Monday, 1 June 2009

Mental Model Literature Review - Part 3

This post continues my contextual/literature review and attempts to synthesise the older seminal work and newer work together.

Back in 1983 Gentner and Stevens presented a typology of methodologies to study mental models. They use the word 'eclectic' which, I deduce, may be where some current writers on the subject may have led to an imprecise use of the term mental model which I'll address in this post. The methodologies that Gentner and Stevens (1983, p2) present are:

  • protocol analysis

  • traditional cognitive psychology experiments

  • developmental studies

  • expert-novice studies

  • simulation of possible psychological models

  • comparison of the results of that simulation with what humans do

  • field observation

  • comparison across cultures

  • comparison across time within the same culture

  • designed field observation

Where Johnson-Laird's 1983 book focused upon the explication of mental models Gentner and Stevens' 1983 book focused upon application. Although these books are 25 years old they are still cited by current writers as the roots of mental model research. As interaction design emerged as a design discipline in its own right during the 1980s, cognitive scientific research informed its understanding of human behaviour. With the research came notable cognitive scientists such as Donald Norman. With the maturation of the discipline the tools for an interaction designer began to use personas in order for the design team to come to understand their target users. Personas are archetypes, derived from field observations, that embody the characteristics of each target user for the designers. The can then use the persona as a character mindset they can enter when they need to see the design from a user's point of view. Personas are not the same as a mental model

Over the last decade industry have identified different methodologies in order to understand the user in order to aid the design for them. One such methodology is affinity diagrams of user behaviours taken from ethnographic data. It is Indi Young from UX design company Adaptive Path who in recent years has been advocating affinity diagramming. It is a thorough methodology that interaction industry insiders find very rewarding. It is also problematic as this methodology is also referred to as mental models. In her 2008 book Mental Models: Aligning Design Strategy with Human Behaviour defines mental models as:

“Mental Models are simply affinity diagrams of behaviors made from ethnographic data gathered from audience representatives” (Young, 2008, pp2).

Not every interaction designer accepts this interpretation of a mental model. Nathan Kendrick of Designmap, in two replies to a posted question raised by Tom Dell'Aringa on the Interaction Design Association discussion board, disputes that this is nothing more than a task model. Dell'Aringa was asking for explication upon a mental model example in Young's book that uses an alignment diagram. Kendrick's response identifies with a cognitive scientific definition of the term, he says,

"This is a task model. Not mental model... mental models are the users (sic) understanding of how a system works. This particular method is to bridge user research and site IA. (…) the key difference meaning the task model has undergone analysis, synthesis, and ultimately organized by someone other than the user. A mental model is a user's messy, sometimes illogical understanding of how something operates." (Dell'Aringa, 2007)

In an podcast interview with Jared Spool, Indi Young defended her interpretation of the term against the confusion it causes within the discipline:

"It doesn't cause as much confusion as it causes debate. In HCI they teach a very narrow definition of mental model, they say it is something the user has in their head of how some particular piece of equipment they are working with works, or how some particular interaction they're going through works. In cognitive science, actually years ago, the definition of mental model had been expanded to mean lots of different concept models, so the idea that they are teaching just this one very narrow definition is a little dismaying. But if people can just open their mind to the idea there are lots of different ways to use that phrase mental model then it makes a lot of sense or the type of models that we're interpreting. We're building these mental models not out of something we're thinking, but we're letting the tasks build themselves, so basically all we're doing is interpreting this data that's already out there that already exists in other people's heads as to their mental model of how they get something done. So its not a mental model of how something works but its how they get something gets done. So maybe they are using lots of different tools and doing lots of different interactions, it is just a tiny extension of the old HCI definition. So I think that's totally acceptable. People may want to debate it - and say 'well gosh we are going to be confused' - but I don't think that, we're smart monkeys and we wont be confused." (Spool, 2008)

Gentner and Stevens (Gentner & Stevens, 1983, pp1-3) saw mental model research as fundamentally concerned with human’s knowledge of their world being examined and understood through cognitive processing. They identified the research, twenty-five years ago, as having three key dimensions that define and characterize it. These three dimensions inspect the nature of the modelling from the study of the methodology, the domain, and the theoretical approach. I began this post with a list of the various methodologies that they identified, and I will deal with methodology to place Young’s affinity modelling into a context. Before I do I wish to explore the other two dimensions, of which the domain contains the subject for the theoretical approach and selected methodology.

The domain they advance is the context for the experience of the phenomenon. The subject that a person focuses upon to make sense of happens in the real world. The person needs to predict what will happen based upon the affordances of the subject in question, past experiences of similar subjects, and inferences. The domain in which this happens either facilitates the successful dynamic construction of a model, or impedes it. Preece in her book Human Computer Interaction identifies (Preece et al, 1995, p136) the context-dependent nature of a model formed within a domain as a functional model. In her eyes the advantage of being context-dependent is that the mental model is easier to use. The domain can also be identified as the interactive system that the person is involved with and trying to make sense of. Preece in her later book Interaction Design: Beyond Human-Computer Interaction (Preece et al, 2002, p92) expands on learning from being context-dependent within a domain. She makes the important point that as the person develop their learning of how to use and work within the system, it doesn’t mean that the person actually understands HOW that system works. This is a crucial pivotal point that both Norman (1998) and Cooper (2007) address by different terms, they both identify making sense of how to use a system as the user’s mental model but the former calls understanding the system as the system model, whilst the latter calls this the implementation model.

The final dimension raised by Gentner and Stevens is that of the theoretical approach. Twenty–five years ago they identified a confluence of research from cognitive psychology and artificial intelligence that synergised the research of the day. This confluence has certainly developed as cognitive science grew from its multidisciplinary origins. Now recognised as a discipline in its own right it has continued its application across multidisciplinary boundaries, and has influenced interaction design, itself a discipline built upon synergy. Mental models have become a useful tool for the interaction designer, and it is still crucial to maintain a clear understanding of what is meant by the term. Young, with her definition of mental model has created a very useful tool for interaction designers but also has muddied the waters. In interview she is partly correct in citing that cognitive science over the years has expanded the “definition of mental model” but it is not HCI who are teaching “just this one very narrow definition”, it is defining affinity diagramming up using a recognised term. It is essentially a decision made at the publisher’s that is making the confusion, as Young’s methodology is proven, useful and informative, but it is as interaction designer Nathan Kendrick sums it up,

“A mental model is a user's messy, sometimes illogical understanding of how something operates." (Dell'Aringa, 2007)

In this post I wanted to synergise several points into a thread spanning twenty-five years of the literature and place Young’s very useful methodology into a clearer context. I feel that the methods she uses can be very useful in my future practical project to understand and evaluate aspects of the user’s behaviour and attitude towards defining a design model, but to understand a user’s mental model the cognitive scientists have tested methodologies that will be more useful.

Again I must reiterate that this post is not a thesis but a process. This post together with earlier and later posts will be re-read and re-evaluated in order to write a paper to support my PhD project. There are still additional theories I wish to explore such as Suchman’s situated actions, aspects of embodied cognition and Winograd, that I haven’t yet touched upon in these posts so far. There are several useful points I may add to a new post that begins this process, but I foresee the paper being the place where these influences will appear.


DELL'ARINGA, T. (2007) Mental Model question restated. 26 June. Available from: IxDA Discussion Archives [Accessed 13 May 2008].

SPOOL, J. (2008) SpoolCast: Reviewing Mental Models with Indi Young [online]. [Accessed 13 May 2008]. Available from World Wide Web: [Podcast]

See Design Model - statement for a full list of references this post cites beyond those indicated above.

No comments: