This paper introduces the concept of risk-reducing experiments as a basis for designing risk measures. A risk-reducing experiment provides the option to mitigate the impact of less favorable outcomes in a gamble, and the gamble's risk is measured as the increase in value brought about by such an experiment. Two examples are presented, including one based on the concept of expected value of perfect information. Both examples yield familiar risk measures, and extensions of them are discussed. A risk measure derived from a risk-reducing experiment makes explicit the sense in which the riskiness of a gamble is captured. Risk-reducing experiments offer a new approach for conceiving, or choosing among, risk measures.