On this year’s Microsoft Research Faculty Summit, Liz Lawley complained that
Many social-networking sites essentially force users to become part of a huge community, or they force users to choose whether someone else is a friend or not, with no other subtleties defining that relationship
Of course, this direction fits perfectly to my thesis. But more specifically, I get the impression that more “subtleties” are nice and essential but also require a lot of effort by the user. Maybe similiar to metadata that was/is supposed to establish a “semantic web” but needs very simple interfaces to come to real use (delicious’ tag auto complete might serve as a good example). But while “bad” tagging might just mess up your knowledge base, getting the subtle interpersonal relations adjusted wrongly will get you in deeper trouble with your friends (light friends/good friends/best friends).
via experientia via macworld
[this is just a fast article that will be extended later on, hopefully]
Finally, my studies at FH Potsdam come to an end. I will give the presentation of my Master’s Thesis and projects on
Tuesday, March 25th, 2008 at 15 h
in the FH Potsdam Casino.
It has been a tough time untill my book went into press and I’m still quite busy preparing a decent show for you. But, hopefully, you will enjoy it and I will succeed in gaining a proud and honourful Master’s degree.
Buddyguard is helping me with making up a proper guestlist. But you are invited now already, as a reader of my blog!
A public version of my home grown “buddyscanner” is now available! It is a visualisation tool that I built in order to analyse communication log files of my group of test persons. This data can be usually found as a part of your phonebill or it can be extracted out of email archives.
You can give it a try right away: Start the buddyscanner
Of course, the visualisation is not too meaningful until you use your own data. But with this anonymized version you can get an impression of how it looks and works. (If you would like to have a private visualisation (where private means your data and a safe, password protected place) just let me know: blog [at] emotisys.net)
The visualisation can be rearranged to reflect different aspects of the data. It offers items that can be found in the raw data directly (such as the overall duration of communication), as well as computed values like reciprocity. The final value (from the perspective of my thesis), relevance, is available, too. Relevance is similiar to a kind of “rank” or “importance” of that person as it is seen by the machine. Although I’m using rather simple scoring methods, the results were quite meaningful to my test persons, already.
Some additional explanations:
-
In order to rearrange the diagram you need to click into the select boxes at the end of each axis. There is a third box available that is used for the “third” dimension, which is mapped onto the size of each square.
-
Hovering over a data point will load a flyout window with a more fine grained diagram. To keep it opened, you can click onto the according square.
-
In the flyout, a bar for each call/mail is displayed at the day of the year when it took place. The height is related to the duration/size of the event. Light blue means it occured during (usual) work times, dark blue is for the evening and medium for the weekends.
-
You can make some remarks for other users in the comments field if you like to.
-
If you want to keep track of some points across differnt sortings, you can highlight them with the button at the bottom of the flyout.
If you want to see more, express your doubts or have some remarks, don’t hesitate to make a statement below!
The more I get into relation detection via communication data, the more services come to my mind. But of course, I don’t invent this wheel for the first time (Pete Warden’s blog brought a lot of evidence to me): In an article two years from now (already!) ZDnet UK has a nice portrait about the emerging business of email analysis. A positive focus is put on Clearwell Systems because of their special (unique?) ranking algorithm (oha! — I bet Google pays very close attention). Its software
weighs the background data and content of each email for several factors, including the name of the sender, names of recipients, how many replies the message generated, who replied, how quickly replies came, how many times it was forwarded, attachments and, of course, keywords.
Well, so do I… But in the light of a fully grown business, ranking emails gets away from a personal (autonomous) assistant that is just nice to have, handy and good for reflection. With the huge amounts of email produced every day and about every topic relevant to any business process, corporate email archives contain pretty any information a manager, and — more delicately — a prosecutor can desire:
Email has come to be viewed as a source of truth. If you want to know what really happened, you look at the email.
As it became clear to me, too, during my research, collecting and archiving (intercepting?) all electronic conversations improves the the basis for statistical analysis and heuristics and hence the quality of the ranking a lot. A lot of entities (Google, security authorities) are after our data, consequentially.
Pete Warden has to receive an honrable mention once more because his position of “trying to generate a useful index with no human intervention” resonates with my basic motivation, too. I find his blog to be imensly interesting and very relevant for my thesis: Like expoiting the time information inherent to email that I thought of using in some kind of “contact profiling”, all the privacy issues entangled, especially in business context, and drawing profit from the knowledge that accumulates often unnoticed in a company (or workgroup). And he complains about the missing Gmail Api, too. All written in a very comprehensive manner.
Imagine, you’re back from a trip abroad and want to tell your friends about all the fascinating experiences that you have made (And you either don’t have a blog for that purpose or don’t want to publish it publicly). Usually, that means you have to go through your entire address book and select the appropriate persons. However, if your computer knew about your relationships it could help you a lot with this task.
How could an interface for this case look like? Here are some propositions (and some problems to discuss!). …
Here comes the all new and sparkling abstract of my Thesis (old stuff). You might want to have a look at it and give it some comments!
-
In my thesis I propose the idea of a socially aware computer. In order to get to know the user‘s circles of friends, it will mine and analyse the data that is left as traces by her communication, mainly phone call logs and email archives. As a result, a value for personal or subjective importance can be computed for each person in the user‘s network.
This allows for a new arrangement of the personal address book so that more relevant persons can be found more easily – an important feature regarding our ever expanding and globalized personal networks.
Moreover, tasks that require knowledge about the user‘s personal relations can be handled automatically: One is turning the user‘s attention towards old friends that tend to be neglected when he is burried in work or because he is always on the run due to our mobile and flexible times. Another one is managing access to her personal data that she stores online, like photos, travel plans or her activity stream that gets created by recent software like Jaiku or Twitter.
Handling friends and acquaintances in such an environment opens up new challenges that are explored by means of a visual prototype. Different types of displaying, managing, and enriching information about related persons are developped. Results from a user testing will be provided.
As a preliminary study, the data sets of several people have been analysed and plotted into an interactive diagramm in order to investigate the potentials of the communication data given. It also offers the possibility to look for the relevant parameters that determine different types of relations (e.g. best friend or old friend).
To provide a conceptual background, existing social network theories are explored and related to personal, ego-centric ones. I take a closer look onto the whole process of operationalisation, i.e. turning human behaviour into quantifiable data by statistical methods. Finally, implications and problematic consequences of both, the software itself and the concept of the „network society“ in general, are discussed. The felt need to turn our friendships into „social capital“ is one of the most remarkable shifts in the functioning of our societies. Others can make draw profits from this capital if they collect detailed data to establish profiles of us and our relationships. Thus, the whole field of privacy is entangled.
And across all these dynamics, computers become so inseparably intermingeld into our daily social life that borders between our (extended) self and the machine is often hard to determine.
For a healthy relationship, you should leave a short notice for your friends at least once in a while. But — huh — somtimes, days are really packed and you tend to forget things like that anyway…
Why not let your digital companion take on some routine care taking? Computers are well versed with keep alive customs:
The “keep-alive” keyword [...] allows the sender to indicate its desire for a persistent connection.
Here is how it works:
With the socially aware address book (from my Master’s Thesis) it will know who your friends are–and you will be able to describe your social aspirations, too. You then only need to define where you store interesting images or your latest writings or what you’re currently occupied with and the system will start sending off short notices every now and then.
If your friend happens to implement the same digital assistant (second line in the picture), your digital sidekicks might end up in a circle of automatic messages and reponses, chatting along on their own. Bypassing your human existence altoghther…
But your assistant can also propose some more personal messages that require your contribution (as depicted in the last line).
After all, some digital support is better than neglecting your remote friends too much, isn’t it?
(discussion declared open…)
img
The mobile phone as a truely social device makes an ideal plattform for social network visualisations. This gets demonstrated in a very inspiring way by Steven Blyth in his master thesis project at the now discontinued IVREA. As I found out he researched a question in the immediate neighbourhood of mine:
How can the softness and ambiguity of our social worlds be visualized within the computational and binary context of a mobile device?
The Social Fabric is a representation of your social world, displayed as a single visual array [of avatars] on your mobile phone. It does not replace your address book or calendar but keeps you subtly informed [via the body posture or the avatars!] about which relationships are prospering, which you have neglected, and the overall state of your social fabric.
img
In lots of his ideas and writings I found good arguments for what I want to further investigate. A very good point is that ambiguous metaphors can avoid the impression that a computer system could be truely accurate about something that is vague by its nature: social networks. As I am following a rather number based approach at the moment, this is something I will consider (with this Paper by Thomas Erickson from IBM).
He also revived another fascination (deep inside of me and, actually, my thesis proposal) for agents and avatars. In his opinion they are not discarded by history, as one can hear often, but depend on the proper design and sometimes sophisticated technology. The more the latter flourish the more the first can emerge as useful companions.
In contrast to his work, visualisation is supposed to be only one facett of my thesis with further applications building on insights gained by them.
Something left unclear to a certain extent in his text is his profiling method, what I used to call the “long term relation records”. Especially when considering “old friends” and “family members” a good balancing between current communication behaviour and long time habbits can offer new possibilities to deal with the less active parts of our “circles of friends”.
Thanks to one of his co-students at IVREA, Myriel, for poking my nose into this work! It found some good resonance over different media: WMMNA (relates it to GORI), Wired, LIFT 07
Usually, your phone bill is a vast amount of numbers that nobody ever reads actually (secret services left aside). It gives you some interesting details if you search for something particular but it’s hard to get an easy overview over what was happening the last month. Now, this has changed! After some weeks of tinkering with code (mySQL, PHP, HTML and some JavaScript) some visual tools have rolled out of my workshop.
The first simple step sums up all of your time spent calling someone on the phone. Different colours for working hours and leisure time (and for the month under focus) are added for further pattern recognition like collegue/friend identification. First evaluations revealed already that some patterns are really characteristic for particular events in the past. That way, the visual attractiveness of certain patterns leads us to remembering interesting stories attached to these dates (that sometimes have been forgotten already). As a nice Extra the whole plot seems to be somehow related to a powerlaw.
A second graph is more oriented towards science and theory. One of the background-chapters in my Master-Thesis focuses on the (mathematical) structure underlying our social networks. Some (Barabási) say all networks of free choice are governed by powerlaws, others (Watts) think that our network of friends is described better by a bell-curve. Maybe I can deduce in reverse from the pictures I get what type of network is contained in a phone bill. It looks as if we talk a lot to non-friends, so far.
A third (not yet fully matured) version will focus on temporal patterns and therefore plots the month of the year against the hour of the day to locate each call. With this method I want to look for “hot” times with a lot of traffic, usually calm zones and possible dissenters.
The work on this graphic as well as the others shows that rather simple data from a phone bill can generate some complexity when it comes to meaningful visualisation. In order to manage this abundance of information I want to add more options to select and filter the dataset. I also need some means to enlarge the “resolution” (i.e. less information per area) for those points in the graphic that are currently examined by the user.
