Unpacking Principle (Support Theory)
The unpacking principle in support theory describes how people assign higher probability to explicit disjunctions (e.g., dying from heart disease, cancer, or other natural causes) than to coextensional implicit disjunctions (e.g., dying from natural causes), demonstrating subadditivity. This cognitive bias affects probability and frequency judgments, increasing with the number of unpacked components.
THE IDEA
Unpacking Principle
The unpacking principle is a core component of support theory, which addresses how individuals judge probabilities for explicit versus implicit disjunctions. It consists of two parts:
- Additivity: For explicit disjunctions, the probability of the whole is approximately equal to the sum of its parts.
- Subadditivity: For implicit disjunctions, the probability of the whole is less than the sum of its parts when unpacked explicitly.
Together, these lead to nonextensionality in probability judgment, meaning the way an event is described (packed vs. unpacked) alters its perceived likelihood.
Empirical Evidence: Study 1 – Causes of Death
Participants (Stanford undergraduates) estimated the likelihood of various causes of death. They were told that about 2 million people die annually in the U.S. and were asked to assess either:
- An implicit hypothesis (e.g., death from natural causes), or
- A coextensional explicit disjunction (e.g., death from heart disease, cancer, or other natural cause).
Results (Table 25.1):
- Implicit estimate for natural causes: 58%
- Sum of explicit estimates (heart disease: 22%, cancer: 18%, other natural: 33%): 73%
- The unpacking factor (ratio of explicit sum to implicit) was 1.26, indicating subadditivity.
- Mean unpacking factors: 1.37 for three-component hypotheses, 1.92 for seven-component hypotheses.
- Probability judgments were higher and more subadditive than frequency judgments.
Contrast with Other Models
Support theory predicts:
- α ≤ β (implicit focal less than explicit focal)
- γ ≤ δ (implicit residual less than explicit residual)
- β = γ (additivity for explicit disjunctions)
This contrasts with:
- Bayesian model: Assumes extensionality (α=β, γ=δ) and additivity (β=γ).
- Belief function theory: Assumes superadditivity (β ≥ γ).
- Regressive model: Assumes extensionality (α=β, γ=δ) and subadditivity for explicit disjunctions (β ≤ γ).
Support theory's predictions were supported: α < δ and β = γ, while the regressive model's prediction of β < γ was not significant.
Key Implications
- The unpacking factor is a measure of subadditivity.
- In the reported study, subadditivity was greater with more components in the explicit disjunction.
- Monetary incentives did not improve accuracy in the research described.
- The effect appeared in both probability and frequency judgments.
When to use it
Risk management: compare estimates for project failure as a whole with estimates for specific failure modes. Examine omissions and overlap; a higher estimate after unpacking is not automatically more accurate.
Research on medical judgment: examine whether listing possible diagnoses changes probability estimates for the same set of outcomes.
Forecasting: compare an aggregate forecast with a breakdown of contributing factors, checking that the components cover the same outcome without double-counting.
Legal reasoning: examine whether different descriptions of the same possibilities change judgments, while distinguishing that framing effect from new evidence.
FROM THEORY TO EVERYDAY LIFE
In practice
Causes of Death Study
Participants estimated the probability of death from natural causes (implicit) as 58%, but when the same set was unpacked into heart disease (22%), cancer (18%), and other natural causes (33%), the sum was 73%—a clear demonstration of subadditivity. This shows that unpacking a category increases perceived likelihood.
Fault Tree Analysis
In fault tree analysis, listing specific failure modes (e.g., software bug, hardware failure, user error) leads to higher estimated probability of system failure compared to a general 'system failure' category, illustrating the unpacking effect in risk assessment.
A MOMENT TO REFLECT
Put it to work.
Think of a recent decision where you assessed the likelihood of an event. Did you consider it as an implicit category or did you break it down into explicit components? How might unpacking change your estimate?
Why do you think unpacking increases perceived probability? What psychological mechanisms (e.g., availability, ease of recall) might drive this effect?
Can you identify a real-world scenario where the unpacking principle could lead to overestimation of risk? How would you adjust your decision-making to account for this bias?
Often confused with
References
- Fault trees: Sensitivity of estimated failure probabilities to problem representation — Fischhoff, B., Slovic, P., & Lichtenstein, S. (1978)
- Support theory: A nonextensional representation of subjective probability — Russo, J. E., & Kolzow, D. J. (1992)
- A regressive model of probability judgment — Erev, I., Wallsten, T. S., & Budescu, D. V. (1994)
CONTINUE THE CONVERSATION
Source texts
- Heuristics and Biases The Psychology of Intuitive Judgment (Thomas Gilovich, Dale Griffin, Daniel Kahneman)