Alistair Morrison

Fast multidimensional scaling through sampling, springs and interpolation

The term ‘proximity data’ refers to data sets within which it is possible to assess the similarity of pairs of objects. Multidimensional scaling (MDS) is applied to such data and attempts to map high-dimensional objects onto low-dimensional space through the preservation of these similarity relations. Standard MDS techniques have in the past suffered from high computational complexity and, as such, could not feasibly be applied to data sets over a few thousand objects in size. Through a novel hybrid approach based upon stochastic sampling, interpolation and spring models, we have designed an algorithm running in O(N√N). Using Chalmers’ 1996 O(N2) spring model as a benchmark for the evaluation of our technique, we compare layout quality and run times using sets of synthetic and real data. Our algorithm executes significantly faster than Chalmers’ 1996 algorithm, while producing superior layouts. In reducing complexity and run time, we allow the visualisation of data sets of previously infeasible size. Our results indicate that our method is a solid foundation for interactive and visual exploration of data.

Further into the wild: running worldwide trials of mobile systems

Many studies of ubiquitous computing systems involve deploying a system to a group of users who will be studied through direct observation, interviews and the gathering of system log data. However, such studies are often limited in the number of participants and duration of the trial, particularly if the researchers are providing the participants with hardware. Apples App Store and similar application repositories have become popular with smartphone users, yet few ubiquitous computing studies have yet utilised these distribution mechanisms. We describe our experiences of running a very large scale trial where such a distribution model is used to recruit thousands of users for a mobile system trial that can be run continuously with no constrained end date. We explain how we conducted such a trial, covering issues such as data logging and interviewing users based in several different continents. Benefits and potential shortcomings of running a trial in this way are discussed and we offer guidance on ways to help manage a large and disparate user-base using in-application feedback measures and web-based social networking applications. We describe how, through these methods, we were able to further the development of a piece of ubiquitous computing software through user-informed design on a mass scale.

A pivot-based routine for improved parent-finding in hybrid MDS

The problem of exploring or visualising data of high dimensionality is central to many tools for information visualisation. Through representing a data set in terms of inter-object proximities, multidimensional scaling may be employed to generate a configuration of objects in low-dimensional space in such a way as to preserve high-dimensional relationships. An algorithm is presented here for a heuristic hybrid model for the generation of such configurations. Building on a model introduced in 2002, the algorithm functions by means of sampling, spring model and interpolation phases. The most computationally complex stage of the original algorithm involved the execution of a series of nearest-neighbour searches. In this paper, we describe how the complexity of this phase has been reduced by treating all high-dimensional relationships as a set of discretised distances to a constant number of randomly selected items: pivots. In improving this computational bottle-neck, the algorithmic complexity is reduced from O(N√N) to O(N5/4 ). As well as documenting this improvement, the paper describes evaluation with a data set of 108,000 13-dimensional items and a set of 23,141 17-dimensional items. Results illustrate that the reduction in complexity is reflected in significantly improved run times and that no negative impact is made upon the quality of layout produced.

Improving hybrid MDS with pivot-based searching

An algorithm is presented for the visualization of multidimensional abstract data, building on a hybrid model introduced at Info Vis 2002. The most computationally complex stage of the original model involved performing a nearest-neighbour search for every data item. The complexity of this phase has been reduced by treating all high-dimensional relationships as a set of discretised distances to a constant number of randomly selected pivot items. In improving this computational bottleneck, the complexity is reduced to from O(N/sub 1/2 /N) to O(N/sub 5/4/). As well as documenting this improvement, the paper describes evaluation with a data set of 108000 14-dimensional items; a considerable increase on the size of data previously tested. Results illustrate that the reduction in complexity is reflected in significantly improved run times and that no negative impact is made upon the quality of layout produced.

Visualisation of Spectator Activity at Stadium Events

Recent advances in mobile device technology have opened up new possibilities in enhancing the experience of spectators at stadium-based sporting events. In creating novel applications for use in such settings, designers must be aware of the current practices of spectators and of features of the environment at such events that novel applications may seek to exploit. This work forms an early part of the Designing the Augmented Stadium project. Data sets have been collected from spectators, logging the results of Bluetooth scans alongside GPS location. This paper presents an information visualization tool that can be used in the analysis and exploration of this data, to provide insight into the activities of spectators, the relationship between an individual spectator and the crowd as a whole and the suitability of stadium environments for applications based on infrastructure such as mobile ad hoc networks (MANETs) and wireless mesh networking. Various visualization tools are described and example cases are illustrated, using several real-world data sets recorded at football matches.

Configuring Attention in the Multiscreen Living Room

We have conducted a video study of households in Scotland with cohabiting students and young professionals. In this paper we unpack five examples of how mobile devices are used by people watching television. In the examples we explore how screens are used together (a) in a physical ecology, (b) in an embodied way, (c) in an orderly way, and (d) with respect to others. We point out that mobile devices are routinely used to access media that is unconnected and unrelated to media on television, for example for sending and receiving messages, browsing social media, and browsing websites. We suggest that mobile devices are not used to directly enhance television programmes, but to enhance leisure time. We suggest that it is important, when considering mobile devices as second screens, not just to treat these as a design topic, but to pay attention to how they are interactionally integrated into the living room.

Ethics, logs and videotape: ethics in large scale user trials and user generated content

As new technologies are appropriated by researchers, the community must come to terms with the evolving ethical responsibilities we have towards participants. This workshop brings together researchers to discuss the ethical issues of running large-scale user trials, and to provide guidance for future research. Trials of the scale of 10s or 100s of thousands of participants offer great potential benefits in terms of attracting users from vastly different geographical and social contexts, but raise significant ethical challenges. The inability to ensure user understanding of the information required to provide informed consent and problems involved in making users aware of the implications of the information being collected all beg the question: how can researchers ethically take advantage of the opportunities these new technologies afford?

Probabilistic Formal Analysis of App Usage to Inform Redesign

Evaluation and redesign of user-intensive mobile applications is challenging because users are often heterogeneous, adopting different patterns of activity, at different times. We set out a process of integrating statistical, longitudinal analysis of actual logged behaviours, formal, probabilistic discrete state models of activity patterns, and hypotheses over those models expressed as probabilistic temporal logic properties to inform redesign. We employ formal methods not to the design of the mobile application, but to characterise the different probabilistic patterns of actual use over various time cuts within a population of users. We define the whole process from identifying questions that give us insight into application usage, to event logging, data abstraction from logs, model inference, temporal logic property formulation, visualisation of results, and interpretation in the context of redesign. We illustrate the process through a real-life case study, which results in a new and principled way for selecting content for an extension to the mobile application.

A Comparison of Distribution Channels for Large-Scale Deployments of iOS Applications

When conducting mass participation trials on Apple iOS devices researchers are forced to make a choice between using the Apple App Store or third party software repositories. In order to inform this choice, this paper describes a sample application that was released via both methods along with comparison of user demographics and engagement. The contents of these repositories are examined and compared, and statistics are presented highlighting the number of times the application was downloaded and the user retention experienced with each. The results are presented and the relative merits of each distribution method discussed to allow researchers to make a more informed choice. Results include that the application distributed via third party repository received ten times more downloads than the App Store application and that users recruited via the repository consistently used the application more.

Using location, bearing and motion data to filter video and system logs

In evaluating and analysing a pervasive computing system, it is common to log system use and to create video recordings of users. A lot of data will often be generated, representing potentially long periods of user activity. We present a procedure to identify sections of such data that are salient given the current context of analysis; for example analysing the activity of a particular person among many trial participants recorded by multiple cameras. By augmenting the cameras used to capture a mobile experiment, we are able to establish both a location and heading for each camera, and thus model the field of view for each camera over time. Locations of trial participants are also recorded and compared against camera views, to determine which periods of user activity are likely to have been recorded in detail. Additionally the stability of a camera can be tracked and video can be subsequently filtered to exclude footage of unacceptable quality. These techniques are implemented in an extension to Replayer: a software toolkit for use in the development cycle of mobile applications. A report of initial testing is given, whereby the techniques use is demonstrated on a representative mobile application.