Environmental psychologists specializing in architectural psychology offer "user needs' assessments" and "post occupancy evaluations" to facilitate communication between users and experts. To compare the efficiency of building descriptions, building walkthroughs, regular plans, simulation, and direct, long-time exposition, evaluation has to be evaluated.

Computer visualizations and virtual realities grow more important, but studies on the effects of simulation techniques upon experts and users are rare. As a contribution to the field of architectural simulation, an expert - user comparison of CAD versus endoscopy/model simulations of a Vienna city project was realized in 1995. The Department for Spatial Simulation at the Vienna University of Technology provided diaslides of the planned city development at Aspern showing a) CAD and b) endoscopy photos of small-scale polystyrol models. In an experimental design, they were presented uncommented as images of "PROJECT A" versus "PROJECT B" to student groups of architects and non-architects at Vienna and Salzburg (n= 95) and assessed by semantic differentials. Two contradictory hypotheses were tested: 1. The "selective framing hypothesis" (SFH) as the subjective theory of planners, postulating different judgement effects (measured by item means of the semantic differential) through selective attention of the planners versus material- and context-bound perception of the untrained users. 2. The "general framing hypothesis" (GFH) postulates typical framing and distraction effects of all simulation techniques affecting experts as well as non-experts.

The experiment showed that -counter-intuitive to expert opinions- framing and distraction were prominent both for experts and lay people (= GFH). A position effect (assessment interaction of CAD and endoscopy) was present with experts and non-experts, too. With empirical evidence for "the medium is the message", a more cautious attitude has to be adopted towards simulation products as powerful framing (i.e. perception- and opinion-shaping) devices.

The possibilities of computer simulation also extend to issues inadequately covered by normative analysis and in particular to dynamic aspects of design such as human movement and circulation. The paper reports on a framework for addressing two related problems, (a) the simulation of fire escape from buildings and (b) the simulation of human movement on stairs. In both cases we propose that current evaluation techniques and the underlying design norms are too abstract to offer a measure of design success, as testified by the number of fatal accidents in fires and on stairs. In addition, fire escape and stair climbing are characterized by great variability with respect to both the form of the possible designs and the profiles of potential users. This suggests that testing prototypical forms by typical users and publishing the results as new, improved norms is not a realistic proposition for ensuring a global solution. Instead, we should test every design individually, within its own context. The development of an affordable, readily available system for the analysis and evaluation of aspects such as fire escape and stair safety can be based on the combination of the technologies of virtual reality and motion capture. Testing of a design by a number of test people in an immersion space provides not only intuitive evaluations by actual users but also quantitative data on the cognitive and proprioceptive behaviour of the test people. These data can be compiled into profiles of virtual humans for further testing of the same or related designs.

1. How do architects currently use CAD systems to produce drawings?

2. What are the effects of current CAD usage on product and performance?

3. What are the possible causes of current CAD usage?

4. What are the capabilities of the CAD medium and how can they be used efficiently?

The above four questions were addressed through the qualitative, quantitative, and cognitive analysis of data collected during an ethnographic study of architects working in their natural environment. The qualitative and quantitative analysis revealed that users missed many opportunities to use strategies that delegated iteration to the computer. The cognitive analysis revealed that missed opportunities to use such delegation strategies caused an increase in execution time, and an increase in errors many of which went undetected leading to the production of inaccurate drawings. These analyses pointed to plausible cognitive and contextual explanations for the inefficient use of CAD systems, and to a framework to identify and teach efficient CAD strategies. The above results were found to be neither unique to the CAD domain, nor to the office where the data were collected. The generality of these results motivated the identification of seven claims towards a general theory to explain and identify efficient strategies for a wide range of devices. This thesis contributes to the field of architecture by providing a detailed analysis of real-world CAD usage, and an approach to improve the performance of CAD users. The thesis also contributes to the field of human-computer interaction by demonstrating the generality of these results and by laying the framework for a general theory of efficient strategies which could be used to improve the performance of users of current and future computer applications.