5.2.9 Speech-Act based office Modeling aPprOach (SAMPO)

SAMPO is an approach to development of office information systems based on a language/action view of the office (Hirschheim et al., 1995:199). An office is seen as a system of communicative action (Auramäki et al., 1988:126), and an information system is considered a technical implementation of a social system.

A main reason for including SAMPO in the discussion is that it has been assessed as constructivistic, both by Krogstie (1995:143) and Hirschheim et al. (1995:199). Another reason is a fundamentally different set of model concepts, claimed to be more appropriate for modeling office work than traditional fact-based approaches (Auramäki et al., 1988:126).

The relevance of SAMPO to enterprise modeling is that office work is a significant part of most enterprises. The purpose of modeling in SAMPO is to understand the domain (office analysis) before constructing an office information system (i.e., it can be associated with the ISE strand). Type of work is office work, although the approach probably can be used with any type of work involving social actors.

In SAMPO, two basic domains are modeled: The entity domain (ED) and the action domain (AD). ED consists of static entities, such as items of transactions (artifacts) and agents. AD includes dynamic entities, called acts, that either can be instrumental acts or speech acts. Instrumental acts change the ED. Speech acts are symbolic deeds, being utterances with associated meaning. Speech acts are the basic units of communication, and can have a complex structure. Auramäki et al. (1988:128ff) describe the structure of speech acts in more detail.

Models in SAMPO are expressed using a number of more or less informal tables and graphs. The two main types of models are:

In addition to the two types of graphs, there are a number of tables describing purposes, effects, media, etc. of discourses and individual speech acts. The modeling process of SAMPO is not discussed in a structured way in (Auramäki et al., 1988), but the following activities can be identified (and iteration is stressed):
  1. Characterization of discourse:
    1. Initial interviews with users in order to define discourse type.
    2. Identification of main speech acts of the discourse.
    3. Naming of discourse entities, e.g., agents.
    4. Further clarification of speech acts and instrumental acts in tables (purpose, frequency, timing, etc.).
    5. Construction of activity tables (functional units of the organization).
    6. Construction of predicate tables to control the flow of the discourse.
  2. Analysis of discourse:
    1. Analyze coherency of discourse by creating a conversation graph.
    2. Analyze the completeness of discourse (finding conditions for termination of processes through transformation of conversation graph into reachability tree).
    3. Describe coordination of commitments (modeling how speech acts influence each other).
    4. Analyze discourse ambiguity (concerning performance, commitments, and roles).
The foundation of SAMPO is claimed to be constructivistic. This claim stems from SAMPO taking a language/action perspective on modeling, with explicit support for modeling of the UoP in terms of communicative acts. Data modeling is considered to describe the rules which govern the uses of signs and symbols in organizational behaviors and thereby attach specific meanings to the organizational vocabulary (Hirschheim, 1995:189). The constructivist flavor of this is that rules are socially constructed and maintained through their very application, and modeling according to SAMPO may be considered modeling of a process of social construction of reality. However, SAMPO does not aid the enterprise modeler much in the modeling process, as the proposed modeling process is very general and focuses on representation (as for several other approaches, e.g., OMT). The sense-making process is not discussed in any significant detail in the surveyed literature. Also, the modeling language (both the symbols and the underlying concepts) does not lend itself easily to application by inexperienced modelers, as also recognized by Hirschheim et al. (1995:232). Complicated modeling languages restricts participation in the construction process.