Decision Trees

Hey students! 👋 In this lesson, we're going to dive into the fascinating world of decision trees - one of the most powerful tools for making smart choices when facing uncertainty. By the end of this lesson, you'll know how to construct decision trees from scratch, perform rollback analysis to find the best decisions, and incorporate probabilities and utilities to make more informed choices. Think of decision trees as your personal GPS for navigating complex decisions - they'll help you see all possible paths and choose the one that leads to the best outcome! 🎯

Understanding Decision Trees and Their Components

A decision tree is essentially a visual roadmap that helps you analyze decisions involving uncertainty and multiple possible outcomes. Picture it like a family tree, but instead of showing relationships between people, it shows the relationships between decisions, chance events, and their consequences.

Every decision tree has three main building blocks that you need to master. Decision nodes are represented by squares (□) and show points where you have control - where you can choose between different options. For example, if you're deciding whether to study medicine or engineering, that choice point would be a decision node. Chance nodes are shown as circles (○) and represent events beyond your control, like whether it will rain tomorrow or whether you'll pass an exam. Finally, outcome nodes (also called terminal nodes) appear at the end of each branch and show the final result of following that particular path.

The branches connecting these nodes represent the different options available at decision nodes or the possible outcomes at chance nodes. Each branch from a chance node must have a probability assigned to it, and these probabilities must add up to 1.0 (representing 100% certainty that one of the outcomes will occur).

Let's consider a real-world example that many students face: deciding whether to take a gap year before university. Your decision tree might start with a decision node offering two choices - "Take Gap Year" or "Go Straight to University." Each of these branches would then lead to chance nodes representing uncertain outcomes, like whether you'll gain valuable experience during your gap year or whether you'll struggle to get back into academic mode later.

Constructing Your First Decision Tree

Building a decision tree requires careful planning and systematic thinking. Start by clearly defining your decision problem - what exactly are you trying to decide? Then identify all the major decision points and uncertain events that could affect your outcome.

When constructing the tree, always work from left to right, following the chronological order of events. Begin with your initial decision node on the left side of your diagram. From this node, draw branches representing each available option. Label each branch clearly with the specific choice it represents.

Next, determine what happens after each decision. If the outcome is certain, you can go straight to a terminal node. However, if there's uncertainty involved, you'll need to add a chance node. From each chance node, draw branches for all possible outcomes, making sure to assign realistic probabilities to each branch.

Here's a practical example: Imagine students is deciding whether to start a small online business while in school. The decision tree would begin with a decision node offering "Start Business" or "Don't Start Business." If you choose "Start Business," you'd face uncertainty about whether it will succeed. This creates a chance node with branches like "Business Succeeds" (perhaps 30% probability) and "Business Fails" (70% probability). Each of these branches would then lead to outcome nodes showing the consequences - maybe financial gain and experience versus lost time and money.

Remember to keep your tree balanced and comprehensive. Include all realistic options and outcomes, but don't overcomplicate things with extremely unlikely scenarios. The goal is to capture the essential elements of your decision without creating an overwhelming diagram.

Mastering Rollback Analysis

Rollback analysis is the mathematical heart of decision tree evaluation - it's how you determine which path through your tree offers the best expected outcome. The process works backwards from right to left, starting at the outcome nodes and rolling back toward the initial decision.

The technique involves calculating expected values at each chance node by multiplying each outcome's value by its probability, then summing these products. At decision nodes, you simply choose the branch with the highest expected value. This systematic approach ensures you're making decisions based on mathematical logic rather than gut feelings.

Let's walk through a detailed example. Suppose students is considering whether to invest £1000 in cryptocurrency. The decision tree shows two options: "Invest" or "Keep Money Safe." If you invest, there's a 40% chance the investment doubles (outcome: +£1000), a 35% chance you break even (outcome: £0), and a 25% chance you lose everything (outcome: -£1000). If you keep the money safe, you earn 2% interest (outcome: +£20).

For the "Invest" branch, the expected value calculation would be: (0.40 × £1000) + (0.35 × £0) + (0.25 × -£1000) = £400 + £0 - £250 = £150. The "Keep Safe" option gives you £20 with certainty. Since £150 > £20, rollback analysis suggests investing is the better choice mathematically.

However, this is where the beauty of decision trees really shines - they don't just give you a single "right" answer. They help you understand the trade-offs involved. While investing has a higher expected value, it also carries significant risk of loss, which might not be acceptable depending on your personal circumstances.

Incorporating Probabilities and Utilities

Real-world decision making often involves more than just monetary outcomes. This is where utilities come into play - they allow you to assign values that reflect your personal preferences and risk tolerance, not just financial calculations.

Utility theory recognizes that the value of an outcome isn't always proportional to its monetary worth. For example, losing £1000 when you only have £2000 total feels much worse than losing £1000 when you have £100,000. This is called diminishing marginal utility - each additional pound becomes less valuable as your wealth increases.

To incorporate utilities effectively, start by identifying all the different types of outcomes your decision might produce. These could include money, time, stress levels, learning opportunities, or personal satisfaction. Then assign utility scores to each outcome on a consistent scale (often 0-100 or 0-1).

Consider a student deciding between two summer opportunities: a paid internship worth £3000 or an unpaid research position at a prestigious university. The monetary calculation is straightforward - £3000 versus £0. But when you factor in utilities, the research position might offer valuable learning experiences (utility: +40), networking opportunities (utility: +30), and enhanced university applications (utility: +50), while the paid internship offers financial security (utility: +60) but limited learning (utility: +10).

When assigning probabilities, base them on reliable data whenever possible. If you're estimating the chance of rain, check weather forecasts. If you're evaluating business success rates, research industry statistics. For our earlier cryptocurrency example, you might research historical volatility data rather than guessing at probabilities.

Remember that probabilities should reflect your best understanding of reality, not your hopes or fears. It's natural to be optimistic about positive outcomes or pessimistic about negative ones, but accurate probability assessment is crucial for effective decision making.

Conclusion

Decision trees provide a structured, logical framework for tackling complex choices involving uncertainty. By systematically mapping out decisions, chance events, and outcomes, then using rollback analysis to evaluate expected values, you can make more informed choices that align with your goals and risk tolerance. The incorporation of probabilities and utilities allows you to move beyond simple financial calculations to consider the full range of factors that matter to you personally. While decision trees can't eliminate uncertainty from life, they can help you navigate it more confidently and rationally.

Study Notes

• Decision Tree Components: Decision nodes (□), chance nodes (○), and outcome/terminal nodes represent the three building blocks of any decision tree

• Construction Process: Work left to right chronologically, starting with initial decision node, then adding branches for each option and subsequent chance events

• Rollback Analysis Formula: Expected Value = Σ(Probability × Outcome Value) for each branch from a chance node

• Probability Rules: All probabilities from a chance node must sum to 1.0; base estimates on reliable data when available

• Utility Theory: Assign non-monetary values to outcomes reflecting personal preferences and diminishing marginal utility

• Decision Rule: At decision nodes, choose the branch with highest expected value after rollback analysis

• Tree Balance: Include all realistic options and outcomes without overcomplicating with extremely unlikely scenarios

• Evaluation Steps: (1) Calculate expected values at chance nodes, (2) Select best options at decision nodes, (3) Work backwards from outcomes to initial decision