An Examination of Contemporary and Computational Techniques for Enhancing the Analysis of Parent-Infant Interactions.

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

Micro-coding involves manually categorising behaviours at a high temporal resolution. When applied to parent-infant interactions, micro-coding can provide vital insights into various aspects of behaviour, mental health, and infant outcomes. However, while rigorous, manual coding is a time-consuming process, rendering it inefficient when processing high volumes of data. This research therefore explores contemporary and computational methods for enhancing the scalability, efficiency, and breadth of parent-infant interaction analysis.

Based on data from two cohort studies, this research analyses videos of parent-infant interactions, predominantly captured using wearable headcams during interactions within the home setting. The first project investigates the efficacy of assessing short segments of behaviour as opposed to full observations, termed ‘thin-slice sampling’, as a viable alternative to alleviate the manual coding burden. The study quantifies the validity of this approach for a range of behaviours and measurements, providing guidance on which slices to target for coding. The second project evaluates the use of an automated facial coding software – FaceReader – as another alternative to manual coding. The study advances understanding of the validity of such software for analysing naturalistic interactions, and offers eight recommendations for future researchers in the field. The final project investigates the association between automated facial coding and depression in parents, effectively addressing the need to assess the practical applicability of this method in the realm of mental health.

This work discusses key themes pertaining to the benefits of contemporary analysis techniques, while also acknowledging the unique biases and challenges linked to these methods. This research concludes that thin-slice sampling and automated facial coding are efficient and valid approaches for alleviating the coding burden, and advocates for an integrated approach consisting of both traditional and contemporary techniques in future research.
Date of Award19 Mar 2024
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorIan Nabney (Supervisor), Helen E Bould (Supervisor) & Rebecca Pearson (Supervisor)

Keywords

  • parent-infant interactions
  • behaviour
  • ALSPAC
  • thin-slice sampling
  • facial expressions
  • automated facial coding
  • depression

Cite this

'