Mapping the Futures a Choice Opens

Why it attracts me

When a choice comes up, I picture several futures growing out of it almost automatically. I like turning that picture into a drawing. A decision tree puts the branches on paper, where I can question them, instead of leaving them as a vague feeling.

The idea

A decision tree has two kinds of branch points. At a decision point, I choose. At a chance point, something I do not control happens, and I can attach a rough estimate of how likely each outcome is (Probability, Risk & Uncertainty). At the ends of the branches sit the results, with some sense of how much each one matters to me. Working backward from the ends, I can compare choices by what they tend to lead to, not just by how they feel today.

The most useful part is often not the final answer. It is finding the points where new information would change my choice. Those points tell me what to learn before deciding, or which option keeps the most doors open (Early Choices That Are Hard to Undo). Learning from what each branch actually brings is the subject of Learning What to Do in Each Situation.

What I think (and don't know)

I think trees are honest about something people often hide: most choices lead to several possible futures, not one. I am less sure how far to trust the numbers on them. The estimates are mine, and they can be wrong. Some things I care about, such as love or meaning, do not fit neatly at the end of a branch.

Where it connects

A branching picture also appears in physics. Hugh Everett's 1957 approach to quantum theory is often read, especially in later many-worlds interpretations (The Many-Worlds Reading of Quantum Physics), as describing a branching structure, and those readings are still debated. The likeness is in the pictures only. A decision tree's probabilities are my estimates about an uncertain choice. They are not the probabilities that quantum physics gives for measurement results, which come from a rule called the Born rule (The Rule Behind Quantum Probabilities). And my decisions do not create worlds. I keep both pictures, and I keep them apart. Branching possibilities are also one meaning that a family name inspired by Everett and Nash carries for me (A Family Name Inspired by Everett and Nash).

An example

My Bedtime Story Engine prototype is a small branching tree. A tiny dragon named Pip sets out to return a lost moon lantern, and at each step the child picks one of two choices, such as stepping through a glowing door or looking for a clue first. A path takes up to five choices and ends in one of two endings. Several paths rejoin along the way, which keeps the tree manageable. The design problem is the same one a decision tree faces: enough branches that choices matter, and enough structure that the whole thing stays coherent. Even a story this small shows how quickly branches multiply, and why some pruning is part of good design.

Questions I am still carrying

• How far ahead is it worth drawing the tree before the guesses become too rough?
• How do I value a branch that mainly teaches me something?
• When does mapping futures help, and when does it become a way to avoid choosing?

Questions I'm still exploring

• How far ahead is it worth drawing the tree before the guesses become too rough?
• How do I value a branch that mainly teaches me something?
• When does mapping futures help, and when does it become a way to avoid choosing?

Sources and further reading

• Stanford Encyclopedia of Philosophy, "Decision Theory"
• Howard Raiffa, Decision Analysis: Introductory Lectures on Choices under Uncertainty (Addison-Wesley, 1968)