I’ve been writing and thinking about about mental models a lot this week as I sharpen up my literature view for Friday’s deadline. In HCI and cognitive science more generally, knowledge is understood to be highly organised and it seems reasonable to suggest there is some organising principle at work in how we approach tasks in daily life. How exactly this organisation is configured and under what circumstances it is engaged when confronted with digital interfaces is a matter of considerable debate however. One major influence has been the idea that knowledge is arranged as a network in a kind of topology of thought. An alternative idea is that knowledge is organised in a series of schemata. A schema is a set of mental instructions developed through experience that helps us to do things in everyday life, such as fill a car with petrol. The idea is that various schema are in operation when we set out to do things in the world. This concept breaks down somewhat when confronted with flexibility and improvisation. How can a fixed set of instructions let me fill up any car? Surely any single schema would be so context specific as to be useless in different circumstances?
There is a way around this problem however if you think about mental models as less a set of instructions than ways of organising experience. Don Norman showed that it’s perfectly possible to act efficiently with technology whilst having a wildly inaccurate mental model of how it works. I am able to drive my car to the station while having only a very vague idea of how the engine and gears work, and can happily send and receive emails without knowing much about my SMTP server configuration or POP3 protocol. Norman reckons this means I have a good functional model that lets me do these things efficiently, but a poor structural model of how those technologies actually work. In design this is called satisficing, a cognitive heuristic theorised by Herbert Simon, originally used as a way of setting acceptability thresholds for decision making.
In design research mental models have been used for years as a way of gathering system requirements. The idea is that if you can obtain a detailed picture of how people go about booking a theatre ticket, or brushing their teeth you can design products and systems that better match how people behave and you’ll sell more stuff. In some ways it has proved highly successful in design but I suspect this is in spite of, rather than because of, mental modelling. In other words, the process of structuring research draws on well established techniques in ethnography and is so effectively executed (when done well) that any discrepancy between participant and designer is compromised by methodological rigour. It assumes that there is a single correct answer to how people do things and the designs that emerge are taken to have been proved useful. I suspect there are any number of failed design projects where the research has suggested a way of doing things that has proved totally incorrect. There’s not much room in design research for mental models that change over time, that take account of ambiguity, or that incorporate hybrid behaviours and activities such as texting while driving, or cooking while dancing.
It gets even more difficult when thinking about how people interact with digital systems. Since mental models are used to design those systems, reversing the process to try and work out what people think about Facebook for example or how they understand their own browsing activity seems to call for different methods. The focus of what I’m up to is developing creative ways of getting at that knowledge, ways that leverage design methods and ways that acknowledge how different people are and how dependent on context and setting any kind of qualitative research is. Yvonne Rogers has proposed that meaningful research is conducted ‘in the wild’ meaning in natural settings and that design offers a way of doing this.