Research Interests

My research lies in the areas of mobile computing, sensing, wireless networking and the investigation of the interraction between the computing and the society. In particular I am interested in:

Current Projects

Anticipatory Mobile Computing

Modern mobile computing devices, such as smartphones, are equipped with an array of sensors: accelerometers, GPS chips, light and proximity sensors, microphones, cameras, barometric pressure and heart-rate sensors. With the help of machine learning sensor data can reveal high-level context of the user, e.g. whether a user is at home/work, running or walking, who is the user talking to, we can even infer the user's stress level. If we accummulate more data, and push machine learning models further we can predict a user's future context. Thus, we can predict the next place a user is about to visit, or her physical activity in the next hour. In our research we propose and investigate Anticipatory Mobile Computing, which aims to bring actionable decisions based on the past, present and predicted context information on a mobile device. This new paradigm opens up a whole new gamut of applications, from navigation systems that take into account predicted congestions, to personalised assistant applications that prepare us for an unexpected meeting with a potential business partners, or proactive health care advisor apps that tackle depression, diabetes and substance abuse. For a through overview of the field of anticipatory mobile computing please take a look at our ACM Computing Surveys article.
Key stages in the anticipatory mobile computing depicted on an example stress management application: collecting sensor data, processing it in order to infer context, predicting future events and using past, present and future to make intelligent autonomous decisions

Digital Behavioural Change Interventions

The global population's health is severely affected by diseases that could be curbed and prevented if the patients are to adhere to certain behavioural practices. Such diseases include obesity, substance abuse, stress-induced hearth diseases, diabetes, to name a few. Behavioural change interventions are a well-studied means of inducing a positive change in a patient's behaviour. However, these interventions are costly to administer, and do not scale to millions of people who suffer from the above diseases, since the interventions require regular meetings with behavioural therapists. Mobile phones represent a promising platform for scalable, timely and personalised digital behaviour change intevention (dBCI) delivery. First, the phones are very personalised devices carried by their owners at all times, and thus can be used to deliver the information at anytime and anywhere. Second, the phones are equipped with sensors that can provide a clear picture of a user's state and context. This makes phones suitable for remote diagnosis of the user state, and delivery of information at the right context. For example, weight loss information can be delivered to a user as soon as he enters a restaurant. Finally, phone applications naturally scale to millions of users with little additional cost. In our work we develop UBhave, the first framework that supports generic dBCI authoring, delivery and administration. We invite interested behavioural researchers to contact us if they are interested in piloting a dBCI study using our framework.
Overview of the UBhave framework.

Mobile User Attention Management

We live in increasingly interactive worlds, where our attention quickly switches from one thing to another. We work on PCs, get text messages via mobile phones, get notification on our running performance via smartwatches, and while on a train read Facebook updates on a smartphone. However, all these devices and services competing for our attention lead to annoyance and stress in case notifications arrive at inappropriate moments, for example while we are at a business meeting, on a bicycle, in a theatre. In this research we try to identify good moments to deliver information to a mobile user. First, we show a relationship between the context in which a user is, such as her location, physical activity and the time of day, with her response time to a notification and the sentiment towards being interrupted. We then build a system, an Android library, termed InterruptMe, that uses sensors on a smartphone to recognise the user's context and infer if a user is interruptible or not. We test the library in a real-world application, and release it as an open source software. You can read more about InterruptMe in our Ubicomp'14 paper.
InterruptMe - an open source library for notification management on Android devices.

However, the context is not the only thing that determines user's interruptibility. We also examine the role of the sender-receiver relationship and the content on a message on user's receptivity towards this message. We find that mobile users react differently to messages sent by their friends and family, than to generic messages sent by system services, for example. Further, we examine how the current task that users are working on impacts their sentiment towards being interrupted. Challenging tasks that users are highly concentrated on, are more difficult to interrupt than the ones that are a part of a user's everyday routine. In our future work we aim to automatically recognise which tasks can be interrupted easily, and implement an intelligent notifications scheduling system that minimises user frustration. You can read more about this study in our Ubicomp'15 paper.
Activity - Interruptibility
Time to react to a mobile notification depends on the activity that a user is engaged in.