C3RI Lunchtime Research Seminar - Physiological Correlates to Online Behaviour with Fatima Isiaka
Event contact Rachel Finch
Speaker: Fatima Isiaka (C3RI PhD Student – Computer Science) – hosted by Dr Kassim Mwitondi
Title: Physiological Correlates to Online Behaviour
The user experience of web application and browsing is mostly studied via tangible measures such as task completion time, surveys and comprehensive test, which are based on subjective approach. The main purpose of our study is to use physiological measures such as skin conductance response, pupil dilation, and skin temperature, as tertiary indicators of emotional responses (stress status). For this we propose a novel method for detecting naturally arising structures in human physiological data attributes during online activities. Fifteen human physiological attributes were captured singularly and in combination using an integrated novel tool (PHYCOB I) for identifying patterns in the data.
Captured attributes form a set of input predictors to PHYCOB I and performance comparisons are made with conventional predictive methods such as Neural Network and Logistic Regression. The Forward Search algorithm (Atkinson, 1984) is used alongside other methods to validate the results. Results show that PHYCOB I outperforms the conventional models in terms of both accuracy and reliability. These results are confirmed by multiple runs of the Forward Search algorithm and Principle Component Analysis on the data and cross-model comparisons which show that the average recoverable natural structures for the three models with respect to accuracy and reliability are more consistent within the PHYCOB I environment than with the other two.
The key advantages of our proposed tool include its resistance to over-fitting a feature achieved through regularisation by integrating physiological parameters with eye movement attributes of users. It also provides an automated way of assessing human stress levels while dealing with specific web contents. This is an important achievement in that it can be able to predict what contents on a webpage curse stress induced emotion in users during interaction.