Understanding complex descriptions

Abstract

User-system interaction is starting to receive the attention it deserves. A picture is starting to emerge from work in cognitive ergonomics of the requirements for user interfaces, for example for text editors (Card et al., 1983), command languages (Barnard et al., 1981) and programming language syntax (Green, 1980). In a short sketch of the research opportunities that exist, Shneiderman (1982) lists several similar issues, such as menu selection, on-line assistance, etc. He also draws attention to issues such as overcoming anxiety and fear of computer usage. I shall not comment on such social and organisational issues, but I want to suggest that attention to ergonomic issues of the first type neglects a range of important human factors in system use. These operate at a conceptual level beneath the external (e.g. graphical or pretty-printed) characteristics of the interface. The problem I want to address relates to these. It is this: how can an information system be designed so that the complex descriptions it supports and which are manipulated by the users can be rendered into forms that are harmonious with the users' mental models?

In the first part of the paper I shall discuss a conceptual architecture for information systems which emphasises the understandability of the descriptions they embody. The architecture consists of an internal description which may be manifested by applying two classes of transformation, forming separate conceptual and external interfaces. In connection with the conceptual interface, I discuss the relevance of current work in cognitive science, especially text linguistics. I shall ignore the external interface, as this has received the most attention in the cognitive ergonomics literature.

In the second part of the paper, I shall concentrate on a class of complex descriptions of particular relevance to software engineering, specifications. A specification is usually thought of as a text and a specification language as a language. A specification, however, is simply a description of a reality that does not exist yet and a programming support environment is, amongst other things, an information system for manipulating such descriptions. Thus in addition to the effects of syntactic and physical factors (e.g. layout) on the understandability of specification, we should also be examining the conceptual models required of the specifier by different specification language.