ADR-030: design-rationale joins the topics

Context

The owner is considering extending this record with argument-development schemes: questions, claims and objections, in the manner of IBIS. Systems of that kind have been built since 1970, and the literature on them is mostly a literature of why they went unused. The owner asked for that literature to be held here, as the cautionary evidence for the extension:

• Kunz and Rittel’s 1970 working paper, which introduced IBIS;
• gIBIS (Conklin and Begeman 1988) and QOC (MacLean et al. 1991), the two best-known notations and tools;
• Conklin and Burgess Yakemovic (1991), a field trial where capture was sustained;
• Grudin (1988, 1994) on why groupware fails, and in particular the disparity between who does the work and who gets the benefit;
• Buckingham Shum and Hammond (1994) and Lee (1997) on what design rationale costs and when it repays the cost;
• Shipman and Marshall (1999) on why users will not formalize.

No topic held them. The nearest words each say something else:

• society-and-governance is law, politics and institutions;
• epistemology is knowledge and justification, not tools for recording reasons;
• cognition is the mind as information processing, and covers the overhead argument only from the side of the person paying it.

The anthology’s human-ai-interaction is “how people work with a model through an interface”. These works predate the models and are about people recording their reasons for each other, so no anthology topic can hold them either.

Decision

Add design-rationale: recording the reasoning behind a design or a decision as it is made, and why systems for doing so are or are not used. That includes:

• argumentation notations and tools: IBIS, gIBIS, QOC and their successors;
• design space analysis and argumentation-based rationale;
• the cost and benefit of capture, and who bears each;
• premature formalization and the cognitive overhead of explicit structure;
• the groupware adoption problems that the design-rationale literature cites as its own.

All nine works in the batch take design-rationale as their only tag. Grudin’s two papers are about groupware in general, not rationale capture. They are filed under this word because the rationale literature takes their cost-and-benefit argument as its central explanation, and that is the question the owner is asking of them.

Alternatives considered

• society-and-governance, epistemology or cognition. Each is the nearest wrong word for some of the batch and the wrong word for the rest.
• human-computer-interaction, or cscw. Broader, and it would hold Grudin and Shipman and Marshall without the stretch noted above. But the owner’s question is rationale capture, and a word for all of HCI would invite filing under it whatever an interface paper happens to be about. Widen the word if the holdings outgrow it.
• Separate words for argumentation tools and for their failure modes. With nine works, one word is enough. The failure-mode literature is about the tools, and is not read apart from them.

Consequences

• A later argument-development scheme in this record, if the owner adopts one, has its prior art and its known failure modes under one tag page.
• Work on argumentation as logic or rhetoric, as opposed to its capture in a system, still goes under logic or philosophy-of-language.