Interruption management

Abstract

As most of us spend a substantial amount of time on computers nowadays for either work or personal tasks, it is very easy to realise that interruptions occur on a very regular basis. Such interruptions often tend to cause an information and cognitive overload due to the inundation of data entering our desktop which disrupt us from our current task. Looking at a few years back, people used to be able to interrupt you only by calling or physically walking into an office for example. Nowadays, through various technological advancements, interruptions can occur via e-mails, instant messaging, mobiles and countless other mechanisms. We focus on the area of interruption management on one's personal desktop and attempt to find a solution to alleviate as much as possible any disruptive interruptions by understanding the user's desktop context and current activities. This framework will then be able to decide if the user should be interrupted or not accordingly and if so, find the most opportune time. The solution we propose is based on machine learning technology, in particular Naive Bayesian Classifiers through extraction of information regarding the user's desktop activities as he is working on a task. After a period of training, the system recognises activities on the user's desktop which are of a non interruptible nature (i.e. user is in a busy state) and activities during which the user is available. We also present a framework to aid the user in reducing interruptions found in today's Internet environment and aim to decrease the resumption time once the interruption has been attended to. Our preliminary evaluation results indicate a good potential in the system to be used in the real-world. Participants in our evaluation gave positive feedback after using the system between three to five days. It emerges that the system is very quick to learn the user's desktop activity trends as it only required the participants a few hours (between 2-4 hours) for the system to become useful.