Artificial Intelligence (AI) is reshaping how we understand parent-child interactions—especially in early childhood development. But how accurate, scalable, and ethical are these tools?
Hosted by DSPI, the Global Parenting Initiative (GPI) webinar, Seeing Parenting Clearly? The challenges and potential of AI in early childhood assessment, brought together GPI researchers, who are using cutting-edge AI, to answer these questions by assessing early interactions and leading discussion around the promise – and perils – of this technology.
The researchers are part of the GPI project which identifies and measures playful parenting using machine learning, including two pioneering early career researchers – Jeremiah Ayock Ishaya and Irene Uwerikowe – who are both GPI Playful Parenting Scholars. Presentations focused on four core aspects of AI in early childhood assessment:
-
‘An AI–Neural Network Ensemble Model for Predicting Mother–Infant Synchrony - using machine learning to measure movement-based synchrony and flag dyads needing support’ – Dr Daniel Stamate, Goldsmiths, University of London; University of Manchester
-
‘Scalable AI Solutions for Analysing Parent-Child Interactions: A Feasibility Study - what it takes to build usable, scalable, ethical AI systems for developmental research’ – Jeremiah Ayock Ishaya, PhD student at Stellenbosch University
-
‘Evaluating Facial Emotion Recognition Models in Low-Resource Settings - can emotion recognition tools work accurately across diverse, underrepresented contexts?’ – Irene Uwerikowe, PhD student at Stellenbosch University
-
‘The Promise and Perils of AI in Global Parenting Assessments – a reflective session on AI's limitations, biases, and global potential’- Dr Caspar Addyman, Stellenbosch University
Key takeaways from the webinar:
-
AI can analyse parent–child videos with up to 90% accuracy, cutting down the time and cost of traditional assessments and opening doors to scalable, low-cost tools—especially valuable in under-resourced settings (Jeremiah Ayock Ishaya)
-
Off-the-shelf emotion recognition tools often misinterpret real-world interactions, particularly with babies and in diverse communities. More reliable, context-aware approaches are needed (Irene Uwerikowe)
“This event showed both the remarkable progress and the critical questions we must grapple with as AI enters the space of parenting assessment. It’s exciting to see GPI Playful Parenting Scholars at the forefront of this work—pushing boundaries while staying grounded in ethics, equity, and care for families globally,” commented Dr Inge Vallance, Head of Research at the GPI.
Hal Cooper, research manager on the GPI and doctoral student at DSPI, reflected: “My research focuses on how parents can support their children's early learning so today’s seminar was really useful to learn about the ways that we can assess parent child interactions and the quality of those interactions. As someone who has been the person doing human-led scoring, finding a way that we can give some of that labour essentially to AI is really promising for reducing the amount of time that we have to spend doing these kinds of assessments ourselves.”
Global Parenting Initiative (GPI) is a five-year collaboration of universities, foundations, and implementing partners with the aim of providing access to free, evidence-based, playful parenting support to every parent, everywhere, so that they are equipped with the knowledge and tools to help their children realise their learning potential and to prevent child sexual abuse, exploitation, and family violence.
Further information
