2.4.1 From Heidegger to knowledge technologies
Because each transformation from one ‘knowledge state’ to another (Figure 2) is an act of interpretation, there is no such thing as objective knowledge representation, or indeed objective classification or codification of any sort (in software or any other medium): there is always a viewpoint. This leads to the view that information and communication systems cannot be thought of as neutral; in their formal structures and operations they embody the goals and perspectives of their developers.
Similarly, attempts to make tacit knowledge explicit run a number of unpredictable risks. Think of the fabled centipede that became paralysed when it tried to think about how it walked. We can perhaps relate a little more closely to the example of sociology students who report that learning how people use non-verbal cues to take turns in conversations has a debilitating impact on their own social skills! These examples point to the impact of asking people to reconceptualise their activities using abstract, descriptive ideas. They illustrate the codification process of moving from tacit to explicit knowledge. The analytical process of studying complex behaviour provides researchers with the vocabulary they need to discuss it, but this symbolic representation is qualitatively different from the tacit, embodied skills being described.
This subtle change in the quality of information is one of the reasons why it is so hard to build expert systems for problems that are not already very well understood, with well-defined boundaries for the nature and numbers of variables that can arise, and the methods that can be used to cope with them. The most successful expert systems ‘simply’ (they are still complex) manage large networks of interdependences between variables which are cognitively too complex for humans to track easily.
A more down-to-earth manifestation of the formalisation problem is that people often find it hard to fill in forms with predefined checkboxes and questions. The form serves the specific purpose of structuring information from the messy world into predefined, abstract categories. Mismatches arise when we do not think about our work according to those categories, when questions are asked at the wrong level of detail, or when we are not asked for information which we deem essential to a proper understanding of what we do. We end up ‘shoe-horning’ information about an embodied activity into decontextualised, symbolic representations ideal for computational analysis and recombination with other data sources. Naturally, we wonder about the value of the data and how our form-based answers will feed into subsequent decision making.