Events
Previous month Previous day Next day Next month
See by year See by month See by week See Today Search Jump to month
The role of politeness in the computational modelling of tutor diagnosis and feedback provision Print
Location:
LKL Auditorium
Host/Speaker:
Kaska Porayska-Pomsta

Date and Time:
Thursday, 29 March 2007, 12:30 - 14:00

Speakers’ ability to recognise socio-psychological states of their interlocutors and to accommodate them whenever possible are crucial in facilitating successful communication (e.g. Grice, 1989; Brown and Levinson, 1987; Clark, 1995). The theory of linguistic politeness (Brown and Levinson, 1987) provides us with rich insights into how such considerations as another person’s ‘face’ may affect the linguistic mediation of interlocutors’ goals, wants and needs. Regardless of the communicative domain, social and cultural norms as well as the context of immediate situation are all determinants of the way in which we communicate our messages and the extent to which our actions are intended and interpreted as face threatening. Until recently the exploration of social, cultural and psychological factors in communication belonged to the realm of theoretical research in social science and social linguistics. However, recent surge of technological developments and the recognition of the potential benefits of technology use for educational purposes led to a an exploration of the potential of politeness theory as a basis for improving human-computer interactions (HCI) (e.g. Bickmore and Cassell, 2001; Johnson et al., 2005), as a basis for interpreting human tutors actions (Person et al., 1995), as well as for managing artificial tutors’ feedback provision within Intelligent Learning Environments (ILEs) (Moore et al., 2004; Johnson et al., 2004). The appeal of the theory to the research in HCI and ILEs is two-fold:

  1. the theory provides a formal explanation of linguistic variation and
  2. it relates speakers’ diagnoses of their interlocutors’ psychological states to concrete communicative strategies for acting upon such diagnoses.
The first is an essential element of any system involving natural language; the second is crucial to a systems ability to reason about the communicative and, in the case of ILEs, educational viability of its output.

In this talk I will discuss the impact of the theory of linguistic politeness on the way in which successful human-computer tutorial interactions may be facilitated and how it can be modelled formally to serve as the basis for the design of ILEs that are both more socially and emotionally intelligent. I will discuss how the theory can be used to facilitate classification of human tutors’ feedback strategies in order to support automatic selection of artificial tutors’ feedback. I will present the results of a small-scale evaluation of the model suggesting that automatic selection of tutorial feedback based on politeness consideration bodes well in comparison with the feedback produced by a human tutor in similar situations. Finally, based on recent studies of human tutor-student interactions conducted in collaboration with colleagues in Edinburgh, I will discuss its potential impact on the way in which diagnoses of learners’ affect may be conducted automatically. I will conclude with remarks on the potential impact of this approach on the way we may investigate requirements of human-computer tutorial interactions in general.

Back