D. STELLARnet Incubator

Incubators are targeted at early-stage researchers. The idea is to support promising researchers with mobility or “incubation” scholarships to spend up to few months in another influential laboratory or research centre to foster integration and the establishment of close relationships. This instrument will support cross-fertilisation while working collaboratively on selected exploratory research actions.

STELLARnet is launching an Incubator Programme, aimed at supporting promising early career researchers with mobility or “incubation scholarships” to spend some months in another research institution or enterprise with the aim of either fostering the incubation of new ideas/projects, or exploiting new prototypes/services/approaches.

An Incubator must involve:

an early career researcher (not in his/her early PhD stages), who is interested in a topic/idea/prototype relevant in the field of TEL
a host institution, interested in the same topic/idea/prototype.

It is a necessary condition that the host institution and /or the junior researcher are within STELLARnet.

Within the First Call (launched in December 2009 and closed on January, 15th 2010), one Incubator was approved (see details below). The activities of the Incubator took place in Summer-Autumn 2010.

Incubator Title	Person & Institution	Host Institution
CoMoCo - Combining Gaze data with audio and action logs to build a computational model of collaboration quality	D. Diziol Mullins (University of Freiburg)	EPFL (c/o Prof. P. Jermann)

Summary (provided by D. Mullins at the end of the activities)

The Incubator addressed one of the major challenges in the area of TEL, more specifically in the area of computer?supported collaborative learning: the dilemma of providing too little structure to students’ interaction on the one hand and overscripting collaboration on the
other hand. Adaptive collaboration support is regarded as a possible solution. However, the development of adaptive support is difficult as it is necessary to define and assess meaningful indicators for the quality of students’ collaboration.
Along these lines, the main goal of the Incubator was to develop a computational model for collaboration quality based on action, audio and gaze data. To reach this goal, we adapted an interaction model to assess collaboration quality based on dialogue (rating scheme). The model contains 6 dimensions: collaboration flow, mutual understanding, time management, technical coordination, task division, & participation symmetry. We applied the rating scheme to analyze the collaboration quality of two contrasting dyads.
In a second step, we analyzed specific gaze features in detail for the 2 dyads mentioned above in order to link the analysis of interaction quality with the gaze patterns. Based on this analysis, we were able to identify gaze, action, and audio correlates that are connected to the rating dimensions. For instance, we could show that the cross recurrence, in other words, the amount of time the dyad partners look at the same things at about the same time, is closely related to the dyad’s collaboration flow; that the fixation of the remaining time can indicate the quality and amount of the dyad’s time management; and that a sustained audio signal as well as the distribution of speech are important indicators for participation symmetry.
The collaboration led to a joint publication that will be presented at CSCL 2011 in Hong Kong.
Future plans are to evaluate these gaze footprints with a bigger sample to statistically confirm our initial results.

ATTENTION: the instrument has been merged with a WP4 instrument and has now changed name; the new mane is "MOBILITY FELLOWSHIP".
The Call for Mobility Fellowship was launched in January 2011. Go to: http://www.stellarnet.eu/instruments/doctoral_consortium/mobility_programme/ for more info about this new instrument!

The Incubator addressed one of the major challenges in the area of TEL, more specifically in the area of computer?supported collaborative learning: the dilemma of providing too little structure to students’ interaction on the one hand and overscripting collaboration on the
other hand. Adaptive collaboration support is regarded as a possible solution. However, the development of adaptive support is difficult as it is necessary to define and assess meaningful indicators for the quality of students’ collaboration.
Along these lines, the main goal of the Incubator was to develop a computational model for collaboration quality based on action, audio and gaze data. To reach this goal, we adapted an interaction model to assess collaboration quality based on dialogue (rating scheme). The model contains 6 dimensions: collaboration flow, mutual understanding, time management,
technical coordination, task division, & participation symmetry. We applied the rating scheme to analyze the collaboration quality of two contrasting dyads.
In a second step, we analyzed specific gaze features in detail for the 2 dyads mentioned above in order to link the analysis of interaction quality with the gaze patterns. Based on this
analysis, we were able to identify gaze, action, and audio correlates that are connected to the rating dimensions. For instance, we could show that the cross recurrence, in other words, the amount of time the dyad partners look at the same things at about the same time, is closely related to the dyad’s collaboration flow; that the fixation of the remaining time can indicate the quality and amount of the dyad’s time management; and that a sustained audio signal as
well as the distribution of speech are important indicators for participation symmetry. The collaboration led to a joint publication that will be presented at CSCL 2011 in Hong Kong.
Future plans are to evaluate these gaze footprints with a bigger sample to statistically confirm our initial results.