Scientific hypotheses can be said to fail when they lead to predictions that do not match the results found in
experiments. Alternatively, experiments can be regarded as failures when they do not provide helpful information about nature. However, the standards of what constitutes failure are not clear-cut. For example, the
Michelson–Morley experiment became the "most famous failed experiment in history" because it did not detect the motion of the Earth through the
luminiferous aether as had been expected. This failure to confirm the presence of the aether would later provide support for
Albert Einstein's
special theory of relativity.
Wired magazine editor
Kevin Kelly explains that a great deal can be learned from things going wrong unexpectedly, and that part of science's success comes from keeping blunders "small, manageable, constant, and trackable". He uses the example of engineers and programmers who push systems to their limits, breaking them to learn about them. Kelly also warns against creating a culture that punishes failure harshly, because this inhibits a creative process, and risks teaching people not to communicate important failures with others (e.g.,
null results). Failure can also be used productively, for instance to find identify ambiguous cases that warrant further interpretation. When studying biases in machine learning, for instance, failure can be seen as a "
cybernetic rupture where pre-existing biases and structural flaws make themselves known". == In popular culture==