structured problems Uncertain, dynamic environments Information-rich environments where situational cues change rapidly Cognitive processing that proceeds in iterative action/feedback loops Multiple shifting and/or competing individual and organizational goals Time constraints or stress High risk Multiple persons involved in decision
Rasmussen’s Skill-, Rule-, and Knowledge-based performance model High Novice Low Performance Attentional Demand Analytic Intuitive Expert Automatic
Situation Awareness – “skilled behavior that encompasses the processes by which task-relevant information is extracted, integrated, assessed, and acted upon” (Kass, Herschler, & Companion, 1991). Levels of SA (Endsley, 1988) • Level 1 – Awareness of information • Level 2 – Comprehension of its meaning • Level 3 – Projection of future status SA is Difficult to measure: Self-report measures - Only aware of what you are aware of Performance-based measures – Intrusive, measure affects performance
Factors Affecting Loss of Situation Awareness • Attention • attentional demands of controlled processes (k-based performance) • Pattern Recognition • inability to perceive pattern of cues (recognition-primed DM) • Workload • tasks too demanding or too many at once • Mental models • inadequate understanding of system or state • Working Memory • failure to adequately “chunk” information Examples: • Commercial plane crashes in the Everglades when aircrew becomes fixated on a warning light while the plane slowly descends into the ground. • Outfielder for the Mets tosses ball to a fan after making the second out while runner on base easily scores.
Improving Situation Awareness n n n Cue Filtering – eliminate irrelevant cues (clutter) that interfere with accurate assessment of situation Augmented Displays –displays that highlight or overlay actual information to make it more salient Spatial Organization – arranging displays to capitalize on spatial relationships (e. g. , pop-out effect) Automate Status Updates – as the environment changes the system should warn the user of change Train Users to Improve Attention?
Methods for Improving Decision Making • Redesign – System should make the decision options obvious • Problem: No system can rule out all bad options • Training – Train people to be better decision makers • Focus on process measures – not outcome measures • Reward correct each DM step, outcome maybe delayed • Provide feedback on consequences of bad, as well as good decisions • Problem: Decision making is typically task-specific • Decision Support Systems – Makes available expert knowledge for DM • Problem: People tend to mistrust DSS, or can’t use for novel problems
Decision Aids Click on dice for decision aid based on Expected Value Theory Decision Tables/Trees – provides calculations of possible outcomes that would result from different choices n Expert Systems – computer programs that use experts’ knowledge of concepts, principles, and rules n Decision Support Systems – any interactive system that allows you to input problem information which it uses to formulate a solution based on complex algorithms n