Improving online style-guides and guidelines
Mikael Ericsson
Department of Computer and Information Science
Linköpings universitet
581 83 Linköping
Sweden
miker@ida.liu.se http://www.ida.liu.se/~miker/
ABSTRACT
Empirical studies (quantitative and qualitative) were performed in order to assess the perception of different forms of online presentation of formalised design knowledge (e.g. guidelines). The results show that guidelines should be imperativerather than declarative. Qualitative data also suggests that examplesshould be used for guidance.
KEYNOTE ADDRESSES
PANELS
KEYWORDS:
Design support, guidelines, online support
INTRODUCTION
Making guidelines and style guides useful in professional work is a difficult task, as evidenced by scientific studies and practitioner comments: guidelines are often claimed to be too general, too specific, hard to access, hard to locate, etc. (Chapanis, 1990; Tetzlaff & Schwartz, 1991) Taking the knowledge online, inserting it into the designers work environment is one approach trying to solve some problems. There are numerous prototype systems in the form of hypertext guidelines, example-databases and intelligent commenting agents. Those make the knowledge available and accessible in the design situation and show the feasibility. There is also evidence that designers prefer online over paper based guidelines (Fox, 1992). However, it is still unclear what form should be used for presentation. This paper presents a study of different linguistic forms for design knowledge presentation.
EMPIRICAL STUDIES -- RESULTS
As part of an investigation of different behaviour of commenting agents delivering guideline knowledge in the form of comments (Ericsson, 1996), we assessed the effects from different linguistic variations.
Sixteen subjects were observed using a simulated (Wizard-of-Oz) commenting agent in a design support system. Different commenting behaviour was tested, and the overall usefulness evaluated. Two linguistic variants (mood) were used--a declarative (pointing out a flaw) and an imperative (suggesting how to remedy a flaw) --with the intent to keep their semantic content otherwise similar. The interaction was logged and recorded on video, and the subjects rated the agent with respect to usefulness, understandability, system competence, disturbance and perceived stress. Perceived mental workload was measured using RTLX. Questionnaire questions relevant for guideline form addressed the comprehensibility of the comments, the perception of the competence of the tool and the participant's ability to find objects referred to in a comment.
The participants performed a one-hour design task, and were then asked to rate and to describe their impression of the system and the design support. The participants were
IMAGE imgs/HCI98CC23.gif
ORGANISATION OVERVIEWS
INDUSTRY DAY
DEMOS
VIDEOS
DOCTORAL CONSORTIUM
POSTERS
--22--
HCI'98 Conference Companion
then shown three short video clips from their design work, and asked to perform the same kind of rating for this specific situation.
We found three significant effects from different mood settings. Participants using the declarative system rated comments as more difficult to understand than those using the imperative system in general (6.3 vs. 2.8, F(1,10)=14.04, p<0.05) and on average in the specific situations (5.7 vs. 2.6), F(1,24)=10.1, p<0.01). Those using the imperative system rated the system competency lower than the declarative group (4.7 vs. 5.8, F(1,10)=4.0, p<0.10). Participants in the declarative mood reported a higher MWL than those in the imperative group (52 vs. 37.6, F(1,35)=16.7, p>0.01).
As part of a study on the use of and need for (requirement) support for formalised design knowledge, we interviewed 8 professional systems developers. The semi- structured, open-ended interviews contained questions about guideline availability, use, demand and form. The interview answers and comments show that users prefer accessing the "essence" of the design knowledge first, without complete rationale. However, they would like to have the possibility to immediately access examplesthat illustrates the guideline (positively or negatively).
CONCLUSIONS AND DISCUSSION
Guidelines presented online in a design situation should be imperativerather than declarative. Imperative comments result in lower MWL and are rated as easier to understand. The fact that an imperative system is considered less competent is similar to others, showing that systems, which are easier to understand, are considered less competent (cf. Wærn and Ramberg, 1995).
The preference for imperative forms implies that comments must be "situated" in the particular design moment. A comment phrased in a specific form (imperative) may be easier to relate to the particular design moment than a comment phrased in a general form (declarative), even if the degree of contextualisation is rather low.
ACKNOWLEDGEMENTS
This work was conducted in collaboration with Magnus Baurén, Prof. Yvonne Wærn and Jonas Löwgren. Thanks to the participants of our studies for their time and efforts. This work was financially supported by the Swedish Research Council for Social Sciences and the Humanities (HSFR).
REFERENCES
Chapanis, A. (1990) Specifying human computer interface requirements. Behaviour
and Information Technology, 9(6):479-492.
Ericsson, M. (1996) Commenting tools as design support - a Wizard-of-Oz study.
Licentiate thesis, Linköping Studies in Science and Technology. No. 576.
Linköpings universitet, Sweden. http://www.ida.liu.se/~miker/research/ Fox, J. A. (1992) The effects of using a hypertext tool for selecting design guidelines.
In Proc. of HFS'92, pp. 428-432.
Tetzlaff, L. & Schwartz, D.R. (1991). The use of guidelines in interface design. In
Proc. ofCHI'91, pages 329-333.
Wærn, Y. & and Ramberg, R. (1995) People's perception of human and computer
advice. Computers in Human Behavior.
