Blending SFL and Activity Theory to Model Communication and Artefact Use: Examples from Human-Computer Interaction
Rebekah Wegener [+]
RWTH Aachen University
Jörg Cassens [+]
University of Hildescheim
Systemic Functional Linguistics (SFL) and Computer Science have a long history together. This ranges from Winograd's blocks world (Winograd, 1971), inspired by Halliday's work on grammar, over applications in Natural Language Generation (Bateman, 1998), through to our own work. For example, we have used concepts from SFL to help model contextualized, ambient intelligent systems ranging from the use of abstract concepts (Cassens and Wegener, 2008) and of parameters of context (Wegener et al., 2008) for requirements engineering to a stratified view on context modeling with a special focus on the semantic stratum (Butt et al., 2013) and its practical application in intention-aware automatic doors (Kofod-Petersen et al., 2009). Based on these experiences and our own work on the use of Activity Theory (Leont’ev, 1978, Nardi, 1996) within the same application domain, we propose an approach to requirements elicitation and modeling of contextualized, ambient systems that combines the tool-centric perspective from the cultural-historic approach to activity theory with the communications and meaning-centric perspective of SFL. For applications in ambient systems, SFL is an attractive model of language because it foregrounds language as a social semiotic and thus frames spoken and written communication as just one avenue for meaning making. However, it also raises some problems. Firstly, SFL focuses on human communication and requires cognition for most of the categories that it posits; for the others, it presupposes language. Secondly, SFL is language centric. Despite the assumption that language is a social semiotic, the inclusion of other modalities is in many cases quite a challenge. SFL assumes spoken or written language as the baseline and other modalities are often treated as the context. This has interesting implications for context and how it is modeled as well. Lastly, SFL has an under-developed model of culture and consciousness. Despite being well positioned to capture both aspects, neither have been adequately integrated into the theory, meaning that while we have a model of language as a social semiotic, we have no model of the semiotic agent or the environment of the semiotic agent. By comparison, Cultural Historical Activity theory (CHAT) is action centric and focuses on action rather than meaning. Because of this, it has no model of semiosis or meaning making and thus no way to analyse action beyond the representation of the action itself. It also lacks an elaborated model of context and a stratified view of interaction. CHAT does however have a strong model of culture that can accept a layered view of context. By focusing on action, it has developed a modelling of the activity sequence that can take GSP and other SFL analyses and provides a clear unit of analysis. Perhaps the greatest strength, particularly from a human-computer interaction perspective, beyond a model of activity, is the explicit inclusion of non-human agents as potential actors. Interestingly, for an approach focused on activity, Cultural Historical Activity theory has a well developed model of cognition that is social in focus and this means that we have an opportunity to integrate a powerful model of language with a similarly focused model of interaction and the mind. Because they share underlying assumptions, it is possible to integrate the two theories at abstract and applied levels. SFL has a history of building on and integrating other theories if the core assumptions are compatible (e.g. Halliday's integration of Berstein's theory of social organisation). Most importantly, they are non-competitive models. That is, they do not seek to explain or model the same sphere of human behaviour, rather they complement each other, meaning that by combining them they together explain and model more than each alone. Here, we set out our integration of the two theories at the abstract and practical level, citing examples from our previous and current projects in human-computer interaction.