Empathy is thought to be reduced in videoconference (VC) compared to face-to-face (F2F) therapy. To empirically test this possibility, therapist and client ratings of empathy and its correlates were examined in VC and F2F sessions. Forty-eight students in clinical training programs were randomly assigned to the role of ‘therapist’ or ‘client’, forming 24 dyads. Each dyad completed a simulated clinical session in F2F and in VC and reported on empathy and telepresence. Both therapist and client ratings of empathy were significantly lower in VC compared to F2F. Therapist empathy correlated with telepresence, particularly its physical subcomponent, and therapeutic presence, but these associations were not found for clients. The relationship between empathy and telepresence was investigated further in a real-world study. Twelve clients met with clinical psychology trainees over VC and reported on empathy and telepresence. The associations between therapist empathy, total telepresence, and the physical component of telepresence were replicated. Again, client ratings of empathy and telepresence did not correlate. These results provide the first empirical evidence that empathy is reduced in VC teletherapy compared to F2F and highlight the connection between the degree of empathy felt by therapists and their sense of telepresence during teletherapy.

Different healthcare services can be provided to patients using virtual agent platforms. Various nonverbal markers of empathy (e.g., direct gaze) have been integrated into the design of these agents. However, how distinct markers interact to influence the empathy perceived from these agents remains to be investigated. This study examines how agent variations in gaze direction, facial expression, trunk lean, and gender affect the perception of empathy. Participants watched video clips of virtual healthcare professionals showing 12 behavior combinations (as seen through the eyes of patients in pain) and rated perceived empathy. A pain face (congruent with that of patients in pain) was perceived as more empathic than a neutral face, regardless of gaze direction, trunk lean, and gender. Perceived empathy from a neutral face was higher for female agents or with a direct gaze and a forward-or backward-trunk lean. These results highlight the primary role of facial expressions in empathic human-agent interactions.

Virtual agents are increasingly used to study the perception of affective states. Their facial configurations are mostly prototypical and derived from visual and motor modalities of psychological theories and motion capture. However, facial configurations perceived daily often stem from complex affective states and reach beyond prototypical visual and motor representations. Interactive virtual agents are a promising tool to unveil users' internal (i.e., conceptual) representations of their complex affective states. "is paper presents a new method to study facial configurations that combines customizable faces of interactive virtual agents and multivariate analysis procedure. The results of a proof-of-concept project on affective states mixing pain and emotions show the usefulness of this approach for research in affective computing.

The lack of high-quality in-game player data has been a challenge in affective computing and player experience research, as the currently available datasets do not fully capture the multimodal nature of the player-game interaction. The following paper addresses this issue by introducing a multimodal collection of video game player data: the FUNii Database. The database contains in-laboratory physiological, behavioral and subjective data from 190 players acquired while they played Ubisoft's Assassin's Creed: Unity and Assassin's Creed: Syndicate. The database also includes demographic data from the players and Fun Trace ratings, a unique tool for assessing player experience on a continuous scale following a game session. The database offers opportunities to study physio-behavioral responses in a gaming context, player profiling, player experience modeling or adaptation strategies for affective adaptive games.

Affective computing, with its potential to enhance human-computer interaction, is experiencing an expansion of its use in many areas such as health care and the gaming industry. One obstacle to its widespread adoption can be the high cost requirement for biofeedback. Indeed, typical laboratory setups are often expensive which makes them out of reach for many. This paper explores lower-cost alternatives to expensive laboratory solutions. Data from several recent studies totaling over 200 hours of physiological recordings are leveraged to compare high-end solutions to a lower cost one. Heart rate, electrodermal activity, facial action units, head movement, and eye movement - five of the most used bio-behavioural signals - have their respective higher and lower cost sensors compared. The resulting comparison illustrates that lower-cost solutions are not drop-in replacements. While a correlation of 0.62 between electrodermal activity readings was found, notable differences between reported heart rate readings over small timescales were also observed. Head tracking recordings shared similarity (0.51), but eye tracking did not (0.18). As for facial action units recognition, only those linked to smiling had significant correlation (around 0.48). These results should broaden the range of contexts in which biofeedback could be exploited. This aim may be fulfilled by informing the reader of the extent of lower cost solution applications.