Halliday & Matthiessen (1999: 112):
The projection of ideas has also played an important rôle in hearer modelling in AI and computational linguistics. From a systemic-functional point of view, hearer modelling corresponds to one aspect of the interaction base: the interactants in a dialogue have to be able to assess and model the way their experiences diverge. This is one measure of overall interpersonal distance between them. In hearer modelling, this tends to be construed ideationally by means of cognitive projection, where the projected idea is represented as a separate partition or space within overall knowledge: ‘the speaker believes —> the hearer knows —> …’. Such models may involve long projecting sequences (often called “nested beliefs”) …