Bongard

Bongard problems are visual puzzles created by Russian computer scientist Mikhail Moiseevich Bongard. They present two sets of visual patterns or images, labeled as Set A and Set B. The challenge lies in identifying a discriminating feature, a visual rule or concept, that is present in all images in Set A but is absent in all images in Set B, or vice versa. The discriminating feature is often subtle and not immediately obvious.

Bongard problems are designed to test and develop skills in pattern recognition, abstract thinking, and hypothesis formation. Solving them involves observing the images, formulating possible rules, testing those rules against the images in both sets, and refining or rejecting rules until a satisfactory discriminating feature is found. There is typically no single "correct" answer, as multiple valid solutions may exist depending on the level of abstraction considered.

The problems are often used in the field of artificial intelligence to explore and evaluate machine learning algorithms and their ability to learn visual concepts. The difficulty of Bongard problems can vary greatly, ranging from relatively simple to extremely challenging, making them a versatile tool for cognitive research and educational purposes. The inherent ambiguity and the need for flexible thinking contribute to their enduring appeal as a test of human intelligence.