n. 1. Gratuitous assumptions made inside certain programs, esp. expert systems, which lead to the appearance of semi-intelligent behavior but are in fact entirely arbitrary. For example, fuzzy-matching input tokens that might be typing errors against a symbol table can make it look as though a program knows how to spell.
2. Special-case code to cope with some awkward input that would otherwise cause a program to choke, presuming normal inputs are dealt with in some cleaner and more regular way. Also called 'ad-hackery', 'ad-hocity' (/ad-hos'*-tee/).
See also ELIZA effect.