# IBM Watson
**Entity class:** Question-answering system / evidence ranking
## Definition and history
IBM Watson answered open-domain quiz clues by generating candidates, weighing evidence, and estimating confidence before deciding whether to answer. [[articles/Intelligence Inside|Intelligence Inside]] uses its 2011 public demonstration as a stage in the lineage of conversational access to predictive and retrieval machinery. The system's visible natural-language response depended on many internal scoring and evidence-selection operations. A voice or sentence at the output could make that deeper architecture appear singular.
## Mechanism and limits
Confidence thresholds connect evidence assessment to an action under uncertainty. Candidate generation, passage retrieval, scoring, and answer selection are separable technical steps. The system illustrates how language can serve as an interface to a larger apparatus of ranking rather than the origin of that apparatus.
## Relationships
**Developer:** [[wiki/IBM|IBM]]. **Earlier interfaces:** [[wiki/ELIZA|ELIZA]] and [[wiki/CALO|CALO]]. **General mechanism:** [[wiki/Predictive Engines|Predictive Engines]]. **Article route:** [[articles/Intelligence Inside|Intelligence Inside]].
## Sources
[IBM Research, *Introduction to This Is Watson*](https://research.ibm.com/publications/introduction-to-this-is-watson); [[articles/Intelligence Inside|Intelligence Inside]].