A scientometric cognitive study of neural network research: Expert mental maps versus bibliometric maps

The method's utility for studying expert views in general is examined, and its strengths and weaknesses as a tool for improving validation studies of bibliometric maps involving subject experts are examined.

Abstract

This paper reports on a quantitative analytical methodology which deals with perceptions of scientific experts regarding the intellectual shape and contents ('cognitive structure') of their scientific domain. This study examines the method's utility for studying expert views in general, and, more specifically, its strengths and weaknesses as a tool for improving validation studies of bibliometric maps involving subject experts. The main premise is that expert views are based on their internal knowledge structures ('mental schemes') of which relevant features can be captured in quantitative data. This approach allows a rigorous and systematic way of studying mental schemes across subject experts. Spatial representations of their data ('mental maps') provide insight in properties underlying those knowledge structures. Data from different experts are reconciled to construct a 'common' mental map which displays a group view. This study includes a test to establish the validity of individual mental maps and common mental maps. The methodology is applied to the views of 14 researchers in the field of neural network research and related areas. Key-findings are: (i) mental maps can provide valid representations of expert mental schemes, (ii) experts sharing the same subject field are more likely to share views, (iii) expert judgements of bibliometric maps are affected by the structure of their own mental schemes, as well as (iv) by their views regarding the utility of those maps, and (v) common mental maps and a bibliometric co-word map based on the same set of items differ significantly, showing a resemblance on main features only.