Video content recommenders are becoming wide spread in the age of ubiquitous access to the internet and web 2.0. But how does the context of video consumption e ect content preference? This paper argues for a greater understanding of the impact of viewing context upon preference adaptation in the new age of multi platform, mobile video entertainment. This paper advocates an approach which investigates preference adaptation from the perspective of users self moderating content choices in response to perceptions of the current viewing context. The author suggests ethnographic study of naturalistic content consumption behaviours as a possible methodology to uncover insight into this area, which could inform design requirements for future video recommenders operating in cross context environments.
Video content recommenders have been with us for some time. Popularised by
TiVo1 on the set top box and MovieLens2and Net ix3 on the internet we have
seen video content recommender engines, (as well as peer recommendations)
propagate onto many of the worlds most popular web 2.0 video, movie and TV
web sites, see [
Web 2.0 applications are at the forefront of an increasing trend towards
video content aggregation and personalised recommendations. The technology
has established itself at a time when mainstream ubiquitous access to high speed
internet has exploded. This has fueled an entertainment revolution with access
to video services and long tail content over a multiplicity of mobile devices in
addition to more traditional content delivery routes such as broadcast television.
The next step in the enhancement of these systems (which is well underway) is
the joining up of services across devices and networks in order to enable a
consistent brand message and user experience. There are many examples of evidence
for this trend in the marketplace with content providers and service operators
alike o ering singular branded video content delivery propositions across all of
their television, internet and mobile services, [
A user utilising a video content service across a range of devices and viewing contexts may have expectations regarding the types of content they wish to consume within each of those situations. We can imagine a member of a family when home alone, taking the opportunity to watch one piece of preferred content over another purely because they know it is of interest only to themselves. They do so under the rational that the other content whilst of equal interest is also enjoyed by the family as a whole. Therefore further opportunities to watch that item are more likely to present themselves at some other time with the family group. Another example could be a user choosing not to watch a movie to their preference on a train commute. The reasons being that the time available, mobile device screen size, and context of sitting on a train surrounded by strangers would (from their perspective) spoil the experience in contrast to watching the same content at home on the settee in front of their wide screen television.
As operators and broadcasters look to unify video content services across devices and environments, several fundamental theoretical questions are raised in relation to the contexts of video consumption which impact upon content selection decisions and therefore the role of recommenders. In order to provide a good experience to the user, a video recommender acting as part of such a service would need to take consideration of the nuances of context. This raises a number of issues surrounding not only if recommender outputs can be ltered to provide the best utility to a user within a given viewing context, but also if the construction of a single user model is valid for a system collecting information from many di erent viewing situations.
We must therefore ask, do the contextual factors which surround use in different socio-technical environments in uence the video content selections users make at any given time? If this is the case, how can video content selections collected from within a speci c context be usefully applied within a user model with the intention of providing recommendations across a landscape of ever changing contexts of use? Finally, even with an e cient recommender system in place how can recommendation selection and presentation be optimised to cope with the conditions imposed on a viewing experience by those same contextual factors? These questions need to be answered to ensure the success of future video content recommenders operating across devices within ubiquitous mobile environments. 3
Addressing research problems in this area requires a consideration of what
context actually is. This short overview is not the place to delve deeply into the
literature in relation to de nitions of context, however many researchers agree
that there still remains considerable confusion surrounding the notion of what
context is [
This as a useful way in which to consider context in future video recommenders as historic restrictions on access to content due to the constraints of broadcast schedules and device connectivity are being rapidly eroded through technological and commercial advances. This approach addresses a world where we could view any video content anywhere. Choices are made in response to users own perceptions of the viewing context and predictions for the experience to follow, which in turn are based on past viewing experiences in others contexts perceived as similar.
Approaching the problem of context from the perspective of user perceptions
has many precedents in the literature. As example the concepts of situatedness
[
The hypothesis under consideration is that inter-contextual cues manifest within a consumption situation are used by the user in the formulation of their own perceptions of context, upon which they then base decisions to mediate their own content selection and consumption behaviours.
The research direction advocated in this paper relates to user research activities currently underway aimed at identifying those aspects of context which in uence video consumption behaviours and content selection. An ethnographic study is currently being conducted which is following a range of individual users through a two week period. An array of qualitative and quantitative data collection methods are being employed in an attempt to capture a picture of each users video content selection behaviours across the range of sources they consume video content from. In parallel the study will describe the setting (physical, social and technical) within which they consume. The goal is to attempt to identify self moderated patterns of content preference adaption and the important factors within each viewing situation which may signify the inter-contextual cues (upon perception of which) the user has responded by adapting their content choices.
This study is being carried out as part of a PhD project investigating the wider issue of user experience optimisation for future video content recommenders. The overall goal of the research is to investigate the possibility of a framework for a video content personalisation and presentation system which can operate across devices and consumption contexts whilst providing the best possible experiences for users.