Moving the objective measurement of child emotions and behaviours from the lab to real world settings

Lead Supervisor
Dr Johnny Downs
Clinical Senior Lecturer
Child and Adolescent Psychiatry, Institute of Psychiatry, Psychology and Neuroscience (IoPPN), King’s College London

Dr Petr Slovak (2nd)
Lecturer in Computer Science, King’s College London
Dr Oya Celiktutan (3rd)
Lecturer in Engineering (Robotics), King’s College London

Professor Emily Jones (Birkbeck), Professor Emily Simonoff (King’s College London), Professor Max Van Kleek (Oxford)

Project Details

Child Psychiatry lacks robust outcome measures. Currently clinical trials are evaluated by subjective accounts of symptoms or by clinical global ratings. Gold standard measures consist of subjective clinical judgements during brief face-to-face assessments with multiple informants. Double blinded trials are nearly impossible to implement within psychological interventions trials, and so response and observer biases play a significant impact on trial quality.  Furthermore, many psychological trial outcomes are based on globally observed behaviours, and do not detect more subtle changes that may be the antecedents of longer-term, meaningful clinical change: context-dependent behaviour is the norm, with poor to moderate agreement between informants reporting on behaviour in different settings.  Hence child mental health research and clinical practice has a pressing need for objective, accurate and holistic outcomes, that can be evaluated in everyday settings. 

Remotely collected video, auditory and sensor-based assessments (collectively called Remote Monitoring Technologies, RMT) conducted within the “usual settings” of home or school, have the potential to provide more sensitive and unbiased objective measurements that index meaningful change. For example, analysis of audio recordings of parent-child interaction (on the level of content or paralinguistic cues) could provide information about attachment or (changes in) parenting style; movement detectors can inform about the number, rate and timing of overactive or repetitive behaviours and link to other measures such as eye gaze and verbal output. In addition, objective measures allow the inclusion of behaviours/symptoms not directly observable, such as heart rate variability and galvanic skin response. However, existing platforms are expensive, difficult to set up in participants homes (if not confined to lab studies only) and would likely struggle with users acceptance and adherence rates. 

This project is enabled by a novel data capture infrastructure, developed by one of the supervisors (Slovak) in collaboration with the Personal Data and Privacy group (Prof Max Van Kleek) at Oxford University CS. Building on cheap, off-the-shelf Internet of Things components (Raspberry PIs), the platform creates a modular mesh of small devices (e.g. 4x4x6 inches) that can be connected to a range of sensors (audio, video, movement), be simply distributed in-situ without complicated installation (e.g., participants houses), extended by wearable sensors, and collect data in privacy preserving and user-empowering way (e.g., raw data stored locally and only selected indicators transmitted elsewhere). However, before we can move to real-world deployment of such RMT platform in the context of child mental health research and clinical practice, a number of technical, psychological, and socio-technical research challenges need to be addressed within a controlled, lab-based environment to identify plausible indicators and establish their validity.

This PhD will thus first undertake a series of investigations into the applicability and feasibility of using the above platform to accurately capture objective longitudinal data collected via multimodal sensors including audio, visual, wearable accelerometer and physiological sensors. These will be situated in the context of child and family studies (see Data source 1 below). A number of cross comparison studies will be made between existing “gold standard” objective capture methods currently used in child and family research laboratory-based studies and the data collection modules developed here. A second, parallel strand of work, will then draw on established human-computer interactive methods and user-centred approaches to examine how such IoT technologies, including avatar/ robot driven approaches can be deployed in the home. The candidate will explore design concepts on how to maintain the important balance between providing accurate collection of psychological/physiological outcomes, providing robot driven assistance to support engagement, and sustain long-term acceptability in-situ.  Finally, the PhD candidate will draw on their findings from their lab based cross-comparison studies and their HCI acceptability to identify design patterns which could tackle ethical and policy implications of home based IoT remote monitoring to provide a road-map for privacy-preserving IoT technologies in sensitive contexts. 


Data source 1: The PhD student will have access to cross-comparison data from a number of KCL based child and family studies (e.g. SuperStaars Study, AIMS-2) using the child and family research testing suite at the IOPPN. These studies all have existing ethically-approved protocols for using video / audio monitoring technology, and currently are able to introduce additional RMT to complement their existing lab monitoring. Data collection has begun and recruitment will continue over the course of the PhD. 

Data source 2: Using qualitative research methods within an HCI framework, the researcher will collect data from co-facilitated workshops and interviews with multiple stakeholders, including; recruited families, academics, clinicians and educational professionals; to gather insights and data, through inclusive design methodology and co-creation techniques.


Translational Psychiatry, Remote Monitoring Technologies, Human Computer Interaction, Multimodal data, Child and Adolescent Mental Health