People prefer known odds to unknown ones.
Given a choice, people pick the option whose outcome they can estimate over one with missing information, even when the unknown option is just as good. Hidden prices, vague estimates, and unclear results push people away.
Daniel Ellsberg described the effect in 1961 with a thought experiment about an urn. It holds 30 balls: 10 red, and 20 that are some unknown mix of black and white. Betting on red has a known one-in-three chance; betting on black has, on average, the same chance, but nobody knows the exact odds. Most people bet on red.
His results became known as the Ellsberg paradox, because they break the standard assumption that people treat unknown probabilities as their best estimate. Later work found the effect strongest when the missing information is called to attention, and weaker when people feel competent in the domain, as Chip Heath and Amos Tversky showed in 1991.
People tend to choose options whose chance of a good outcome is known over options where it is unknown, even when the unknown option is no worse. Missing information reads as risk, so people steer away from it or stop to look for more.
Most people prefer to bet on the 10 known red balls rather than on black balls of unknown number, though both offer the same average chance. The missing information alone makes the second bet feel worse.
A hidden price is an unknown, and unknowns lose to a competitor with a visible number. Show prices, or at least ranges and examples, so buyers can estimate the cost.
A spinner with no estimate leaves people unable to judge whether to wait. A progress bar or a time estimate turns an ambiguous wait into a known one.
Delivery dates, response times, and refunds shown as ranges are better than nothing. Even an honest window removes the ambiguity that makes people hesitate at checkout.
A button whose result is unclear is an ambiguous bet. Previews, confirmation summaries, and undo make the outcome known before people commit.
A new feature, plan, or vendor has no track record, so people stick with what they know. Trials, guarantees, and evidence from others reduce how unknown it feels.
Ratings and reviews replace an unknown with a known quantity. A product with no reviews can lose to one with mediocre reviews, because some information beats none.
Heath and Tversky found people bet more readily on ambiguous events in areas where they feel knowledgeable. Explaining how something works can make an uncertain option feel safer.
AI output is uncertain by nature. People are more willing to rely on it when they can see how confident it is and what it is based on.
An answer with no sign of how reliable it is feels like Ellsberg’s unknown urn. Show confidence, sources, or the basis for an answer, so people can judge it.
An agent whose next step is unknown makes people hesitate to let it act. Show the plan before it runs, and let people approve or edit the steps.