Joint Cognitive Systems

Anomaly R esponse 79

Recognizing Anomalies

Formal accounts of diagnosis often over look the demands of recogniz ing, out of

many changing data channels , which changes or la ck of ch ange are anomalies, and

which are less significant (or are even distractions). The inherent v ariability of the

physical world means that changes constantly occur in the monitored proc ess that

could be relevant, in principle.

When artific ial intelligence algorithms firs t attempted to au tomate d iagnosis, the

programs relied on p eople to feed them the findings to be e xplain ed (Woods et al.,

1990). Historica lly, system developers usually labeled these as “consultants” or as

“optional advice-givers”; bu t they w ere developed in fact to move toward

autonomous sys tems ...