TY - GEN
T1 - Revealing and Interpreting Crowd Stories in Online Social Environments
AU - Kiefer, Chris
AU - Yee-King, Matthew
AU - D'Inverno, Mark
N1 - The work reported in this paper is part of the PRAISE (Practice and Performance Analysis Inspiring Social Education) project which is funded under the EU FP7 Technology Enhanced Learning programme, grant agreement number 318770.
PY - 2015/7/26
Y1 - 2015/7/26
N2 - The underlying patterns in large scale social media datasets can reveal valuable information for interaction designers and researchers, both as part of realtime interactive systems and for post-hoc analysis. Music Circle is a social media platform aimed at researching the role of community feedback in online learning environments. A large dataset was collected when the platform was used as part of a Massive Open Online Course (MOOC). We developed a novel analysis technique for observing global patterns in the behaviour of students. The technique employs network theory techniques to view student activity as an interconnected complex system, and observes the temporal dynamics of network metrics to create timelines which are clustered into groups using unsupervised learning methods. This approach highlighted global trends and groups of outliers that needed further attention or intervention.
AB - The underlying patterns in large scale social media datasets can reveal valuable information for interaction designers and researchers, both as part of realtime interactive systems and for post-hoc analysis. Music Circle is a social media platform aimed at researching the role of community feedback in online learning environments. A large dataset was collected when the platform was used as part of a Massive Open Online Course (MOOC). We developed a novel analysis technique for observing global patterns in the behaviour of students. The technique employs network theory techniques to view student activity as an interconnected complex system, and observes the temporal dynamics of network metrics to create timelines which are clustered into groups using unsupervised learning methods. This approach highlighted global trends and groups of outliers that needed further attention or intervention.
UR - https://ceur-ws.org/Vol-1407/
UR - http://www.scopus.com/inward/record.url?scp=84938523881&partnerID=8YFLogxK
M3 - Conference proceeding
AN - SCOPUS:84938523881
T3 - CEUR Workshop Proceedings
SP - 47
EP - 52
BT - Proceedings of the First International Workshop on AI and Feedback (AInF 2015) co-located with the 24th International Joint Conference on Artificial Intelligence (IJCAI 2015)
PB - CEUR-WS
T2 - 1st International Workshop on AI and Feedback, AInF 2015
Y2 - 25 July 2015 through 27 July 2015
ER -
