2Outline Definitions Decisions and alternatives Characterizing decisionsDecision making strategiesDecision making phasesImplications for decision support
3Definitions Choice about a “course of action” -- Simon Choice leading to “a certain desired objective”-- ChurchmanKnowledge indicating the nature of a commitment to action-- Holsapple and Whinston
4Simon’s Model of Problem Solving Decision-making consists of three major phases---intelligence, design, and choice [Simon]H.A. Simon The New Science of Management Decision. Harper and Row, NY.Newell, A., & Simon, H.A. (1972). Human Problem Solving. Prentice-Hall, Englewood Cliffs, NJ.
5ExampleA farmer with his wolf, goat, and cabbage come to the edge of a river they wish to cross. There is a boat at the river’s edge, but of course, only the farmer can row. The boat can only handle one animal/item in addition to the farmer. If the wolf is ever left alone with the goat, the wolf will eat the goat. If the goat is left alone with the cabbage, the goat will eat the cabbage. What should the farmer do to get across the river with all his possessions?
6Phase I: Intelligence Problem Identification and Definition What's the problem?Why is it a problem?Whose problem is it?
7Phase II: Design Problem Structuring Generate alternatives Set criteria and objectivesDevelop models and scenarios to evaluate alternativesSolve models to evaluate alternatives
8Problem Solving State Space Search Initial State Goal State Operators Choosing representation and controlling the application of operators requires decision making
10States and Operators State = <Farmer/Boat location, Wolf location, Goat location, Cabbage location>Operator<L,L,L,L> ----> <R,R,L,R>…..
11Phase III: Choice Solution Determine the outcome of chosen alternativesSelect the/an outcome consistent with the decision strategy
12Decisions and Alternatives where do they come from?how many are enough?Evaluationhow should each alternative be evaluated?how reliable is our expectation about the impact of an alternative?ChoiceWhat strategy will be used to arrive at a choice?E.g., DxPlain
14Human Cognitive Limitations (Harrison, 1995) Retain only limited information in short-term memoryDisplay different types and degrees of intelligenceThose who embrace closed belief systems restrict information searchPropensity for risk variesLevel of aspiration positively correlated to desire for information
15Common Perceptual Blocks (Clemen, 1991) Difficulty in isolating a problemDelimiting the problem space too closelyInability to see the problem from various perspectivesStereotypingCognitive saturation or overload
17Decision Making Strategies OptimizingSatisficingQuasi-satisficingSole decision ruleSelection by eliminationIncrementalism and muddling throughConsider some well-known decision problems- Health care reform- Faculty hiring process
18Decision Making Strategies ConsiderationsIndividual-focused vs. organization-focused decisionsIndividual vs. group decisionsExpensive-to-change vs. inexpensive-to-change decisions
19Optimizing Goal: select the course of action with the highest payoff estimation of costs and benefits of every viable course of actionsimultaneous or joint comparison of costs and benefits of all alternativeshigh information processing load on humanspeople do not have the ``wits to maximize'' [Simon]
20Observations Given high cost in time, effort, and money Decisions are made under severe time pressure (``fighting fires'')Optimization on stated objectives may result in sub-optimization on unstated, less tangible objectivesTherefore, people oftenDo not consider all alternativesDo not evaluate all alternatives thoroughly and rigorouslyDo not consider all objectives and criteriaPlace more weight on intangible objectives and criteria
21SatisficingDecision-makers satisfice rather than maximize [Simon]. They choose courses of action that are ``good enough''---that meet a certain minimal set of requirementsTheory of bounded rationality: human beings have limited information processing capabilitiesOptimization may not be practical, particularly in a multi-objective problem, yet knowing the optimal solution for each objective and under various scenarios can provide insight to make a good satisficing choice
22Sole Decision Rule``Tell a qualified expert about your problem and do whatever he (she) says---that will be good enough'' [Janis and Mann]Rely upon a single formula as the sole decision ruleUse only one criterion for a suitable choicee.g., do nothing that may be good for the enemyImpulsive decision-making usually falls under this category
23Selection by Elimination Eliminate alternatives that do not meet the most important criterion (screening; elimination by aspects)Repeat process for the next important criterion, and so onDecision-making becomes a sequential narrowing down process
24Selection by Elimination ``Better'' alternatives might be eliminated early on---improper weights assigned to criteriaDecision-maker might run out of alternativesFor complex problems, this process might still leave decision maker with large number of alternatives
25IncrementalismOften, decision-makers have no real awareness of arriving at a new policy or decisiondecision-making is an ongoing processthe satisficing criteria themselves might change over timeMake incremental improvements over current situation and aim to reach an optimal situation over timeUseful for ``fire-fighting'' situationsFrequently found in pluralistic societies and organizations
26Heuristics and BiasesHeuristics are “rules of thumb” that can make a search process more efficient.Most common biases in the use of heuristicsAvailabilityAdjustment and anchoringRepresentativenessMotivationalA. Tversky and D. Kahneman “Judgement Under Uncertainty: Heuristics and Biases.” Science, 185:
27Example 1 Which is riskier (probability of serious accident): a. Driving a car on a 400 mile trip?b. Flying on a 400 mile commercial airline flight?
28Example 2 Are there more words in the English language a. that start with the letter r ?b. for which r is the third letter?
29Availability “what is easily recalled must be more likely” Inability to accurately assess the probability of a particular event happeningAssess based on past experience which may not be representativeStructured review and analysis of objective data can reduce availability bias
30Example 1A newly hired programmer for a software firm in Pittsburgh has two years experience and good qualifications. When an employee at Au Bon Pain was asked to estimate the starting salary she guessed $40,000. What is your estimate?a. $30,000 - $50,000?b. $50,000 - $70,000?c. $70,000 - $90,000?
31Example 2A newly hired programmer for a software firm in Pittsburgh has two years experience and good qualifications. When an employee at Au Bon Pain was asked to estimate the starting salary she guessed $80,000. What is your estimate?a. $30,000 - $50,000?b. $50,000 - $70,000?c. $70,000 - $90,000?
32Adjustment and Anchoring Make estimates by choosing an initial value and then adjusting this starting point up or down until a final estimate is obtainedMost subjectively derived probability distributions are too narrow and fail to estimate the true variance of the eventAssess a set of values, instead of just the mean
33ExampleWhat is the most likely sequence of gender for series of children born within a family?- The sequence of BBGGBG, BGBBBG, BBBBGG?
34ExampleMike is finishing his CMU MMM degree. He is very interested in the arts and at one time considered a career as a musician. Is Mark more likely to take a job:a. In the management of the arts?b. A medical management position?
35RepresentativenessAttempt to ascertain the probability that a person or object belongs to a particular group or class by the degree to which characteristics of that person or object conform to a stereotypical perception of members of that group or class. The closer the similarity between the two, the higher is the estimated probability of associationSmall sample size biasFailure to recognize regression to the mean (predicted outcomes representative of the input?)
36MotivationalIncentives, real or perceived, often lead to probability estimates that do not accurately reflect his or her true beliefsNon-cognitive, motivational biasesDifficult to address through the design of a DSSSolicit a number of estimates from similar sources, both related and unrelated to problem context
37Summary: Heuristics and Biases Heuristics are rules of thumb that we use to simplify decision making.Overall, heuristics result in good decisions. On average any loss in quality of decision is outweighed by the time saved.But, heuristics can cause biases and systematic errors in decision making when they fail.In addition, we are typically unaware of the heuristics and biases, and fail to distinguish between situations in which their use is more and less appropriate.
38Evaluation Metrics Effectiveness: what should be done Easier access to relevant informationFaster, more efficient problem recognition and identificationEasier access to computing tools and modelsGreater ability to generate and evaluate large set of alternativesEfficiency: how should it be doneReduction in decision costsReduction in decision time for same level of detail in the analysisBetter quality feedback
39Implications for Decision Support Different people will use different strategies at different times for different kinds of decisionsWhich decision strategy to engineer in a decision support system?Multiple strategies may be used in making a decision
40Implications for Decision Support Is there an ``optimal decision strategy’’ for each problem?What are the information processing requirements for each decision-making strategy?Which strategy do decision-makers favor, When, and Why?
41Value of DSS easier access to information Increase the bounds of rationalityeasier access to informationidentify relevant informationincrease ability to process information