Description
Show pairings of a candidate cause and an outcome, ask how strongly one predicts the other, compare judgements with delta-P, and use blocking designs to test cue competition.
People judge whether a feature, alert or action predicts an outcome from the cases they see, and they weight those cases unevenly. In the basic task, the four cells of a 2x2 table (cause present or absent by outcome present or absent) are shown as a summary or one case at a time; judgements are compared with delta-P, the difference between the outcome rate with and without the cause, and regressed on cell frequencies. People weight the cause-and-outcome cell most and the neither cell least, which McKenzie argues can be sensible when events are rare. Cue-competition designs show that credit is shared: once A predicts the outcome, a later companion B gets little credit (forward blocking), and learning about A alone can retrospectively reduce B's credit (backward blocking). The inverse base-rate design shows attention shifting to distinctive cues of rarer categories.
Attributes
- Discipline
- Behavioral Science & Economics
- Loss type
- Unmigrated
- Stage
- Design, Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Choose a candidate cause and outcome relevant to the product question, for example a warning badge and a failed transaction.
- Contingency version: present the four cell frequencies as a table or as a sequence of cases, varying each cell systematically across conditions.
- Ask for a rating of how strongly the cause predicts or causes the outcome, keeping predictive and causal wording distinct.
- Compute delta-P for each condition and regress judgements on cell frequencies to estimate each cell's weight.
- Cue-competition version: train cue A with the outcome, then A and B together, and compare ratings of B with a control group that saw only the compound (forward blocking); reverse the phases for backward blocking.
- Base-rate version: train two categories with a shared cue plus distinctive cues at unequal frequencies, then test with conflicting distinctive cues.
Sources
- Handbook of Cognition (Lamberts & Goldstone, 2005)
Also known as
- Covariation judgement task
- 2x2 contingency judgement with cell-frequency analysis
- Delta-P judgement task
- Forward and backward blocking design
- Relative validity design
- Inverse base-rate (highlighting) design
Categories
- Analytics & Behaviour
- Behavioral Science & Economics