Conceptual Models – Johnson, Henderson

Friday, January 20th, 2006

Johnson, J. and Henderson, A. 2002. Conceptual models: begin by designing what to design. interactions 9, 1 (Jan. 2002), 25-32.

designing what the system is to users – is this also the role of task analysis, a design message, all that stuff about making sure you’re on the same page about why a system matters? i guess it’s making that concrete? but it could come from the task analysis – thinking of how users will interact with the system suggests it’s underlying structure.
thinking at the system level, not the task level.

choosing a clear model for the system so that users aren’t confused.
weighing tradeoffs between simplicity and power.
metaphors, analogies, concepts, relationships, and mappings.
not the user’s mental model, but a tool for designers that will hopefully make the formulation of mental models easier and less prone to error.
idealized view of the system, the ontological structure, the mechanism by whish users accomplish tasks.

the argument strikes me as a little bit captain obvious, but it does make me think about things like: what are the actual conceptual pieces that i am trying to build? right now my idea is pretty abstract, with just a vague goal of bringing about a stronger sense of location, landscape, interconnection… but what does that mean? if i shy away from the folder, what am i offering instead? if i’m talking about a halo, what does that really mean? how will it be accessed, and what sort of pervasive metaphor will tie the separate pieces together? the idea seems to rely on the transformation of these dimensions into entities of their own.
is that right?
how does that work?

*more response needed to end of paper*


Thursday, December 29th, 2005


Krishnan, A. and Jones, S. 2005. TimeSpace: activity-based temporal visualisation of personal information spaces. Personal Ubiquitous Comput. 9, 1 (Jan. 2005), 46-65.
Available online here


I got this paper months ago, but only just got around to reading it. It’s a good thing. It’s quite relevant to what I’m doing and offers a lot of good points to consider. It gives a very good overview of other work and literature on personal information spaces, and talks about the different ways that information can be organized and visualized. It’s main points are that people naturally think of information in terms of activities and that interactions have clear temporal characteristics that can be helpful for organization and recall. The work of Abowd and Mynatt on the idea of ‘everyday computing’ – “the scaling of ubiquitous computing with respect to time” is drawn on to support these points. Existing systems are said to be either activity-based OR temporal, not both, and they also tend to focus on specific application areas rather than personal information as a whole. The system that is presented in the paper seeks to address these issues by creating an activity-based, temporally filtered information space that populates itself as users interact with their files. They harp a lot on the fact that they are augmenting the existing hierarchical file system, not replacing it, and they see that as a major strength. All in all, the work seems sound to me, the paper reads easily, and though their prototype is in a pretty early stage, users found it to be quite helpful and enjoyed the ability to see their workflow over time.

My reaction

A lot of the points that are raised here really support my working hypotheses. I have thought a lot about time-based filtering of files (the whole idea of capturing metadata about “what else I was looking at while I was looking at this” is based on it), and it’s good to see other work that supports that idea. They also talk a lot about the limitations of hierarchical filing systems and do a good job of articulating a call for alternatives. The survey of other options is helpful, and they make a fairly compelling case that time and activity are good central filters. They do not, however, convince me that other filters and visualizations aren’t needed. They don’t really try to convince me of this, actually, focusing on the potential for their tool as a productive augmentation to existing hierarchical filing systems. The trouble is, existing hierarchical filing systems don’t offer means of visualizing along other axes – by personal relationship, by tags that cut across folder boundaries, etc., and their main arguments for keeping the hierarchical system intact are that a) people are familiar with it, and b)it is easier for people to find infrequently used or very specifically organized data within it. Fair enough, but that doesn’t convince me that another representation couldn’t address these issues better, and the result is that users who want to benefit from both of these systems need to do organizational upkeep in both. Timespace does a good job of making a lot of the metadata collection automatic, but there is still tweaking to be done, and so users will have to do that as well as maintain their current filing habits, habits which, as the authors concede, might not be all that effective. Other tools could be made to fill in the remaining gaps, but that would just make the burden of use higher.
The argument that is starting to take shape in my head is something like this: a lot of work has been done to show the need for new ways of organizing and visualizing personal information, and many effective tools have been developed that show a great deal of promise in addressing these needs (TimeSpace, Haystack, Relo, Data Mountain, Snarf, delicious). The majority of these tools, however, focus on only one potential axis of organization and attempt to make that the central axis that the user sees, despite the fact that data suggests that different axes are useful in different contexts, and it therefore seems worthwhile to explore the potential of developing a way of combining the axes by way of a shared central metaphor. I think that the key to this new metaphor is the realization that all of the axes involve thinking of categorization as a task of defining relationships rather than destinations. The RDF-based approach of Haystack and Relo show great promise in this regard, but the interface still relies on a hierarchical layout and the cognitive burden of assigning the metadata is still too high for the average user. I think there’s a better way.
dun, dun, duhhhhhh!!

the key is that it has to do with picking out the types of relationships that are meaningful, the axes along which information is connected, and then visualizing those as a means of connecting the other methods. that way every piece of information, everywhere, can be used to find it’s friends without having to open up seventeen different filtering tools.
this is what i’m calling halo for now, because i see it as an aura of relatedness that exists around everything.
so is the key difference that i’m getting at the idea that the locus of categorization should be accessible from each unit, at any time?


They have two pretty helpful lists, one of organizational methods and one of visualization methods as enumerated by Schneiderman. Both of these refer to the ‘network’ model in a way that seems strange to me. Networks are described as methods that illustrate the relationships between files. That doesn’t seem very descriptive to me. Isn’t the whole problem How those relationships are illustrated? It seems to me that some of the other methods fall under the ‘network’ category, but maybe I’m missing the point?
They say that activity-based organization frameworks are missing from Schneiderman’s taxonomy, but aren’t they pretty much covered under hierarchical?
I never really get how you add just one file to a project…


“Users’ personal information spaces… are fluid entities, evolving over time, aned supporting multiple user activities that may require different perspectives of the same underlying information structure” (46).
“Considering the characteristics of personal information spaces and user information behaviour, users may benefit from overviews of their infomation collection that relate their personal documents, activities and interests together over time and provide a context for their past, present, and future work”(47).
“Both hierarchical and network models primarily focus on the organisation of information per se, rather than the user’s interaction with that information” (49).
lots of justification for temporal organization (49-51)
“Studies of user information behaviour have revealed many difficulties that users face in organising and managing their personal information collections – difficulties that arel ikely to be exacerbated as those collections grow in size and complexity. As the bounds of the collections stretch over multiple devices and across time – as in the everyday computing paradigm – the need for appropriate user support becomes increasingly pressing. The widely adopted hierarchical model of information organisation has many limitations, not the least of which is that it requires substantial user effors to maintain a herarchy so that it facilitates the higher-level tasts that the user is addressing” (52).
“Given that personal information spaces are almost always too large to be fully displayed on the user’s screen, interactive visualisation techniques can be applied to present overviews, allow navigation within the space, enable the user to focus on specific segments, and access details of information items when necessary” (52).