.png)
Affective Computing
Affective Computing (AC) is a transdisciplinary area of research that looks at ways for artificial systems to detect, recognise, and interpret human users’ feelings (“affect”) and emotional signals. Extended versions of AC seek to use this affective and emotional information to adapt the functionality of these systems: adapting its services, assistance, and interactions in a way that responds to users’ emotional needs, much like humans do with other humans. This affective data can be physiological (e.g. EEG, EDA, ECG, facial expressions), behavioral (e.g. bodily gestures, body language, task-based behavioral dynamics) and can also be complemented with subjective assessments (ethnography, self-reports).
How we work with Affective Computing
At DICE Lab, we take both a theoretical and applied approach to study a range of multimodal affective detection approaches. First, we look to understand which modalities, signals and/or data can accurately capture a user’s affective state, particularly in niche demographics (i.e. elderly populations, and/or populations with cognitive impairments) or in specific contexts (i.e. being a passenger in an automated vehicle, or interacting with a social robot). Secondly, we seek to understand the relevance of these affective signals in the real-time adaptation of these technologies (socially assistive robots, autonomous vehicles) in a way that best serves positive outcomes associated with the human-technology interaction, such as motivation to interact (with cognitive training or a robot companion) or trust in automated systems (such as an automated vehicle).
Current Projects
Socially Affective Robots for Digitized Cognitive Interventions (SARP-DCI)
This project looks to investigate the effects of a social robot partner, which adapts its behaviours and interactions based on a human partner’s affective and emotional states, on the long-term viability (acceptance, adherence, motivation, efficacy) of digitised cognitive training therapy. In combination with interactions with a social robot and a cognitive training task, this project seeks to find out: (1) the optimal combination of multimodal signals that can best capture a user’s affective/emotional states, particularly when interacting with these tools, (2) the effects of real-time adaptation of these technologies (robot interactions, characteristics of training tasks) based on these affective states on various behavioural and performance measures and (3) the viability of this approach in pre-clinical populations with Mild Cognitive Impairments.