Lecture 1
SDM is about choosing between options that affect long-term strategy. These decisions
always involve risk or uncertainty. What you choose often depends on what others choose
too (that is the strategic part).
DM: process of picking one option from a set of mutually exclusive alternatives.
- Input: ideas, values.
- Throughput: process → focus of this course.
- Output: decision.
Decision theory: formal theory made of statements and propositions that explain how
decisions are made, axioms → theorems → hypotheses → models. It describes and
explains decision-making processes. Three versions of decision theory:
1. Descriptive: how people actually make decisions (behavioral).
2. Normative: how people should make decisions (fairness, ethics).
3. Prescriptive: how people could make better decisions (efficiency, inclusiveness).
Rational: choosing the best option based on preferences and values.
Preferences are hidden but revealed through choices. Preferences must be:
- Complete: you can compare any two options.
- Transitive: if A > B and B > C → then A > C.
Violations of preferences:
- Independence of Irrelevant Alternatives (IIA): adding a third option shouldn’t change
your preference between two others.
- Cyclical preferences: this breaks transitivity.
Types of Decision-Making problems:
- Certainty: you know the outcome of each option.
- Risk: multiple outcomes, with known probabilities.
- Uncertainty: multiple outcomes, probabilities are unknown.
- Structural ignorance: you don’t even know the possible outcomes.
Decision matrices:
- Rows: options
- Columns: criteria
- Cells: scores/ utilities
How to come to a decision, you can use:
- Lexicographic: choose based on the most important criterion first.
- Elimination by aspects: remove options that fail on key criteria.
- Sequential: compare options step-by-step.
- Regret rule: minimize the pain of choosing badly. You compare each option to the
best score in its column.
- Step 1: for each column, find the maximum value.
- Step 2: subtract each option's value from that maximum.
- Step 3: for each row, find the maximum regret.
, - Step 4: choose the option with the smallest maximum regret.
- Maxmin rule: select the best of the worst. For each option, you look at the worst
outcome it could give you.
Effects:
- Decoy effect: a weak option can influence how attractive others seem.
- Unique attributes: a special feature can sway decisions (only one option has a
specific feature).
,Lecture 2
Complex decision-making
DM is complex when: there are many actors, conflicting interests, many alternatives,
different perspectives, uncertainty is involved etc. When this is the case, our usual heuristics
are not enough. Heuristics are mental shortcuts or rules of thumb (these are flexible, used in
everyday decisions, based on experience and often good enough). But for complex
decisions, heuristics can fail → we need structural methods/ decision analysis.
There are two types of decision analysis:
1. Descriptive: studies how people actually make decisions. Shows that humans are
often not rational.
2. Prescriptive: suggests how people should make decisions. Helps to identify the ‘best’
decision. Assumes rationality and complete information.
- The aim is to support DM, so people make better decisions. It helps, but it
does not decide for you.
What makes a good decision:
- Considers all alternatives.
- Includes stakeholders.
- Includes their values.
- Uses scientific knowledge.
- Understands consequences.
- Is consistent.
- Is transparant.
- Ideally leads to consensus.
Multi-Criteria Decision Analysis (MCDA)
MCDA = family of techniques for decisions with multiple objectives and multiple options.
SMART is one technique in this family.
Alternative: the options you can choose
Attribute / Criterion: a feature used to evaluate alternatives
Value: the score of an alternative on a criterion
Objective: the overall goal (e.g., maximize income)
Weight: importance of a criterion
Axioms of MCDA (SMART), these must hold for SMART to work:
- Decidability (Completeness): you can compare any two options.
- Transitivity: if A > B and B > C, then A > C.
- Summation: strength of preference must be consistent.
- Solvability: unknown values can be determined by asking the decision maker.
- Finite boundaries: best and worst values must be finite (no infinity).
, Steps of MCDA:
1. Define and structure the decision problem. What is the decision? Who are the
stakeholders?
2. Identify objectives and attributes. Build objectives hierarchy (value tree).
3. Identify options. Generate alternatives.
4. Predict outcomes. For each option, determine performance on each criterion. Include
uncertainty.
5. Elicit preferences. Determine weights using swing weighting.
6. Integrate everything. Use weighted additive rule. Rank options. Sensitivity analysis.
7. Discuss results with stakeholders. Look for consensus. Possibly iterate.
Step 1-3: structure the problem.
Step 4-6: getting the information + calculations.
How to structure the problem: objectives → criteria → attributes → indicators.
Measuring performance of alternatives:
- Qualitative criteria → can’t be measured → direct rating. Steps:
1. Pick the best and worst alternative for that criterion.
2. Assign 0 to the worst and 100 to the best.
3. Rate the others in between based on preference strengths.
Give each option a score and make it so that gaps reflect how much better an option
is compared to another option.
Example: nice colleagues
SDM is about choosing between options that affect long-term strategy. These decisions
always involve risk or uncertainty. What you choose often depends on what others choose
too (that is the strategic part).
DM: process of picking one option from a set of mutually exclusive alternatives.
- Input: ideas, values.
- Throughput: process → focus of this course.
- Output: decision.
Decision theory: formal theory made of statements and propositions that explain how
decisions are made, axioms → theorems → hypotheses → models. It describes and
explains decision-making processes. Three versions of decision theory:
1. Descriptive: how people actually make decisions (behavioral).
2. Normative: how people should make decisions (fairness, ethics).
3. Prescriptive: how people could make better decisions (efficiency, inclusiveness).
Rational: choosing the best option based on preferences and values.
Preferences are hidden but revealed through choices. Preferences must be:
- Complete: you can compare any two options.
- Transitive: if A > B and B > C → then A > C.
Violations of preferences:
- Independence of Irrelevant Alternatives (IIA): adding a third option shouldn’t change
your preference between two others.
- Cyclical preferences: this breaks transitivity.
Types of Decision-Making problems:
- Certainty: you know the outcome of each option.
- Risk: multiple outcomes, with known probabilities.
- Uncertainty: multiple outcomes, probabilities are unknown.
- Structural ignorance: you don’t even know the possible outcomes.
Decision matrices:
- Rows: options
- Columns: criteria
- Cells: scores/ utilities
How to come to a decision, you can use:
- Lexicographic: choose based on the most important criterion first.
- Elimination by aspects: remove options that fail on key criteria.
- Sequential: compare options step-by-step.
- Regret rule: minimize the pain of choosing badly. You compare each option to the
best score in its column.
- Step 1: for each column, find the maximum value.
- Step 2: subtract each option's value from that maximum.
- Step 3: for each row, find the maximum regret.
, - Step 4: choose the option with the smallest maximum regret.
- Maxmin rule: select the best of the worst. For each option, you look at the worst
outcome it could give you.
Effects:
- Decoy effect: a weak option can influence how attractive others seem.
- Unique attributes: a special feature can sway decisions (only one option has a
specific feature).
,Lecture 2
Complex decision-making
DM is complex when: there are many actors, conflicting interests, many alternatives,
different perspectives, uncertainty is involved etc. When this is the case, our usual heuristics
are not enough. Heuristics are mental shortcuts or rules of thumb (these are flexible, used in
everyday decisions, based on experience and often good enough). But for complex
decisions, heuristics can fail → we need structural methods/ decision analysis.
There are two types of decision analysis:
1. Descriptive: studies how people actually make decisions. Shows that humans are
often not rational.
2. Prescriptive: suggests how people should make decisions. Helps to identify the ‘best’
decision. Assumes rationality and complete information.
- The aim is to support DM, so people make better decisions. It helps, but it
does not decide for you.
What makes a good decision:
- Considers all alternatives.
- Includes stakeholders.
- Includes their values.
- Uses scientific knowledge.
- Understands consequences.
- Is consistent.
- Is transparant.
- Ideally leads to consensus.
Multi-Criteria Decision Analysis (MCDA)
MCDA = family of techniques for decisions with multiple objectives and multiple options.
SMART is one technique in this family.
Alternative: the options you can choose
Attribute / Criterion: a feature used to evaluate alternatives
Value: the score of an alternative on a criterion
Objective: the overall goal (e.g., maximize income)
Weight: importance of a criterion
Axioms of MCDA (SMART), these must hold for SMART to work:
- Decidability (Completeness): you can compare any two options.
- Transitivity: if A > B and B > C, then A > C.
- Summation: strength of preference must be consistent.
- Solvability: unknown values can be determined by asking the decision maker.
- Finite boundaries: best and worst values must be finite (no infinity).
, Steps of MCDA:
1. Define and structure the decision problem. What is the decision? Who are the
stakeholders?
2. Identify objectives and attributes. Build objectives hierarchy (value tree).
3. Identify options. Generate alternatives.
4. Predict outcomes. For each option, determine performance on each criterion. Include
uncertainty.
5. Elicit preferences. Determine weights using swing weighting.
6. Integrate everything. Use weighted additive rule. Rank options. Sensitivity analysis.
7. Discuss results with stakeholders. Look for consensus. Possibly iterate.
Step 1-3: structure the problem.
Step 4-6: getting the information + calculations.
How to structure the problem: objectives → criteria → attributes → indicators.
Measuring performance of alternatives:
- Qualitative criteria → can’t be measured → direct rating. Steps:
1. Pick the best and worst alternative for that criterion.
2. Assign 0 to the worst and 100 to the best.
3. Rate the others in between based on preference strengths.
Give each option a score and make it so that gaps reflect how much better an option
is compared to another option.
Example: nice colleagues
