| looking at email Readings:
How do you think email archives can be used to create personal histories? Do you think the significance of particular emails is cmoputationally accessible or does it require human interpretation? What makes particular emails or patterns of email significant? What are the potential personal and social functions of a visualized email archive? What are some ways of classifying one's contacts? Are these overlapping categories? hierarchical categories? do they change over time? Sketch a design for an email based contact management interface. How does it deal with evolving relationships over time? Visualising Email Archives As outlined in the readings this week, email has become a ubiquitous component of daily life for almost all those who have regular access to a computer and internet connection. It's uses are varied, and connected to many facets of daily life, from work contact, social interaction to newsletters and information services (such as online ticket confirmation, or MIT's 'reuse' recycling list). Despite it's unsuitability for the tasks, it is also used for document transfer, memos and information storage (for example, travel guides now often recommend that people backpacking around the world, scan their important documents and email them to their own account to be printed in emergencies in an internet cafe). Ducheneaut and Bellotti, ascribe this email software 'overloading' to its embedded nature in people's daily workflow. Despite this richness of use, email programs do little to distinguish one type of email from another or aid users in obtaining an overview of their interaction patterns and usage histories. Email can be interpreted and visualised to present some form of personal history. Either as an aid to a user's understanding of their own email usage history, or, as Donath proposes, as a form of display, analogous to offline examples of home decoration or a scrap book. I think the personal histories created from email archives are not like diaries or blogs as they are fundamentally derived from statistics of email use and semantic analysis rather than a user defined narrative. They are more analogous, even when the visualisations are edited by the user before 'publishing', to measures of pulse or a family tree (or the online example of MoodStats). Donath identifies three of the possible ways to organise visual representations of email; temporal sorting, social network patterns and by topic. Other metrics which might prove interesting to visualise include physical location of each sender, purely personal value (such as monetary wealth or social status - although these values would have to be obtained from external sources) or interest (such as a comparison of links suggested in email to browser history). This significance of any one particular email is I think a computationally challenging task, as the content is often brief, partially formed or part of a longer conversation. Certainly some semantic analysis may provide useful results, but I think success is more likely if email are analysed in the context of longer conversational threads. Further social information and analysis such as Warren Sack's "Conversation Map" can be achieved if a sender's email is analysed in its entirety, or even groups of senders (such as all those working on a project). Style, spelling and word choice along with a few formatting options are the few non language based forms of expression in email. Analysis of these may lead to patterns in non standard vocabulary or grammar use indicative of social group emphasis. It is also possible that an e-mail's start and end comments may give some indication of familiarity. In particular if someone includes a nickname for themselves, this may provide an indicator of perceived familiarity. Subject lines can often slide out of context with the content of messages, and so seem misleading to analyse. Presenting patterns of communication between two people or in a small group may act to strengthen bonds, and increase communication linked with the direct feedback on email traffic. In current email software, users can easily loose sight of online relationship development, as well as their overall usage and time investment in email interaction. Emphasising these aspects visually can fulfil an organisational role, provide a rich 'habitat' and even encourage further investment of time. In an asynchronous communication medium such as email, providing a visual history of dialogues can be vital in maintaining conversational momentum, context and interaction dynamic. It seems to me that the visual organisation of email contact, should allow an informative overview of interaction, and highlight important email based on past interaction for a user as a form of organisation or provide a more abstract habitat. It is worth noting that this form of organisation may be detrimental to interaction with predominantly offline acquaintances that user is trying to remain in contact with online. It is therefore useful to add a further visual indication of a user's desire to remain in contact, as well as the measured past performance. In her paper Visualizing Email Archives, Donath argues persuasively for not storing email in folders (as users find this cognitively difficult), but instead ascribing any number of user defined attributes (similar to the functionality of the "virtual folders" implemented in the Evolution email software package). I would go one step further and suggest a sliding scale (or even a parameter surface) for attribute strength in cases where the choice is not binary (e.g. a sliding scale from not at all to entirely work related, a flag for "about fishing", a sliding scale for funny, etc.). This allows people to be indecisive and still sort things accordingly. In their research, Ducheneaut and Bellotti found that category choice in sorting email depended to some extent on role. Sorting by project activities suited researchers, sorting by department suited company workers and personal versus private divisions suited entrepreneurial workers, all to a greater or lesser extent. It is clear that categorisation in email organisation must not be mutually exclusive (for a work colleague may also enjoy fishing), must be user defined (although low cost in terms of time invested and complexity) and must be adaptable over time to suit shift in user perceptions and role. There seems very little cost in keeping information on all contact with anyone, the challenge is then to provide better sorting and categorisation and search functions (again if cost in the use of search can be reduced sufficiently - see for example the search in the iTunes software) to a user. For my email visualisation design, I decided to use the premise of spatially sorting correspondents according to user defined categorisation as a starting point. If an email conversation can be represented visually, and clustered with other similar conversations in terms of categories, I think an easily understandable environment could be created.  My first step is to
return to my email visualisation sketches for the abstract
representations assignment in week 2 of this course. Here I was
interested in emphasising the fractal like nature of long-term
interaction by emphasising the wax and wane of emails in such a
contact. The visualisation space was composed of shapes, each
representing a user-sender relationship and was meant to act as a
backdrop (or habitat in this week's parlance) for a more standard
compose/reader window. A texture component was added for each email
interaction, and the previous texture components for that sender shrunk
according to the time lapsed since the last email sent or received.
Thus the shape gave indication of frequency of interaction. I like this
approach, and have decided to try and extend this system. I propose an
email visualiser that first sorts and removes junk and information only
emails (such as subscribed emails) to a separate viewer. This allows me
to concentrate on visualising email interactions. In my original design, each sender relationship was
ascribed a texture by the user, thus allowing for personalisation of
the environment, and labelling of relationships the user deemed
important (weather or not their email traffic was high). I want to
continue this idea, but to include a generic pattern for textured
shapes that the user has not assigned their own pattern to. Following
the argument presented that emails are composed only of text, and that
visualisation should reflect this, I chose to try modifying the font
characters themselves to form pattern 'tiles' of varying complexity.I wanted the overall complexity of an email history shape to reflect frequency and longevity of interaction so that it is easy to see the tempo of an email exchange over time. I decided to compliment this by linking the complexity of a pattern 'tile' (the element added, around which the previous tiles are shrunk and placed) to the frequency of exchange in some unit of time (such as a day). I wanted to show imbalance in email sent and received, and thought that the symmetry of a tile might show this with the symmetry of the letter forms equating to symmetry in the exchange.  These meaningless characters that form the tile patterns can possibly be built from a set of components and simple kerning and symmetry rules. As an email contact (of initially unknown importance) is formed, and develops, dots and punctuation evolve into complex shapes as dialogue progresses.  I wanted to present temporal information through symbolic
representation of an email exchange fading in luminosity with time
(~months). It would darkened again slightly if the user revisited an
email (not linear, something like logarithmic, so each return visit
caries more weight, to reflect growing importance of an old email). The
shape would also darken more strongly for a copy from and paste to
another page action, and similar indicators of current use. This sounds
similar to Hill, Hollan, et al.'s work with EmailWear.I considered the possibility of mixing two colours to represent the ratio of questions unanswered (measured possibly by the number of '?'s) between the user and sender. For example, yellow for questions not responded to, blue for questions asked of you. In terms of the visualisation as a usable environment, I think access to the overview screen should be available through a number of possible routes. Each should be optional, tuneable to a users needs.
I envisage a program that initially sorts email according to automatically generated rules based on semantics, past cc affiliation, signature (if identifiable) etc.. Then allow a user to quickly reweigh a message (thus updating the automatic rules) using a low cost method such as a ToolGlass.  |