principles of information systems session 08 decision and choice
TRANSCRIPT
Principles of Information SystemsSession 08
Decision and Choice
2
Problem Identification and Solving
Chapter 7
It’s much easier to just plain decide. Never mind – nothing is going to change your mind. I did that once when I was a student at MIT. I got sick and tired of having to decide what kind of dessert I was going to have at the restaurant, so I decided it would always be chocolate ice cream, and never worried about it again – I had the solution to that problem. Richard Feynman, ‘Surely You’re Joking, Mr Feynman’
4
Overview
Learning objectives
1. Introduction
2. The nature of decision – individuals
3. The nature of decision – groups
4. Uncertainty and risk
5. Decision-making theories and approaches
6. Methods of supporting decision-making
7. Summary
5
Learning objectives• Explain why individuals and groups need to make
decisions and identify the main themes involved in decision-making
• Describe the cognitive aspects of decisions and recognise the cognitive biases that can affect individual thinking and decision-making
• Identify the issues involved with decision-making in groups and describe some models and techniques for group decision-making
• Explain how uncertainty and risk affect decision-making
• Distinguish between programmable and non-programmable decisions
6
Learning objectives
• Describe several decision-making theories and approaches, including classical decision theory, behavioural approaches and multi-criteria decision-making
• Use different techniques to represent the logic of decisions, including matrices and rules
• Describe some problems and approaches for strategic decision-making
• Describe what needs to be done to automate the process of decision-making and describe some methods for decision support
What did you decide today?
List ALL of the decisions you made since waking up this morning!
8
• Whether to get out of bed…• What to have for breakfast…• Whether to come to this lecture…
• Whether to throw out or recycle the drink cans…• Which sound system to buy…• Who to vote for in the guild election…
• What career you are aiming for…• Where you want to live next year…
9
What affected your decision-making?
• Uncertainty• Risk• Intuition• Emotion• Calculation• Morals• Ethics• Competing values• Other people• …
10
Introduction
• Decisions are a part of everyday life-Decisions and judgements are made constantly, even subconsciously
-Decision-making applies to the most banal and the most important situations
• Decision-making involves choosing between alternatives
• Decisions can involve: -Risk and uncertainty-Multiple possible outcomes-Competing values-Other stakeholders
1. Introduction2. The nature of decision - individuals3. The nature of decision - groups4. Uncertainty and risk5. Decision-making theories and
approaches 6. Methods of supporting decision-
making7. Summary
11
The nature of decision - individuals
• How should a person make a decision? • This is one of the most fundamental life questions of all,
affecting daily personal and professional activity and existence.
1. Introduction2. The nature of decision -
individuals3. The nature of decision - groups4. Uncertainty and risk5. Decision-making theories and
approaches 6. Methods of supporting decision-
making7. Summary
12
Rational decision – a decision that is based on the available information, weighing
advantages and disadvantages in a logical manner.
Intuitive decision – a decision that is based on feelings or values rather than
being made rationally.
Values – individual moral and social beliefs and principles.
The large size coffee tin is better value, but if I buy that I won’t have
enough money for cheese
I can’t really afford those
shoes but they make me feel
great!
I won’t buy genetically
modified food products
13
Rational versus intuitive
• In practice, most decisions are likely to be a mixture of both emotion and reason, or at least validated against the other mode.
• Analysing your values first can make a subsequent choice among alternatives more meaningful.
-For example, you may be trying to decide among a range of countries and destinations for your holiday, but deeper analysis of your values suggests that you would rather stay and relax at home
14
Cognition and informatics
• Cognitive science is the study of how humans think and is highly relevant to informatics.
• Cognition refers to the mental processes involved in processing information, such as:
-memory
-problem solving
- intelligent reasoning
-pattern recognition.
• These are all functions that can be replicated, supported or extended using computing technologies.
15
Cognitive aspects of decisions
• In decision-making cognitive ability relates to the capacity that an individual has effectively to process information.
-e.g. Good chess players can visualise several moves ahead and construct scenarios based on the opponent’s best replies.
• The ability to memorise and recognise familiar patterns, recognise the relevance of new information and to make connections all help increase decision quality.
16
Multiple intelligences in cognitive science
• Emotional intelligence is now widely accepted as a distinctive quality affecting the management of one’s own emotions and interpersonal relationships.
- ‘bedside manner’, creative ability, political nous, empathy, etc
• The decisions made in any situation may be more or less intelligent and the quality of the intelligence may involve social and emotional competencies as well as analytic ones.
Bias – a preference that is not supported by logic.
Biases in thinking can reduce the quality of a decision.
Heuristic – a shortcut or rule of thumb that helps you guess an
answer if you don’t know.
Guessing strategies
18
Cognitive biases
• Many types of cognitive biases have been identified that directly affect decision-making
-Availability
-Gambler’s fallacy
-…. (many more!)
• The classic work on this was done by Amos Tversky and Daniel Kahneman
- ‘The framing of decisions and the psychology of choice’ Science, 1981. 211: 453-8)
19
The Availability heuristic
• Which is a more likely cause of death in the United States - being killed by falling aeroplane parts or by a shark?
?
20
The Availability heuristic - Sharks and jets
• In the USA, the chances of dying from falling aeroplane parts is 30 times greater than dying from a shark attack.
-Most people rate shark attacks as the more likely cause of death.
-Shark attacks are easier to imagine and receive more publicity. Since information about shark attacks is more readily available, the availability heuristic helps explain why people overestimate the chances of dying in this way.
21
The Availability heuristic
• Which member of each pair do more Americans die from:
a) diabetes or homicide
b) car accidents or stomach cancer
c) tornado or lightning
?
22
The Availability heuristic - Diabetes and homicide• More Americans die from diabetes and stomach cancer
than from homicide and car accidents, by about 2:1. - Many people guess homicide / car accidents, largely due to the publicity those receive and thus their availability
23
The Availability heuristic
• Consider the following question.
A word is chosen at random from the dictionary (discarding words with less than 3 letters). Is the word more likely to begin with ‘k,' more likely to have ‘k' as its third letter, or are both outcomes equally likely?
A. A word beginning with ‘k' is more likely to be chosen than one that has ‘k' as its third letter.
B. A word with ‘k' as its third letter is more likely to be chosen than one that begins with ‘k'.
C. It is equally likely that a word that begins with ‘k' will be chosen as it is that a word that has ‘k' as its third letter will be chosen.
?
24
The Availability heuristic –Words with K• About 67% thought A but answer is B, by 2:1
- It is easier to think of words beginning with k than with k in third position
25
The Availability heuristic
• Is where decision makers
“assess the frequency of a class or the probability of an event by the ease with which instances or occurrences can be brought to mind” (Amos Tversky and Daniel Kahneman, 1974)
Availability heuristic – a type of cognitive bias concerning the tendency of people to opt for the answer that involves the least effort in thought.
Representativeness heuristic
“Linda is 31 years old, single,
outspoken and bright. She
majored in philosophy. As a
student she was deeply concerned
with issues of discrimination and
social justice, and also participated
in antinuclear demonstrations.”
Which is most likely?
A. Linda is a bank teller
B. Linda is a bank teller and is active in feminist movement
?
Representativeness heuristic
• “A bank teller & feminist” can’t be more likely than just “a bank teller”
• conjunction fallacy
Bank tellers Feminists
Feminist bank tellers
28
Representativeness heuristic
• Nearly 90% of 86 people said B – why?
• Even when T+K checked that “A” wasn’t interpreted as “Linda is a bank teller and is not active in feminist movement”
• Repeated for other scenarios (accountants, jazz / lose first set and win match etc)
29
Representative bias – war example
What is more likely in next few years?
A. An all out war between the USA and China?
or
B. A war between them triggered by a third country such as Iraq, Libya, Israel or Pakistan?
?
30
War example
• Most people feel the more specific event (B) is more likely than the more general one (A).
“As amount of detail in scenario increases, its probability decreases but its REPRESENTATIVENESS and hence its apparent likelihood may increase”
(Tversky & Kahneman, 1982:98)
• more details…. less probable but more PLAUSIBLE
Representativeness heuristic – a type of cognitive bias where the statistical probability
of something is estimated erroneously, because it is based on a preconceived idea of its representativeness of a particular category.
Representativeness heuristic
32
(1) Is the % of African countries in the UN more or less than 65%?
• What would you guess exactly?
• Now forget that – the wheel spins again!
Anchoring and adjustment
?
33
Anchoring and adjustment
(2) Is the % of African countries in the UN more or less than 10%?
• What would you guess exactly?
• T+K found averages of 45% in (1), 25% in (2)• First bias – insufficient adjustment
from anchor
?
34
Anchoring and adjustment
• This is the tendency when asked for an estimate, to go for a ballpark figure then make adjustments as more information becomes available
• Most people do not revise their initial judgement far enough, leading to a biased estimage
35
Gambler’s fallacy
• A fair coin is flipped three times – HHH• You have to bet $100 on next outcome.• H or T – why?
?
36
Gambler’s fallacy
• No preference is correct• T – thought of as more likely due to run• Assumption that chance sequences must be locally
representative – wrong!
• But what about hot streaks in sport? …
37
Gambler’s fallacy: the hot hand
Which of the following sequences of X's and O's seems more like it was generated by a random process (e.g. flipping a coin) ?
1. XOXXXOOOOXOXXOOOXXXOX
2. XOXOXOOOXXOXOXOOXXXOX
?
38
Gambler’s fallacy: hot hand
1 alternates on half of all occasions
2 on 70%
1 is expected by chance, 2 isn’t
Did you think runs in 1 were too long to be random? This is the hot hand fallacy.
39
The Affect heuristic
• Psychological factors can also act as a bias in decision-making
• Affect means ‘feeling’
• The affect heuristic suggests that having a positive or negative feeling at the time can shape the perception about the risk or benefit of something
- If you are feeling happy, you are more likely to judge something as ‘good’
40
Cognitive biases affecting decisions
• These biases (and many others) introduce unpredictability and irrationality into human decision-making
• This can have significant implications, e.g. decisions on funding:
- More shark nets or more research on diabetes?
41
Post-hoc rationalisation
• A rational explanation for a decision is made after the event in order to justify the decision, although it was not in fact considered in the decision-making
If I buy another ornamental duck
one day I’ll be able to open a museum and
retire
42
Individual decision-making is a complex
process, with rational, emotional and
evaluative components,
and with perceptions, biases, amount of
information and other aspects affecting
choices.
Recap
43
The nature of decision- groups
• Individuals are not always in control of the fate of their decisions. An individual may make decisions, but these may be outweighed by the system in which those decisions are actioned
-For example, you may vote for a minority party, but a major party gets into government
• In group decision-making, processes may have to be built around judgmental decision-making to ensure fairness
1. Introduction2. The nature of decision - individuals3. The nature of decision - groups4. Uncertainty and risk5. Decision-making theories and
approaches 6. Methods of supporting decision-
making7. Summary
44
Group decision-making
• Group decision-making is a critical question for families, organisations, communities and nations
• In informatics:
-when establishing the requirements for a new information system
-when combining responses and comments on a community proposal
-when weighing varying evidence from a range of data sources.
• Political science, organisation theory and social psychology have all addressed this issue and many models exist
45
Styles of group decision-making
ConsensusEveryone agrees, all the relevant information is fully available to all, everyone makes a free choice and all make the same choice.
Autocratic
Decisions are made by people or a process trusted to decide for the group.
One person unilaterally makes all decisions affecting the group.
Representative
step
46
Combining group decisions
• When we need a single outcome of a group process some individual decisions will not have an effect on the outcome – the voter paradox
• The way votes are aggregated also affects the outcome
- Can be a political rather than statistical process, eg collegiate vote vs popular vote, in US electoral system
47
Techniques for debating decisions
• Nominal group technique-A group of people is chosen to address a problem and generate ideas and alternatives
-Individuals list ideas and vote on them-Aim is to avoid dynamics of interacting groups
• Delphi-A panel of experts each make an anonymous, individual assessment on an issue
-The assessments are collected and averaged-Experts again give an opinion, until they converge on an agreed figure
48
Groupthink
• When harmony within a group is high, there is the possibility of poor decisions being made through ‘groupthink’
• Decisions are made on the basis of preserving harmony and solidarity within the group, rather than critically considering alternatives
49
Group decision making varies from consensual to
autocratic.
In between is representative decision-making,
where individuals have their say but some
collective process determines the decision for the
group.
An example is voting.
Recap
50
Uncertainty and risk
• In a closed world situation the complete set of choices and their relative probability is known
-Risk can be therefore be estimated statistically, and decisions based on this
-Many card games are like this
• Not all situations reflect a closed world though. More commonly there would be:
-many more choices
-not all outcomes may be known
-the probability of any given one occurring might be hard or impossible to estimate
• In these situations we have uncertainty
1. Introduction2. The nature of decision - individuals3. The nature of decision - groups4. Uncertainty and risk5. Decision-making theories and
approaches 6. Methods of supporting decision-
making7. Summary
51
_ye
Uncertainty and risk
• Risk can be reduced by taking extra information into account
th• We know that ‘the’ is a much more
common word than ‘thy’ - guessing that the next letter is E is a lower risk strategy
step
52
Risk management
• When extra information is not available, strategies of risk management can help reduce uncertainty
-Assess which risks apply and their relative likelihoods and severity, even just ranking by gut feel rather than anything numerical.
-Then assess the possible outcomes so if they occur, they can be recovered or worked around by some contingency plan.
• So, although taking a risk may be unavoidable, in considering taking a risky decision some analysis can help reduce uncertainty, or at least help to handle it.
53
Who Am I?
5-point clue - I was born in Texas in 1905 and studied at the California Institute of Technology. I died in 1976. Who am I?
4-point clue - A proposed biography of my life was announced in 1971, but was never published as the author was imprisoned for fraud. Who am I?
3-point clue - I inherited my father’s tool company and later began producing films including Hells Angels and The Outlaw. Who am I?
2-point clue - My famous plane the Spruce Goose made just one flight. Who am I?
1-point clue - I was portrayed by Leonardo DiCaprio in The Aviator. Who am I?
(example from http://www.quizzyheights.co.uk/who_am_i.htm)
step
54
Who Am I?
• What is the risk and reward involved in playing a competitive game of “Who Am I” such as this?
• Does all the information have the same value?
55
Who Am I?
• Answer: I am Howard Hughes
• If you guess early you are more likely to be wrong, but greater reward if right
• Usually information has more variability in higher values questions –
- lots of people were born in 1905, but Leonardo de Caprio only played one role in the film.
56
Many personal and professional
decisions also involve some risk, and
uncertainty is often associated with risk.
In many cases where decisions are
uncertain, extra information can reduce
the uncertainty.
Recap
57
Risk perception
• Taking risks is a personal preference and rests on an individual’s emotional comfort with a choice and their relative optimism, pessimism or realism.
• Risks may be perceived differently by different people (and even by the same person under different conditions)
• This means that individuals will vary in their assessment of the likelihood of outcomes and behave accordingly.
58
Risk averse and risk-taking behaviour
• Buying a lottery ticket carries the same odds whoever buys it
I’ll save my $4 because I’m not likely to
win
It’s only $4 and I might win the
jackpot this week
59
You are in charge of a hospital, preparing to combat a disease expected to kill 600 people
If program A is adopted: 200 people will be saved
If program B is adopted:There is a 33% probability that 600 people will be saved and a 67% probability that no people will be saved
WHICH DO YOU CHOOSE?
60
You are in charge of a hospital, preparing to combat a disease expected to kill 600 people
If program C is adopted:
400 people will die
If program D is adopted:There is a 33% probability that nobody will die and a 67% probability that 600 people will die
WHICH DO YOU CHOOSE?
61
Framing and Prospect Theory
• Tversky and Kahneman gave this dilemma to real doctors and found a clear difference, depending on which version they read.
• In the first version, most (72 per cent) chose program A.
-This is a conservative, low risk choice.
• Presented with the decision problem as framed in the second version though, 78 per cent chose D, the higher risk option!
62
Framing and Prospect Theory
• Prospect theory addresses the finding that people have different attitudes to the same risk whether it is framed in terms of loss or gain.
- This is important, as it means that decision-makers may behave inconstantly or irrationally
63
Framing techniques
• Fairhurst and Sarr’s classic book describes several different framing techniques:
-Metaphors
-Stories
-Traditions
-Slogans and catchphrases
-Symbols and objects
-Contrast
-Spin
• Understanding framing is particularly important for those trying to lead opinion or manage perception of meaning
-Politicians, advertisers…
64
In the following quote from the BBC TV series Yes Prime Minister, Sir Humphrey demonstrates to Bernard Woolley how to frame a survey question in order to get the desired response …
http://www.yes-minister.com/ypmseas1a.htm
© 1997-2006 The Yes (Prime) Minister Files
Sir Humphrey: […] Mr. Woolley, are you worried about the number of young people without jobs?"Bernard Woolley: "Yes"Sir Humphrey: "Are you worried about the rise in crime among teenagers?"Bernard Woolley: "Yes"Sir Humphrey: "Do you think there is a lack of discipline in our Comprehensive schools?"Bernard Woolley: "Yes"Sir Humphrey: "Do you think young people welcome some authority and leadership in their lives?"Bernard Woolley: "Yes"Sir Humphrey: "Do you think they respond to a challenge?"Bernard Woolley: "Yes"Sir Humphrey: "Would you be in favour of reintroducing National Service?"Bernard Woolley: "Oh...well, I suppose I might be."Sir Humphrey: "Yes or no?"Bernard Woolley: "Yes"
© 1997-2006 The Yes (Prime) Minister Files
Sir Humphrey: "Mr. Woolley, are you worried about the danger of war?"Bernard Woolley: "Yes"Sir Humphrey: "Are you worried about the growth of armaments?"Bernard Woolley: "Yes"Sir Humphrey: "Do you think there is a danger in giving young people guns and teaching them how to kill?"Bernard Woolley: "Yes"Sir Humphrey: "Do you think it is wrong to force people to take up arms against their will?"Bernard Woolley: "Yes"Sir Humphrey: "Would you oppose the reintroduction of National Service?"Bernard Woolley: "Yes"
© 1997-2006 The Yes (Prime) Minister Files
67
Ethics and decision-making
• Ethics in professional decision-making is very important and it addresses issues that usually cannot be quantified, nor are they simple personality variables.
• ‘Ethics is not about doing what is legal and avoiding what is illegal; it is about the motives and principles which guide problem solving and decision-making in the grey areas where an action may be legal, but may not be right.’
(J. Harrison, Ethics for Australian Business)
68
Risks may be perceived
differently by different
people,
and the same risk may be
perceived differently
according to how it is framed
Recap
69
Decision-making theories and approaches
• Decision-making has been studied in informatics, economics, management, psychology, political science and has been applied and adapted in several other fields.
• In this section we describe some of the main theoretical concepts and models for decision-making that are used in informatics.
1. Introduction2. The nature of decision -
individuals3. The nature of decision - groups4. Uncertainty and risk5. Decision-making theories
and approaches 6. Methods of supporting decision-
making7. Summary
Non-programmable decision – a decision which corresponds to an unstructured or infrequent problem. There no well-known set of procedures to handle the
decision and the situation needs to be treated as a one-off.
Programmable decision – a decision which is routine: it usually occurs frequently, its structure is
clear, and the way to handle it is known. Corresponds to a structured problem.
Semi-structured decision – a decision where some aspects are programmable, but others
required judgment
71
Theories in decision-making
• Normative theories prescribe what should occur in an ideal world
- Classical decision theory
• Descriptive theories describe the behaviour that people actually display
- Behavioural decision theory (also called administrative decision theory)
72
Classical decision theory
• The classical approach to decision-making assumes an ideal situation:
- there is no uncertainty
- all the information is available
- consequences are known
- there is a clear criterion to evaluate the choices
• The theory assumes perfect rationality, so any objective decision maker would make the same choice consistently.
73
Classical decision theory
• Games such as chess have perfect information and a clear goal
-Both players know the entire move history and each other’s objectives, so everything is conducted in an atmosphere of ‘transparency’.
-There are also methods for assessing which moves are stronger than others, although some judgment may be involved in choosing between alternatives.
• Because chess decisions can be modelled according to classical theory, it is possible to program computers to make these decisions to a very high level
74
Utility
• Utility is the perceived value of a decision outcome
• When outcomes can be valued and the probability of those outcomes can be assessed, strategies for choosing can be determined.
• Cost-benefit analysis is a practical application of the concept of utility within economics
75
Cost-benefit analysis
• A new information system that automates reorder requests will cost $50,000 to develop and $1,000 annually over 10 years
• But it will save the $15,000 pa salary of the worker who did the reordering manually
- A cost-benefit analysis shows that the system will pay for itself in 4 years
• An alternative system, costing $25,000 and $2,000 annually is an even better deal, costing $45,000 rather than $60,000 over the 10 year period
76
Behavioural approaches
• Behavioural theories of decision-making describe how people actually make decisions, which is not always in line with the idealised cases posited by normative or prescriptive theories
• Examples include:
- The administrative model of Simon
- The garbage can model of Cohen et al.
77
Administrative model
• Prescriptive models assume that- all information is known and an optimal outcome can be determined
• Simon proposed that practical decision making was actually bound by
- Incomplete information-Uncertainty regarding courses of action and their constraints
• Concept of bounded rationality:-A decision needs only to be ‘good enough’-Given limited time, resources and information, models that make realistic assumptions are of more practical use
78
Garbage can model
• Emphasises the situated context and general messiness of real life organisational decision making:
-Decisions are made in specific contexts of time and place
-With whatever information is available
-And participation from whoever happens to be there
-Decision goals may be ambiguous or unclear
-And people may disagree about how to get there
• No mathematically optimal solution – just ‘muddling through’
79
Classical decision theory assumes an
ideal situation of complete information and
no uncertainty, with perfect rationality.
In contrast, behavioural theories seek to
represent the ways in which decisions are
actually made in practice.
Recap
80
Representing decisions
• There are many different ways to represent the logic of a decision, including:
• Utility matrices• If-then rules• Decision tables and decision trees
Doesn’t rain
Rains
Take umbrella
Don’t take umbrella
It might rain today. Should I take my umbrella?
It might rain today. Should I take my umbrella?
Doesn’t rain
Rains
Take umbrella
Don’t take umbrella
If I didn’t take it and it didn’t rain, I would
be totally happy!
100 1
If I took it and it didn’t rain,
I would be a bit unhappy, but not too
much
10
If I didn’t take it and it rained,
I would be REALLY unhappy!
If I took it and it rained,
I would be quite happy
50
step
Doesn’t rain
Rains
Take umbrella
10 50
Don’t take umbrella 100 1
Totally happy Really unhappy
A bit unhappy Quite happy
It might rain today. Should I take my umbrella?
84
Utility matrix with probabilities
• The numbers in the cells reflect my subjective preferences of taking my umbrella versus possibly getting wet
• But the expected utility also depends on how likely it is to rain – so these probabilities can also be included
It might rain today. Should I take my umbrella?
Doesn’t rainp=0.7
Rains
p=0.3
Take umbrella
10 50
Don’t take umbrella 100 1
It might rain today. Should I take my umbrella?
Doesn’t rainp=0.7
Rains
p=0.3
Take umbrella 10 50Don’t take umbrella 100 1
Expected utility of taking umbrella
= 10 x 0.7 + 50 x 0.3 = 22
Expected utility of NOT taking umbrella
= 100 x 0.7 + 1 x 0.3 = 70.3
Don’t take the umbrella!
step
87
Utility matrix - summary
• Organises choices among actions and possible outcomes in the form of a table
• Each cell of the table reflects the relative utility of that combination of action and outcome
• Probabilities can be included in the possible outcomes
• Overall utility can be calculated to give the decision that maximises utility
88
Rules
• Indicate what action to take in a particular situation or condition:
IF it is raining, THEN take your umbrella
• Or what consequences would be entailed:
IF it is raining and you do not take your umbrella, THEN you will be unhappy
89
Rules
• An ELSE structure can be used to indicate a different action if the condition does not hold:
IF milk levels are low, THEN buy more milkELSE keep monitoring levels
• There may be many possible condition-action pairs in a single rule
• If-then rules are easily programmable and are the basis of many computer programs for supporting decisions
90
Decision tables and decision trees
• A simple way of representing complicated condition-action pairs, such as those found in regulations
• Decision tables and decision trees are very common structures in informatics, especially for representing procedures when analysing requirements for an information system
91
Students must have paid all their due fees to the university in order to receive a grade for the courses in which they are enrolled. Any failure to pay fees to the university will result in a grade of ‘result withheld’ being awarded. Furthermore, for a student to be eligible to be awarded a grade, that student must be recorded as being enrolled in the relevant subject prior to the end of semester and regardless of the attendance record and academic performance in the assessed components o f that subject. A failure to enrol in the subject will result in a grade of ‘result withheld.
What conditions and actions can be identified here?
92
… Furthermore, for a student to be eligible to be awarded a grade, that student must be recorded as being enrolled in the relevant subject prior to the end of semester and regardless of the attendance record and academic performance in the assessed components o f that subject. A failure to enrol in the subject will result in a grade of ‘result withheld’.
Condition: ENROLLED or NOT ENROLLED
Actions: ISSUE GRADE or WITHHOLD RESULT
93
Students must have paid all their due fees to the university in order to receive a grade for the courses in which they are enrolled. Any failure to pay fees to the university will result in a grade of ‘result withheld’ being awarded.
Condition: PAID FEES or NOT PAID FEES
Actions: ISSUE GRADE or WITHHOLD RESULT
94
Decision tree
• The tree could also be drawn with the first branch being ‘Paid fees?’
95
Decision table
• Rules can be generated from the table. There are four rules here, eg:
IF Enrolled AND Paid Fees THEN Issue Grade
• Exactly the same information can be shown as a table:
96
Utility matrices, if-then rules,
decision tables and decision
trees are some of the methods
by which the logic of a decision
can be represented.
Recap
97
Multi-criteria decision-making
• When a decision needs to be made based on many different criteria
• Think about how you will buy your next computer – how do you balance:
- Price?
- Laptop vs desktop?
- Processor?
- Extended warranty?
- Multimedia capability?
- …
98
Which car?
• The Barista is cheapest…• … and the fastest acceleration• But the Yappy is tallest…• And the Git is most fuel efficient…
Which of these attributes is most important to you?
99
Which car?
• Now RANK the attributes instead of using their absolute values
• And WEIGHT the attributes according to which is most important to you
-Here cost is most important, rated at 40%, height is relatively unimportant at 10%
100
Which car?
• The weighted score can now be calculated• This shows the Yappy is worst overall (highest score)
and can be ruled out
101
Multi-criteria decision-making
• There are many different methods for multi-criteria decision-making, based on the assumptions made and the weightings allocated
• Multi-criteria decision-making is well suited to spreadsheet-based analysis, and is widely used in informatics.
102
Prediction
• Many decisions implicitly or explicitly entail forecasting or prediction. Prediction and forecasting is big business
• If an outcome of a decision can be determined in advance it is a safe bet
• In uncertain situations, some things may indicate the causes of outcomes. If the things that cause or determine that outcome can be identified, you are at a betting advantage
-The ability to find such indicators is part of the skill of prediction
103
Multi-criteria decision-making
involves many different and
sometimes competing criteria.
Choices can be represented using
weighted attributes to compare
options.
Recap
104
Strategic decision-making
• Strategic decisions:
- take a long view
- payoffs are usually related to longer term outcomes, not quick wins
- may involve short term ‘sacrifices’ to get a better final position
• Some modelling techniques:
- Game theory
- Prisoner’s dilemma
- Scenario analysis
105
Game theory
• Game theory is the study of decisions when more than one player is involved, and the costs and benefits of actions are not fixed but depend on the choices made by the other players
A banker shows you three boxes, labelled with the number of dollars in each:
The banker secretly removes the bottom of one box, and asks you to put the amount indicated into any two of the boxes:
In this case the banker removes the bottom from the $4 box, and you put money into the $1 and $4 boxes – you get back $1 but lose $4 – a net loss of $3
step
107
Payoff matrix for banker game
• The matrix shows that if you are conservative and play I, you will only ever lose a maximum of $1
• - however the potential gains are not as large as with II and III
• But the banker also knows all this, so can modify strategy accordingly, e.g. by never playing BI, where you always win something
108
Game theory
• The banker game is known as a zero-sum game, because one person’s gains exactly equal the other’s losses
- (the banker’s payoff matrix is exactly the same but reversed)
• There are many different types of game, and game theory has been used in many disciplines such as economics and biology to study strategic behaviour
109
Cooperation or self-interest: The prisoner’s dilemma
• You are one of two prisoners, captured on suspicion of some crime
• You are put in separate cells• The jailer tells you there is enough circumstantial
evidence to convict you for 2 years• However, if you betray the other prisoner, you will be
let off, and the other prisoner will get 5 years• … but the other prisoner has probably been offered
the same deal
WHAT DO YOU DO?
110
Payoff matrix for prisoner’s dilemma
• If you both say nothing, you’ll each get a lesser sentence• But if you say nothing and the other dobs you in, you’ll get
5 years
Can you trust the other prisoner??
111
Prisoner’s dilemma variations
• One common variation of Prisoner’s Dilemma has a lesser sentence (eg 3 years) if you each dob the other in
-How does this change your decision?
• Iterated Prisoner’s Dilemma looks at the long term – how to play to minimise total number of years in jail?
112
Scenario analysis
• Where future states are considered in advance and assessments and decisions made against likely outcomes
• Often called ‘what if’ analysis:- What if interest rates go up 0.25% ?- What if I change my contract to 60% ?
• Spreadsheets are ideal for this- Many spreadsheets have built-in tools to answer this sort of question
113
Strategic decision-making is important when
long-term viability or survival is valued over short-
term gains.
Some long term problem situations are complicated
by other people’s decisions affecting the
consequences of your own. Game theory can
address these situations.
Recap
114
Methods of supportingdecision-making
• As we have seen, many decisions can be formally represented and quantified, either from the nature of the data or from a ranking of similar process.
• When this is done, decision-making becomes amenable to computer-based support
• Decision support systems are a class of system developed primarily within organisational informatics to support management activity
1. Introduction2. The nature of decision - individuals3. The nature of decision - groups4. Uncertainty and risk5. Decision-making theories and
approaches 6. Methods of supporting decision-
making7. Summary
115
Decision support systems
• Expert systems were first developed in the early 1970s, to make logical conclusions from data in the same way as a human expert
-Assumed perfect information
• Decision support systems support semi-structured problems, and combine:
-strengths of computer in analysing data into useful form
-human ability to see relevance in context and other hard to automate aspects
116
Types of decision support systems
• Group decision support systems (GDSS) provide support for process of meetings
-Negotiation support systems are a type of GDSS that focus on facilitating communication in situations where participants have strong disagreements
• Meetings are virtual rather than face to face
-Greater involvement of all members and focus on issues rather than personality
-But can be slower, less expressive of complex arguments, loss of non-verbal cues
117
Decision support systems assist in
organisational decision-making by
automating those parts of a decision that
are amenable to computation.
Many different types of DSS exist, such
as Group DSS and Negotiation SS.
Recap
118
Summary• Decisions are an integral part of everyday life, and a
key topic in informatics
• Decision-making involves many human and psychological factors
• Cognitive biases can influence individual decision making
• Making decisions on behalf of a group involves some collective process of aggregating individual decision
• Risk, uncertainty and framing affect how a decision is perceived
119
Summary
• Classical decision theory is based on a rational approach to decision-making, with perfect information
• Behavioural approaches focus on what actually happens when people make decisions
• Tools and techniques such as matrices, decision tables and decision trees are used to represent the information and probabilities inherent in a decision
• Tools can be incorporated into software for automated decision support
120