Talks by Annie Quandt and Christopher Ellis

Annie Quandt, based in Human Factors, and Christopher Ellis, from the MRL, will present their current PhD work.

Annie Quandt: Analysing Adolescent Interaction with Agents: A linguistic approach

The advanced concept of ubiquitous computing is based on the notion that computing happens everywhere around us and penetrates every aspect of people’s lives. Teenagers are a specific group in which the keen use of technology is a phenomenon that has increasingly become subject to research in different contexts. In the limbo state of having left childhood behind but not having entered adulthood yet, the way adolescents use technology incorporates fascinating aspects such as the forming of identity and the management of relationships. Understanding the role technology plays in teenagers’ interactions is a crucial step towards increasing the user experience.

With interactions no longer limited to human face to face contact as agents are becoming increasingly pervasive in everyday life, there is a need to increase the understanding of human-agent interaction. This research project seeks to investigate adolescents’ interaction with agents such as self-checkouts in the supermarket, satellite navigation systems on smart technology and automated phone systems from a linguistic perspective. With language being a main shared feature of human-human and human-agent interaction, the research questions include:
• What are the perceived differences between human-human and human-agent interaction in speech?
• How do identity and diversity play into effective communication via voice?

The novelty of the research project lies in bringing to bear linguistic approaches and challenges on the broader question around the behaviour of teenagers in choosing and interacting with different agent environments. Though linguistic discourse models exist for human interaction, there is a need for new models that address the specific features of human-agent interaction. For these purposes it is crucial to gain insight into adolescents’ perception, inhibitions and biases in the context of agent interaction. A pilot study is currently being undertaken with eleven- to nineteen-year-olds. One part of this research study aims to determine which agent environments will be at the focus of this project. The comprehensive analysis and study of human-agent interactions, together with teenagers, will help to better understand the role of agent interaction and help to inform future linguistic features, design and functionality of these agents.

Christopher Ellis

In this talk, I will provide a brief overview of the field of recommender systems. I will begin by introducing the historic challenges and notable solutions that have shaped the field over the last twenty years.

In the second half of the talk I will focus solely on music recommendation. The main problem within this area is one of personalisation. The difficulty in addressing it arises from its fragmented nature. We consider a recommendation to be highly personal if it accurately reflects our tastes and is also non-obvious. A recommendation of "Hey, Jude" to a Beatles fan is likely to be accurate but is arguably not very personal. Going the other way, a recommendation of Cole Porter's "Let's do it" may be non-obvious but is not necessarily going to accurately reflect their tastes.

Within academia and the commercial world, it is increasingly being recognised that algorithmic solutions are not necessarily the answer to this apparent dichotomy. In the final part of the talk I will argue that a human lead user-interface approach is the best way to begin to address it.