Saturday, September 24, 2005
behavior verses preferences
Is the future of user centered design watching people, or probing people? When selecting an approach, a key question is: how reliable is behavior, and how reliable is preference?
Traditional usability engineering admonishes researchers to "watch what people do; don't listen to what they say." Ethnography also can stress the need to watch what people do in naturalistic settings, and to discount verbal introspection from users. It can deviate from straight behaviorism by asking users questions about why they do what they do -- it tries to overcome the black box approach. Emotional design research borrows far more from traditional market research by exploring preferences instead of behavior. It may be wary of the validity of verbal declarations of preference, but it nonetheless seeks to develop a model of the user's inner mind, and his or her wants. Many designers of new products argue that past behavior with mature products is a poor guide to understanding what people really want from future products.
I think the variability of people is well illustrated by how they behave on the job. I just finished listening to a BBC Radio 4 program on psychometric job testing. When I lived in the UK, I always thought the extensive use of psychometric testing to screen job candidates rather odd. If you visit a UK bookshop, you will find shelves of books on how to take, and "pass," psychometric tests. The whole exercise degenerates into a guessing game of finding and providing the socially acceptable answer. The flaw in psychometric job testing is the notion that one's stated personal preferences somehow reflect one's future job behavior. The first problem is that what people say they prefer may not be what really prefer. This is a well known problem with any probe. The second problem is that people's preferences may have little bearing on people's behavior. Whether someone says they prefer mercy or justice does not predict if they will fire an incompetent employee. There are many other factors that come into play (perhaps feelings toward potential lawsuits.) And it also confuses the idea of global personal preferences with role behavior. We have all seen examples of the kind grandfather/ruthless executive.
Probes may work in the absence of social influence and role associations. Some products have fewer social and role associations than others, but all products have some (unless intended for hermits.) People may be less able to manipulate their nonverbal responses than verbal responses to probes, but I would doubt their responses are entirely involuntary and thus "objective."
Ethnography is potentially powerful for accounting for the influence of social factors. But by subjecting observation to context, it is not possible to develop a global perspective on people, finding their core life motivations and preferences. Only traditional usability can pretend to offer a global perspective on users. It does this by ignoring preferences, and looking only at a narrow range of user behaviors that are consistent across contexts.
Traditional usability engineering admonishes researchers to "watch what people do; don't listen to what they say." Ethnography also can stress the need to watch what people do in naturalistic settings, and to discount verbal introspection from users. It can deviate from straight behaviorism by asking users questions about why they do what they do -- it tries to overcome the black box approach. Emotional design research borrows far more from traditional market research by exploring preferences instead of behavior. It may be wary of the validity of verbal declarations of preference, but it nonetheless seeks to develop a model of the user's inner mind, and his or her wants. Many designers of new products argue that past behavior with mature products is a poor guide to understanding what people really want from future products.
I think the variability of people is well illustrated by how they behave on the job. I just finished listening to a BBC Radio 4 program on psychometric job testing. When I lived in the UK, I always thought the extensive use of psychometric testing to screen job candidates rather odd. If you visit a UK bookshop, you will find shelves of books on how to take, and "pass," psychometric tests. The whole exercise degenerates into a guessing game of finding and providing the socially acceptable answer. The flaw in psychometric job testing is the notion that one's stated personal preferences somehow reflect one's future job behavior. The first problem is that what people say they prefer may not be what really prefer. This is a well known problem with any probe. The second problem is that people's preferences may have little bearing on people's behavior. Whether someone says they prefer mercy or justice does not predict if they will fire an incompetent employee. There are many other factors that come into play (perhaps feelings toward potential lawsuits.) And it also confuses the idea of global personal preferences with role behavior. We have all seen examples of the kind grandfather/ruthless executive.
Probes may work in the absence of social influence and role associations. Some products have fewer social and role associations than others, but all products have some (unless intended for hermits.) People may be less able to manipulate their nonverbal responses than verbal responses to probes, but I would doubt their responses are entirely involuntary and thus "objective."
Ethnography is potentially powerful for accounting for the influence of social factors. But by subjecting observation to context, it is not possible to develop a global perspective on people, finding their core life motivations and preferences. Only traditional usability can pretend to offer a global perspective on users. It does this by ignoring preferences, and looking only at a narrow range of user behaviors that are consistent across contexts.
Link to this article. posted by Michael Andrews @ 9:25 PM 
