“problem of induction: Problem of justifying the inference from the characteristics of observed instances of a general concept to unobserved instances of the same concept. For example, if all emeralds I have ever seen have been green, what entitles me to draw the inference that all emeralds are green, given that my past observations do not strictly entail (or deductively imply) that are emeralds are green? May we infer that the characteristics of a sample taken from a population are characteristics of the entire population? A quality-control engineer who looks at a sample of 100 lightbulbs produced by a particular manufacturing process and finds that five are defective may conclude that 5% of all bulbs that have been and will be produced by the process are defective. For the engineers inference to be justified, two criteria that must be met are (1) that the sample be random (i.e. every subset of 100 bulbs has an equal chance of being selected for the examination), and (2) that the sample be sufficiently large (in a mathematically precise sense). See also statistics.”
Excerpted from: Stevens, Mark A., Ed. Merriam-Webster’s Collegiate Encyclopedia. Springfield, Massachusetts: Merriam-Webster, 2000.