Detecting Negative Emotion for Mixed Initiative Visual Analytics

CorrFeat: Correlation-based Feature Extraction Algorithm using Skin Conductance and Pupil Diameter for Emotion Recognition

“…Most previous work [14,23,31] solve this problem by manually designing handcrafted features across different sensors and fusing them at decision level [29], which is timeconsuming and often leads to low accuracies. Other works [1,28,33] segment or pad the signals to let them have fixed lengths and train the data with neural networks.…”

Observing and predicting knowledge worker stress, focus and awakeness in the wild

“…Following guidelines from previous studies [80,81,82] and following the preferences of extensive user piloting, we sent via text message a survey request to each participant two times per workday. Pilot participants preferred the usage of text message, in part, due to them being accessible and noticeable anywhere in the office.…”

Guidance in the human–machine analytics process

“…For engagement: Mixed-initiative interaction systems may leverage bio-sensing and other tools of affective computing to provide personalized and just-in-time guidance to mitigate or prevent frustration and increase engagement in an analysis process (Conati et al, 2013;Panwar and Collins, 2018;McDuff et al, 2012).…”

“…For example, Shao et al (2017) used eye-tracking records of which areas of a scatter plot matrix have been explored so far, to inform guidance. Panwar and Collins (2018) use GSR sensing and eye tracking to detect user frustration to provide just-in-time guidance.…”

“…We investigate the potential role of facilitation across lower-level task taxonomies, e.g., Brehmer and Munzner (2013) and more sophisticated models of visual analytics (Andrienko et al, 2018). Specific instantiations of guidance have been reported, for example, helpful interventions when eye-tracking indicates an analyst is exhibiting signs of confusion (Conati et al, 2013;Toker and Conati, 2014;Panwar and Collins, 2018) or when a logging system detects sub-optimal search strategies (Brown et al, 2014). There have also been investigations into the role of machine intelligence in revealing a potential bias in an analysis process (Wall et al, 2017a), by exposing the differences between the data a user has seen and the overall characteristics of the full dataset.…”