Tallying
Also known as: Unit-weight model, Equal weighting
Count the reasons for each option equally instead of weighting them.
What it means
A fast-and-frugal strategy that decides between options by counting the number of cues favoring each and choosing the one with the higher tally, giving every cue equal weight rather than estimating and applying optimal weights. By ignoring how to weight the evidence, tallying sidesteps the estimation error that plagues complex models built from small or noisy samples, and unit-weight schemes have repeatedly been shown to predict about as well as — and sometimes better than — multiple regression out of sample. Its success is another instance of ecological rationality and the robustness of simple rules: precise weights overfit the idiosyncrasies of the data they were tuned on, whereas equal weights generalize. Tallying contrasts with take-the-best, which uses one cue and ignores the rest, while tallying uses all cues but ignores their differing validities. It matters as practical evidence that careful counting can rival sophisticated statistical weighting in real prediction.
Examples
A clinician predicting an outcome by simply counting how many of several warning signs are present, rather than running a weighted risk formula, often does about as well.
A hiring panel scores applicants by counting how many of six requirements each one meets, with no weightings, and predicts next year's performance about as well as the matrix it replaced.
Deciding between two flats, you count how many must-haves each one hits — light, commute, storage, garden — and take the higher tally instead of agonising over what matters most.
First described in Robyn Dawes (1979); fast-and-frugal program, Gigerenzer & Todd (1999).