Final Report Summary - ERIS (Experimental resource for the investigation of dialogue processes)
The three objectives of ERIS were:
(1) to carry out a programme of experiments that addresses key problems in dialogue research, using text-based computer-mediated interaction;
(2) to develop and refine the text-based computer-mediated methodology into a fully-integrated experimental resource;
(3) to distribute the experimental resource to the wider research community.
The fellow's research programme has addressed fundamental questions in dialogue research concerning how coordination in dialogue is established, sustained and develops during interaction. The fellow's research has investigated how the procedural infrastructure of dialogue supports the development of dyad- and group-specific vocabularies. A central focus of the fellow's experimental programme was on the processes involved in miscommunication: how interlocutors identify, signal and resolve problematic understanding. To investigate the interactive and cognitive mechanisms underpinning communication, the fellow's research involved the development of novel experimental techniques using the experimental chat tool resource.
Key findings from the fellow's research programme are listed below:
(1) Interlocutors rapidly conventionalise sub-languages for describing spatial relations, temporal relations, sequences, and also for describing facial features of unfamiliar faces. The sublanguages that emerge depend on the specific interaction history, in particular on sequences of clarification.
(2) As interlocutors become more coordinated with each other, their representations differ qualitatively from those used initially, becoming more systematised and abstract.
(3) Alignment and repetition of representations is exploited by interlocutors as a communicative resource, in particular as a means of clarifying problematic understanding.
(4) When dealing with problematic understanding, interlocutors align more with each other.
(5) As interlocutors become more coordinated, clarification requests querying referring expressions are increasingly interpreted as querying higher-level communicative intentions.
(6) Clarification requests querying high-level intentions cause less disruption as interlocutors become more coordinated, while for clarification requests querying reference, the converse is the case.
(7) Interlocutors develop idiosyncratic procedural routines for managing temporal and sequential coherence in dialogue. Unlike signals used to refer to objects, the basic structure of these routines is complementary.
(8) Convergence on routines does not occur via explicit negotiation of intentions. Instead, routines emerge out of tacit, reciprocal negotiation, via progressive differentiation of interlocking, complementary behaviours, culminating in normative conventions.
(9) These routines become conventionalised by groups of interlocutors, which interlocutors associate with specific conversational partners.
(10) Conventionalisation of these routines, like referential coordination, is achieved via tacit, as opposed to explicit negotiation.
(11) Finally, the fellows research programme has demonstrated the viability of manipulating the affordances of text-based interaction in order to conduct controlled experiments on dialogue.
The project produced one output. The experimental resource for investigating text-based interaction is available to the research community. The experimental resource now allows the fine-grained, context-sensitive, real-time manipulation of the content and timing of participants' dialogue in multi-party dialogue. It provides a suite of techniques that allow experimental control without restricting the interactivity of the dialogue. The level of control is unprecedented, enabling systematic, replicable, and context-sensitive interventions to be inserted fully blind into live dialogue. It has been developed as a core component of the fellow's experimental programme at Stanford University and University of Edinburgh, and has also been used and evaluated at King's College London, and Queen Mary University.
The toolkit offers a novel unified means of experimenting on the lexical, syntactic, semantic, pragmatic and social parameters of interactive dialogue - at the level of the individual, dyad and group. The findings from the research programme and the techniques that are consolidated in the experimental resource are intended for exploitation by a broad spectrum of interdisciplinary research. This includes psycholinguistic and sociolinguistic approaches, dialogue system design, studies of communication in joint action, experimental pragmatics, and studies on the emergence of conventions. The practical implications are far-reaching: informing the design of collaborative technologies for spatial navigation, online learning, and also assistive technologies. The experimental resource can be exploited to analyse change that occurs during an interaction, as the techniques allow the experimental manipulation of the underlying mechanisms that drive these changes. The toolkit simplifies these techniques and is distributed freely to the research community. The toolkit is available at http://cogsci.eecs.qmul.ac.uk/diet/
(1) to carry out a programme of experiments that addresses key problems in dialogue research, using text-based computer-mediated interaction;
(2) to develop and refine the text-based computer-mediated methodology into a fully-integrated experimental resource;
(3) to distribute the experimental resource to the wider research community.
The fellow's research programme has addressed fundamental questions in dialogue research concerning how coordination in dialogue is established, sustained and develops during interaction. The fellow's research has investigated how the procedural infrastructure of dialogue supports the development of dyad- and group-specific vocabularies. A central focus of the fellow's experimental programme was on the processes involved in miscommunication: how interlocutors identify, signal and resolve problematic understanding. To investigate the interactive and cognitive mechanisms underpinning communication, the fellow's research involved the development of novel experimental techniques using the experimental chat tool resource.
Key findings from the fellow's research programme are listed below:
(1) Interlocutors rapidly conventionalise sub-languages for describing spatial relations, temporal relations, sequences, and also for describing facial features of unfamiliar faces. The sublanguages that emerge depend on the specific interaction history, in particular on sequences of clarification.
(2) As interlocutors become more coordinated with each other, their representations differ qualitatively from those used initially, becoming more systematised and abstract.
(3) Alignment and repetition of representations is exploited by interlocutors as a communicative resource, in particular as a means of clarifying problematic understanding.
(4) When dealing with problematic understanding, interlocutors align more with each other.
(5) As interlocutors become more coordinated, clarification requests querying referring expressions are increasingly interpreted as querying higher-level communicative intentions.
(6) Clarification requests querying high-level intentions cause less disruption as interlocutors become more coordinated, while for clarification requests querying reference, the converse is the case.
(7) Interlocutors develop idiosyncratic procedural routines for managing temporal and sequential coherence in dialogue. Unlike signals used to refer to objects, the basic structure of these routines is complementary.
(8) Convergence on routines does not occur via explicit negotiation of intentions. Instead, routines emerge out of tacit, reciprocal negotiation, via progressive differentiation of interlocking, complementary behaviours, culminating in normative conventions.
(9) These routines become conventionalised by groups of interlocutors, which interlocutors associate with specific conversational partners.
(10) Conventionalisation of these routines, like referential coordination, is achieved via tacit, as opposed to explicit negotiation.
(11) Finally, the fellows research programme has demonstrated the viability of manipulating the affordances of text-based interaction in order to conduct controlled experiments on dialogue.
The project produced one output. The experimental resource for investigating text-based interaction is available to the research community. The experimental resource now allows the fine-grained, context-sensitive, real-time manipulation of the content and timing of participants' dialogue in multi-party dialogue. It provides a suite of techniques that allow experimental control without restricting the interactivity of the dialogue. The level of control is unprecedented, enabling systematic, replicable, and context-sensitive interventions to be inserted fully blind into live dialogue. It has been developed as a core component of the fellow's experimental programme at Stanford University and University of Edinburgh, and has also been used and evaluated at King's College London, and Queen Mary University.
The toolkit offers a novel unified means of experimenting on the lexical, syntactic, semantic, pragmatic and social parameters of interactive dialogue - at the level of the individual, dyad and group. The findings from the research programme and the techniques that are consolidated in the experimental resource are intended for exploitation by a broad spectrum of interdisciplinary research. This includes psycholinguistic and sociolinguistic approaches, dialogue system design, studies of communication in joint action, experimental pragmatics, and studies on the emergence of conventions. The practical implications are far-reaching: informing the design of collaborative technologies for spatial navigation, online learning, and also assistive technologies. The experimental resource can be exploited to analyse change that occurs during an interaction, as the techniques allow the experimental manipulation of the underlying mechanisms that drive these changes. The toolkit simplifies these techniques and is distributed freely to the research community. The toolkit is available at http://cogsci.eecs.qmul.ac.uk/diet/