Distinction Bias: Why Options Look More Different Side by Side

The bias comes from a mismatch between two modes of evaluation. When you choose, you are usually in joint evaluation: the options sit side by side and every difference is visible. When you live with the result, you are in single evaluation: one job, one apartment, one television, and nothing next to it. Hsee and Zhang's claim is that people in joint evaluation predict their feelings as if the comparison would stay, and so overpredict how much the difference will matter once it is gone.

Which kind of difference is being judged is the key. Some attributes have a natural reference point and are easy to feel on their own: a success against a failure, a job with health insurance against one without, a book that sold against one that did not. Hsee and Zhang call these qualitative differences, and people judge them about as well alone as in comparison. Others are a matter of degree: 80 buyers against 160, a salary of 60,000 dollars against 70,000, a 72-decibel noise against 78. These quantitative differences are vivid in comparison and nearly invisible alone, because without the alternative there is nothing to tell you whether 80 buyers is a lot or a little. Nobody buying your book feels bad in either mode; 80 buyers feels poor only when 160 and 240 are in view.

Hsee's earlier work on evaluability showed the same mechanism producing preference reversals: a dictionary with 20,000 entries and a torn cover is valued more than one with 10,000 entries and an intact cover when the two are compared, and less when each is seen alone. Distinction bias is what happens when that comparison-driven judgment is used to predict a life lived without the comparison. In the authors' own example, a person comparing an interesting job at 60,000 dollars with a tedious one at 70,000 sees both gaps clearly, chooses the money, and then spends every working day in single evaluation, where 70,000 dollars feels like a salary and tedium feels like tedium.

The record is mixed. In Anvari, Olsen, Hung and Feldman's close replication of 2021, the poem-book study replicated, with joint evaluators again overpredicting the differences between 80, 160 and 240 buyers, while the word-reading study, in which people actually read lists of 10 or 25 negative words, did not. The prediction half of the bias is on firmer ground than the claim about real experience.

Christopher Hsee and Jiao Zhang published "Distinction bias: Misprediction and mischoice due to joint evaluation" in the Journal of Personality and Social Psychology in 2004. The first study is the one the demo reproduces: 249 students imagined selling a book of their own poems on campus. Fifty of them, the joint evaluation group, saw four outcomes together, no buyers, 80, 160 and 240, and predicted their happiness in each on a scale from 1 to 9; the rest were split into four groups that each saw a single outcome. In joint evaluation the predictions climbed steeply with sales: 1.7 for no buyers, 3.3 for 80, 6.3 for 160 and 8.5 for 240. In separate evaluation they were nearly flat above zero: 2.2 for no buyers, then 7.1, 7.6 and 7.9. Both modes agreed only on the qualitative difference, between selling nothing and selling something.

The second study moved to real experience. Three hundred and sixty students either predicted, in joint evaluation, how they would feel after reading 10 or 25 negative or positive words, or actually read one list and reported their mood. Joint evaluators predicted that 25 negative words would be worse than 10, a mean of 3.3 against 4.0; people who read the lists felt no difference, 4.6 against 4.6, and the positive lists showed the same pattern.

The third study put a trade-off on the table: 243 students in China could recall a personal failure and receive a 15-gram chocolate, or recall a success and receive a 5-gram one. Choosing between the two, 65 percent took the failure with the larger chocolate. People simply given one option or the other were happier with the success and the small chocolate, 2.4 against minus 0.7 on the study's scale, and so were the choosers once they had lived through their choice. The chocolate, a quantity, drove the choice; the task, a quality, drove the experience.

The rows below are illustrations, not findings from the studies cited on this page.

The counter is to predict from the mode you will live in: take the comparison away before you decide.

• Evaluate each option alone. Write down how you would feel about option A if it were the only offer, then do the same for B, before comparing.
• Sort the differences. Ask which attributes have a natural reference point you could feel without a comparison, such as interesting against tedious, and which are only a matter of degree, such as 60,000 dollars against 70,000. Give the first kind more weight than the comparison suggests.
• Run the Tuesday test. Picture an ordinary weekday a year into each option. The attributes that show up in that picture are the ones that will matter in single evaluation.
• Ask what you would notice. For any difference of degree, ask whether you would detect it if the alternative had never been shown to you.
• Remember the record. The prediction half of the effect replicates well; the claim about lived experience did not hold up in 2021, so do not use the bias as a license to ignore real differences in degree.

Distinction bias belongs to a family of context effects, and its neighbors work differently.

• It is not the contrast effect, in which a judgment of one thing shifts because of what was seen just before it. Contrast changes the rating in the moment; distinction bias mispredicts how a difference will feel once the contrast is gone.
• It is not the decoy effect, in which adding a third, inferior option shifts choice between the first two. The decoy works on which option wins; distinction bias works on how much the winner is expected to matter.
• It is not the focusing effect, the overweighting of whatever attribute attention happens to be on. Distinction bias is specific to comparison and predicts which attributes get overweighted: the ones that are hard to evaluate alone.
• It is not the less-is-better effect, Hsee's 1998 finding that a smaller option can be valued more than a larger one when each is seen alone. That is a reversal between the two modes; distinction bias is the forecasting error of predicting one mode from the other.