AppSensor
After attending the Mobile HCI 2011, it is time to summarize some of the experiences I had there — and let’s start with the most interesting paper I listened: Falling Asleep with Angry Birds, Facebook and Kindle — A Large Scale Study on Mobile Application Usage by Böhmer et al. (2011).
As we all know, the mobile application market has boomed, mainly due to iOS and Anrdoid-systems and the online market places they have. However, what I haven’t seen yet is a study on the application use and lifetime in a more broad manner. I remember seeing in an paper in the Pervasive 2010 about one game and the usage patterns of it. But, we don’t know so much on how the applications are used, in what situations they are used and more focused, why are they used. We know… they are used.
Böhmer et al. suggest a new sensor type called appsensor, only to focus on application and use of them. This allows us to go deeper in the use, just seeing what applications are popular at what time and what places — like done in the paper. Naturally, one can figure out more uses for appsensors — self organizing maps, data mining, … all the cool applications. Maybe I briefly highlight some aspects observed in this study to demonstrate the value.
Firstly, the sample is based on 4 000 something Android users, mostly from the states. They use all applications about one hour per day, and an average time application is opend… is 72 seconds. Short time… The core functionalities of traditional phones are used even more shortly: communication 47 seconds, maps 45 seconds, productivity — like calender — 61 secs. The “new” emerging things, multimedia, browsers, games, lifestyle apps, clearly are in the better side: 83 secs, 74 secs, 114 secs and 168 secs.
There are spikes when specific application categories become more important: tools category is popular around 6 am to 8 am, communication apps dominate the use 11 am till 8 pm and games are played during the evening. And location affects this too, some apps are popular in airports, and there are differences between Europeans and US users.
In general, what the results indicate is that we can build smarter smartphones. Instead of being a multipurpose device, it seems that certain patterns exists and the next step would be to facilitate these. And, for science — now this data was anonymized, can more data of users make us see even more. Especially this data is important, as similar kind of data surely exists in the HQs of Apple, Google and Nokia: we need to understand what that means as a privacy question.
Tags: application, conference, data
September 10th, 2011 at 1:21 pm
I guess you know about 360 panel at Nokia? http://www.google.com/search?sourceid=chrome&ie=UTF-8&q=360+panel+nokia
There are many entities that track mobile phone usage and appsensor is nothing new, though the paper was still interesting.
September 11th, 2011 at 12:22 am
Yep there are some papers that analyze app usage. Still the paper has two major contributions. First, in line with some other recent work (including our own), the paper shows how researchers can do this kind of work without requiring huge resources or the support of major companies. Second, the results are based on the largest sample I’m aware of (with minimal resources required).
While the paper itself is IMO very cool I beliefe that there is a lot more potential in this approach. E.g. I would love to see work that combines app usage with transportation mode recognition. Thereby it could e.g. be clarified if people REALLY walk while typing or reading (which is basically the motivation of a large corpus of mobile HCI work).
September 12th, 2011 at 7:22 pm
J Kujala: Yep, SP 360 panel has a long study done by Nokia at they provide some interesting details on how the behavior has changes. I was lucky enough to see some data visualisations while working at Nokia. And, I’m sure Apple, Google, Samsung, … have their own engineers & scientist working on this. However, those data sets usually tend to stay inside the company.
And yes, there are some papers focusing on single app or few app usages out in the scientific field, nicely referenced in this paper too. So, the concept of measuring application use is not novel. However, I do like the notion of “appsensor”, makes it sound like an every day activity that we can use as an input for other systems.
September 12th, 2011 at 7:28 pm
Niels: Clearly, there are many items that can be connected to this study, and there are many things that can be digged out from the dataset: luckily, as I understood, the data set will be opened for collaboration. So, maybe next years Mobile HCI will have more on apps and how we use applications — there is science to be done!
January 18th, 2012 at 8:30 pm
Thank you for the feedback and the discussion here. Maybe you are interested in our “everything on one glance” infographic.
Find it here: http://appazaar.net/study/