sufficient relevance

From: Dan Sperber (dan@sperber.com)
Date: Sun May 19 2002 - 20:50:55 GMT

  • Next message: J L Speranza: "From Sperber and Wilson to Hoey and Wodak"

    Stefan Malmberg writes:

    I wonder if it is possible to achieve a closer definition of the terms "adequate
    range of contextual effects" and "minimum justifiable processing efforts" than has
    hitherto been provided by RT.

    What we want is not so much a definition as a criterion. Here is one way of developing one. Remember that all potential inputs are competing for an individual's attention and more generally for her cognitive resources at any one time. Ideally, attentional resources should be allocated only the input  with the greatest expected relevance at the time (I am assuming, at least for the sake of exposition, that we can attend only to one input at a time - otherwise, the criterion might be a bit more complex, but not radically different). Once an input has begun being processed, it may very rapidly loose it status as the one with the greatest expected relevance in the ongoing competition - anyhow, the competition remains open.

    Now, how is expected relevance assessed? Our take has always been that it is not computed. Rather the cognitive system is so organised that it tends to give a competitive advantage to input with the greatest potential relevance. In other terms the competition is biased by a variety of mechanisms shaped by human evolution, by the person's education and history, by the saliencies of the immediate situation that, either in a bottom up or a top-down manner push or pull potentially more relevant inputs towards attentive processing. Take this partial ordering of potential inputs to represent the individual's (of course sub-personal) expectations of relevance. For an external point of view, one can always judge the adequacy of these expectations, and decide that some unattended stimulus would have been more relevant than the attended one. Well in that case, the cognitive system functioned sub-optimally (as it often does): it was aiming at the most potentially relevant stimulus. So "sufficient relevance to receive attention" is determined by all the factors that affect the internal competition. "Sufficient attention to deserve attention" is determined by objective counterfactual considerations: which input, had it been processed, would have proved the most relevant?

    Now, to effect and effort. Again, the general take has been to assume that there is no computation of absolute values, but only comparisons where possible and of use. A competing potential input is helped by a comparatively high level of expected effect and hindered by a comparatively high level of expected effort. This may cause a competing input to win or loose, all depending on comparisons with the competitors. How all this? A matter for empirical investigations, not definitions.

    All this so far was about the First, Cognitive Principle. Regarding the Second, Communicative Principle, the presumption of optimal relevance in its last version mentions only an adequate level of relevance, and does not talk about effect and effort. In discussing specific cases, considerations of (sufficient or adequate, or justified or unjustified) effect and effort may arise unproblematically, as for instance when a remark sufficiently relevant on one interpretation to deserve processing involves, to reach this interpretation, a blatantly unnecessary effort. Given the second principle, this raises the level of expected effect and therefore of overall expected relevance of the stimulus, allowing it to stay in the competition a bit longer and get its interpretation enriched.

    Or something of the sort. Yes, there is plenty more to do.


    Cheers, Dan


    This archive was generated by hypermail 2b29 : Sun May 19 2002 - 20:55:15 GMT