Skill Guide

Decision theory including expected utility, multi-criteria decision analysis (MCDA), and Bayesian decision networks

Decision theory is a formal discipline that integrates statistical inference (Bayesian decision networks), optimization (multi-criteria decision analysis), and utility modeling to select actions that maximize expected outcomes under uncertainty.

It transforms complex, high-stakes choices from subjective guesswork into quantifiable, auditable processes, directly improving ROI on strategic investments and risk management. Organizations leverage it to systematically navigate trade-offs in product development, resource allocation, and market entry, ensuring choices align with long-term objectives and risk tolerance.

1 Careers

1 Categories

9.0 Avg Demand

15% Avg AI Risk

How to Learn Decision theory including expected utility, multi-criteria decision analysis (MCDA), and Bayesian decision networks

1. Master foundational probability and Bayes' theorem. 2. Learn the expected utility theorem, including risk aversion and prospect theory basics. 3. Understand the structure of a decision matrix and simple weighted scoring.

1. Apply MCDA methods like AHP (Analytic Hierarchy Process) or TOPSIS to real project prioritization problems. 2. Build and solve simple influence diagrams or decision trees using software. 3. Avoid common pitfalls: confusing value with utility, neglecting to structure the problem first, and failing to elicit probabilities systematically.

1. Architect multi-stakeholder decision frameworks that incorporate game theory and mechanism design. 2. Integrate Bayesian networks for dynamic learning in operational systems (e.g., predictive maintenance). 3. Mentor teams on decision hygiene, embedding structured decision protocols into organizational culture and strategic planning cycles.

Practice Projects

Beginner

Case Study/Exercise

Personal Major Purchase Decision Analysis

Scenario

You must choose between buying a new car, a used car, or using a ride-sharing service exclusively for the next 5 years.

How to Execute

1. Define criteria: Total Cost, Reliability, Convenience, Environmental Impact. 2. Assign weights to each criterion based on your values. 3. Score each option (New, Used, Ride-share) on each criterion. 4. Calculate the weighted total score and perform a sensitivity analysis on the weights.

Intermediate

Project

Product Feature Prioritization with MCDA

Scenario

As a Product Manager, you must prioritize the next quarter's feature roadmap from a backlog of 20 features with limited engineering capacity.

How to Execute

1. Select 4-5 criteria (e.g., Revenue Impact, Strategic Alignment, Development Cost, Customer Satisfaction). 2. Facilitate a workshop with stakeholders to agree on criteria weights using AHP pairwise comparisons. 3. Score each feature against the criteria using data and expert judgment. 4. Compute rankings, conduct sensitivity analysis, and present the justified prioritized list.

Advanced

Case Study/Exercise

Strategic Market Entry with Bayesian Decision Network

Scenario

A tech company is deciding whether to launch a new AI product in Market A (regulated, high-uncertainty) or Market B (competitive, lower-uncertainty). Key uncertainties: competitor response, regulatory approval timeline, and customer adoption rate.

How to Execute

1. Structure the problem as an influence diagram with decision nodes (launch market, timing), chance nodes (competitor actions, approval), and a value node (NPV). 2. Elicit conditional probabilities and utility functions from domain experts. 3. Use Bayesian network software to compute expected utilities and identify the optimal policy. 4. Define a monitoring plan for key evidence nodes (e.g., early adoption signals) to update probabilities and potentially revise the decision.

Tools & Frameworks

Mental Models & Methodologies

Multi-Criteria Decision Analysis (MCDA)Bayesian Decision Networks / Influence DiagramsExpected Monetary Value (EMV) AnalysisProspect Theory for Risk Aversion Modeling

Use MCDA (AHP, TOPSIS) for complex trade-offs with multiple stakeholders. Apply Bayesian Decision Networks when decisions are sequential and information will be gathered over time. EMV is the baseline for risk-neutral financial decisions. Prospect Theory helps model actual human behavior in high-stakes choices.

Software & Platforms

Excel/Google Sheets (for basic decision trees & sensitivity)GeNIe/Smile (for Bayesian networks)SuperDecisions (for AHP)Python (pymc3, pgmpy for custom modeling)

Use spreadsheets for transparent, collaborative analysis of simple models. Employ specialized Bayesian network software for complex probabilistic modeling and inference. Use Python for integrating decision models into automated systems or for advanced customization.