I'm reminded of a set of papers (here's one of them) I've encountered, which treat human language in an rather interesting fashion. In their model, different individuals have different conceptual spaces, representations of all the concepts and information that the individuals know about. Language maps areas of this conceptual space to individual words and sentences. Communication fails when the words you use map to different areas in your and the other person's conceptual space.

As a way of illustration, see this image. The two images are the conceptual spaces of two people, with the lines marking the areas that different words map to for the two. As they put it in the paper:

We can consider an example illustrated by Fig. 2: a buyer agent expresses the wish for finding an item belonging to a specific category, represented as an area on the diagram. It may very well be that the selling agent understands the query differently and therefore the items considered as candidates by the selling agent differs from the buyers intentions. Within traditional epistemological theory, this situation could be considered through set theory: by a symbol the agents refer to different sets of items. However, as many features (or quality dimensions) are continuous, it appears to be more natural to consider the situation within a continuous multidimensional space. Moreover, the exact features of an item may not be known but they need to be considered as a probability distribution.

One implication is that in business transactions there should be means for checking what is a meant by some expressions via an access to a broader context (cf. symbol grounding). Moreover, rather than relying solely on a standardized conceptual system, one could introduce mechanisms of meaning negotiation. Before two business agents get into negotiation about, for instance, the price of some commodity, they should first check if they agree on what they refer to by the expressions that are used in the negotiation. This concern is valid both for human and computerized systems, even though humans are usually capable of conducting meaning negotiations even when they are not aware of it.

The need for meaning negotiations becomes even more obvious when one considers qualities like being ‘interesting’ or ‘beautiful’. The concepts of interestingness and beauty are highly subjective and the criteria used may depend on very different underlying features. These kinds of concepts are, however, the basis of a vast number of daily business related transactions. For instance, when a person is considering a gift for her spouse, she is usually, at least implicitly, considering the taste of her spouse, for which she has a mental model formed over time.

Personally, I've noticed that there's a great many deal of disagreements that vanish completely after I and the other person finally get our mappings in synch and realize that we agreed all along. After running into this problem several times with some close friends, I've come to the conclusion that I need to apply the heuristic "if somebody seems to be saying something that just doesn't make sense, find out if he's really saying what you think he's saying" more often.