Expected Value Decision Rule
- Difficulty
- Moderate
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- —
Expected value is a decision rule for evaluating actions whose chance of making a difference is small but whose potential impact is enormous. Instead of dismissing an action because it probably will not be decisive, estimate the probability that it changes the outcome, estimate the value created if it does, and multiply the two. MacAskill applies this reasoning to voting: one ballot is unlikely to swing an election, but influencing decisions involving an entire government can be extraordinarily valuable. The rule also applies to consumer choices that only occasionally change a supplier's next order. It is most useful when comparing opportunities consistently rather than treating low probability as equivalent to zero probability.
Origin
Extracted from Deep Dive with Ali Abdaal
How to run it
- 1
Define the consequential outcome
Specify exactly what would change if the action became decisive, such as which government wins an election or whether a supplier orders another crate.
Pro tip Describe the outcome in concrete social, financial, or welfare terms.
- 2
Estimate the pivotal probability
Estimate the chance that your action is the one that changes the outcome, even if that probability is very small.
Pro tip Use ranges when a precise probability would create false confidence.
Watch out Do not replace a very small probability with zero merely because the action usually makes no visible difference.
- 3
Estimate impact if pivotal
Assess how much value would be created or protected in the scenario where the action really does affect the outcome.
Pro tip Include the full scale of the decision rather than only its immediate personal effects.
- 4
Multiply probability by impact
Combine the probability and consequence to obtain the action's expected value.
Watch out Large consequences do not justify assigning an unrealistically high probability.
- 5
Compare available alternatives
Compare expected values while accounting for time, money, location, expertise, and other opportunity costs.
Pro tip For voting, consider whether the election is competitive and whether you have an informed preference.
In the wild
A single ballot is very unlikely to swing an election, but elections determine governmental decisions involving enormous budgets and effects on millions of people. MacAskill argues that multiplying the tiny probability of being pivotal by the value of the resulting governmental difference can make voting comparable to a substantial charitable donation.
→ Voting can emerge as a high-impact use of a small amount of time.
Buying or declining one factory-farmed chicken usually does not visibly affect stock. Occasionally, however, aggregate demand causes a supermarket to order eleven crates rather than ten. In that pivotal case, one consumer choice contributes to a much larger production change.
→ An action can have meaningful expected impact even when most individual instances appear inconsequential.
Common mistakes
Treating unlikely as impossible
A low probability of changing an outcome is not the same as having no expected impact.
Ignoring the size of the outcome
Evaluating only the probability misses cases where a tiny chance is paired with an exceptionally consequential result.
Using inflated probabilities
Expected-value reasoning becomes misleading when hopeful assumptions replace defensible probability estimates.
From the transcript
“once you do the maths I actually think voting comes out as quite a high impact activity”
“the very very low probability of making a difference multiplied by the enormous value if you do have a difference that's called multiplying those two…”
From the episode
Moral Philosopher Will MacAskill on What We Owe The Future