The title is a touch misleading, since the book comprises sections about diagrams and artificial intelligence, but the connection between them which Nakatsu proposes in the final chapter is little more than an idea. Chapters 1-4 and 9 are about diagrams, while 5-8 describe artificial intelligence, a technology that can be used to manage business processes and high-frequency trading on stock markets. Nakatsu’s finale is an argument that diagrams would constitute a more effective interface between users and these "expert systems" than existing methods of giving commands with such software. (His faculty page says he designed Expert-Strategy, a software that provides a graphical user interface to an expert system’s knowledge base.)
The main value to us of his essay lies in his earlier observations about mental models and why diagrams are effective for reasoning when compared with sentential statements:
In an sentential representation we form system descriptions by employing the sentences of a language. A diagram, by contrast, is a type of information graphic that "preserves explicitly the information about topological and geometric relations among the components of the problem." [Larkin/Simon] In other words, an information graphic indexes information by location on a plane. ... For example, a graphical hierarchy can help humans sort through information much more efficiently and understand how the objects of a domain are classified much more rapidly than a verbal description, which must be processed sequentially. (page 57)This comes close to my own description of why the Great Stemma is likely to have been devised and what advantages it offered to its users. Oddly, his discussion of hierarchic diagrams only briefly mentions their use for classifying biological species and is silent about their first use to represent human pedigrees.
In his final chapter, Nakatsu briefly alludes to an earlier paper he published on the effectiveness of diagrams when compared them to an alphabetic tabulation of the important data which could be fed into an expert system.
When asked to comment on possible weaknesses of the hierarchic system, participants were able to come up with a few responses. The most popular response was that the hierarchic interface was more complex and that more training would be required to learn and use it (28 participants). It is interesting that a few participants (6 individuals) suggested that the hierarchic interface might be harmful in terms of biasing the user toward a certain way of using the system. That is to say, the user of a flat system is more actively engaged in trying to understand the relationships in the variables, whereas the user of the hierarchic system is more likely to passively accept what the system teIls him or her to do. However, overall, it was c1ear that the hierarchic interface was highly preferable to the flat one. (page 315)This relates to a point I have referred to myself: diagrams offer a more powerful tool to someone seeking to convince others, because diagrams are more difficult to test analytically. The doctrines in the Great Stemma appear less plausible when explained verbally, but somehow more logical in a neat visualization. Audiences are suspicious, because they are generally aware that diagrams tend to lead to "passive" acceptance and can be inimical to a critical response. It may simply be that we are educated to question what we are told, but we are not trained to question the veracity of what we see.
Nakatsu, Robbie T. Diagrammatic reasoning in AI. Wiley, 2010.
-------- ‘Explanatory Power of Intelligent Systems’. In Intelligent Decision-making Support Systems, 123–143. Decision Engineering. Springer London, 2006. http://link.springer.com/chapter/10.1007/1-84628-231-4_7.
No comments :
Post a Comment