With the emergent focus on human-centered artificial intelligence (HCAI), research is required to understand the humanistic aspects of AI design, identify the mechanisms through which user concerns may be alleviated, thereby positively influencing AI adoption. To fill this void, we introduce “Situation Awareness” (SA) as a conceptual framework for considering human-AI interaction (HAII). We argue that SA is an appropriate and valuable theoretical lens through which to decompose and view HAII as hierarchical layers that allow for closer inquiry and discovery. Furthermore, we illustrate why the SA perspective is particularly relevant to the current need to understand HCAI by identifying three tensions inherent in AI design and explaining how an SA-oriented approach may help alleviate these tensions. We posit that users’ enactment of SA will mitigate some negative impacts of AI systems on user experience, improve human agency during AI system use, and promote more efficient and effective in-situ decision-making.

The accurate classification of psychophysiological data is an important determinant of the quality when interacting with a physiological computing system. Previous research has focused on classification accuracy of psychophysiological data in purely mathematical terms but little is known about how accuracy metrics relate to users' perceptions of accuracy during real-time interaction. A group of 14 participants watched a series of movie trailers and were asked to subjectively indicate their level of interest in a binary high/low fashion. Psychophysiological data (EEG, ECG and SCL) were used to create a binary classification of interest via a Support Vector Machine (SVM) algorithm. After a period of training, participants received real-time feedback from the classification algorithm and perceptions of accuracy were assessed. The purpose of the study was to compare mathematical classification accuracy with the perceived accuracy of the system as experienced by the users. Results indicated that perceived accuracy was subject to a number of psychological biases resulting from expectations, entrainment and development of trust. The F1 score was generally a significant predictor of perceived accuracy.

This article presents a psychophysiological construct of interest as a knowledge emotion and illustrates the importance of interest detection in a cultural heritage context. The objective of this work is to measure and classify psychophysiological reactivity in response to cultural heritage material presented as visual and audio. We present a data processing and classification framework for the classification of interest. Two studies are reported, adopting a subject-dependent approach to classify psychophysiological signals using mobile physiological sensors and the support vector machine learning algorithm. The results show that it is possible to reliably infer a state of interest from cultural heritage material using psychophysiological feature data and a machine learning approach, informing future work for the development of a real-time physiological computing system for use within an adaptive cultural heritage experience designed to adapt the provision of information to sustain the interest of the visitor.