Using Communication Frequency and Recency Context to Facilitate Mobile Contact List Retrieval

Introduction

Technological advances of the last decade have turned mobile phones to small multi-purpose personal computers being equipped with camera, GPS receiver, accelerometer, Bluetooth and other sensors. These devices are now used, among others, to access the World Wide Web, to transfer files, produce multimedia content, as email clients and digital calendars. However, mobile phones remain primarily communication devices (LaRue et al., 2010), supporting the communication needs of their users within their social networks with several tools. As such, it is reasonable that a common task for their owners include searching for a contact in a phonebook or selecting one from a recent-call list (Lee, Seo & Lee, 2010) in order to start a new communication session using one of the provided methods. As contact lists get increasingly bigger and since a significant percentage of contacts are never used (Bergman et al., 2012), the cognitive load on the user increases while trying to retrieve a contact from this repository. This effort is also obstructed by the limitation of the relatively small screen that mobile phones are equipped with. Furthermore, since call logs impart information about use and not lack of use, mobile devices have become good at supporting communication but provide little support for the task of managing social relationships (i.e. deciding who to contact and how frequently), leaving decisions entirely to the users.

At the same time, mobile devices collect a significant amount of data and information about the user's context, including location, the current date and time, the orientation of the device, whether the user of the device is on move and his speed, the user’s current task (e.g. on the phone, messaging), whether the vibration or the silent mode are enabled etc. (Komninos et al., 2011b). The user considers her mobile device a “trusted device”, usually having it close to her, sometimes operating 24 hours per day. Devices also contain a lot of personal information related to the user’s social environment (Toninelli et al., 2008). These are either generated automatically by the device (e.g. a phone list saves the calls that have been made, the time of the day for each call and the duration of each call for the past few days or even weeks) or consist of user-generated content (e.g. SMS/MMS and multimedia files, browser's history, calendar events etc.). Therefore, a mobile device could also be aware of the social environment of the user (social context). The combination of social and mobile context results in a dynamically defined social context, termed the mobile social context (Gilbert, Cuervo & Cox, 2009).

Our work is based on the hypothesis that context mined from personal interactions with a mobile device can be used to aid personal mobile information retrieval tasks. In this paper we attempt to address the problem of contact retrieval when performing an outgoing call, by predicting which contact is the most probable to be called at any time. Though communication often takes place on mobile phones through not just phonecalls and sms, but other networks such as Skype, Facebook, Twitter etc., it is not yet possible to collect all such communication data from the various apps due to varying data access permissions. Such permissions however are available for the most basic communication modes, i.e. phonecalls and short text messages.

While contacts are just one aspect of mobile personal information, considering modern devices’ capacity to store information (several gigabytes are available on most devices) and the fact that additional storage capacity is afforded by cloud services, personal information management is likely to pose significant challenges to users in the future. Hence, methods for managing context with a view to inform retrieval tasks can be applied to multiple retrieval situations. We believe that a solution to such problems can be informed by mobile social context as mobile users seem to adopt different behaviour patterns under different contexts. As an example consider the following scenarios derived from our experiment participants that involve two different context dimensions, frequency and recency: