How Do In-video Interactions Reflect Perceived Video Difficulty?

  • Upload
    nan-li

  • View
    162

  • Download
    1

Embed Size (px)

Citation preview

How do iN-VIDEO inTERACTIONS rEFLECT pERCEIVED VIDEO dIFFICULTY?Nan Li (Presenter), ukas Kidziski, Patrick Jermann and Pierre DillenbourgCHILI Lab, Swiss Federal Institute of Technology Lausanne (EPFL)

Presented at the 2nd European MOOCs Stakeholders Summit,EMOOCs 2015, Mons, Belgium, 18/05

BackgroundLecture video viewing is central learning activity in MOOC has been investigated topredict dropout rates (Halawa et.al 2014; Sinha et.al 2014)predict performance (Jiang et.al 2014)analyze demographic differences (Guo & Reinecke 2014)analyze learner engagement (Kizilcec 2013)mostly analysed at macro-levelNumber of videos watched (Anderson et.al 2014)Engaging time (Guo & Reinecke 2014)Ecole Polytechnique Fdrale de Lausanne

Ecole Polytechnique Fdrale de LausanneProblem Statement

?Do in-video interactions reflect students perceived video difficulty?If yes, then how do different types of video interactions relate to it?

Method

End-video Survey ExampleEcole Polytechnique Fdrale de Lausanne

DataSET DescriptionCourseVideosActive LearnersVisitSessionsResponse ratesDifficultyReactive Programming3622794First26549349.1%2.699Revisiting20550123.7%2.837Digital Signal Processing589086First5834932.8%2.478Revisiting5961012.7%2.593

Ecole Polytechnique Fdrale de LausannePre-processingMismatched video events (e.g. incorrectly logged time)AnalysingVideo sessions with survey responses

Video Interaction ProfilesEcole Polytechnique Fdrale de LausanneInteractiveNon-interactivePausingReplayingMixed-InteractingSkippingExplicit-speedingSilentImplicit-speeding

Video Interaction ProfilesEcole Polytechnique Fdrale de Lausanne2.612.512.642.412.342.302.722.432.522.642.732.602.782.51

Implicit-Speeding ProfileEcole Polytechnique Fdrale de LausanneAn increase of 0.25 video speed results in an average decrease of perceived difficulty by 0.08,( = -0.08, 95% CI = [-0.10, -0.05], p < .0001)

Initial speed has shown negative effect on the perceived video difficulty

EXplicit-Speeding ProfileEcole Polytechnique Fdrale de LausanneAverage speed change = Average video speed - Initial video speed

Non-linear relationship (GAM) Effect saturates at 0.4The amount of average speed increase has a negative effect saturated at 0.4

Pausing ProfileEcole Polytechnique Fdrale de LausannePause Frequency = Number of pauses in a video sessionPause Median = Median Duration of all pauses in a video session

Positive effect (edf = 3.14, p < .0001)Steeper slopePause frequency:Positive effect (edf = 2.44, p < .0001)Flater slope saturated at 4.1 (1 min)Pause median:Pause frequency matters more than duration

Replaying ProfileEcole Polytechnique Fdrale de LausanneSeek Backward Frequency = Number of backward seeks in a video sessionReplayed Length = Video seconds that are re-watched

Negative effect (edf = 1.36, p < .0005)Seek backward frequency:Positive effect (edf = 2.20, p < .0001)Replayed length:Infrequent, large amount of re-watching indicates higher video difficulty

Skipping ProfileEcole Polytechnique Fdrale de LausanneSeek Forward Frequency = Number of seek forwards in a video sessionSkipped Video Length = Amount of video seconds skipped by forward seeks

Negative effect ( = -0.13, 95% CI = [-0.19, -0.06], p < .0005 )Seek forward frequency:Positive effect (edf = 1.56, p < .0005)Skipped video length:Infrequent , large skip suggests higher perceived video difficulty

Discussion & ConclusionLimitationsDiversified motivationDifficulty is measured per video not per contentAlternative ways to externalise difficultyContributionsReveal the empirical trendsInform the design of interventionEcole Polytechnique Fdrale de Lausanne

Questions?

Ecole Polytechnique Fdrale de Lausanne