Poster #13 - Evaluating Objective Assessment of Auditory Household Chaos as a Predictor of Physiological Arousal in Infants

• In Event: Poster Session 09

Abstract

Household chaos is a noted risk factor for child development, affecting both brain and behavior development (Deater‐Deckard et. al, 2009). Most studies typically measure chaos as a stable trait using self-reported surveys. However, chaos is a dynamic feature of the environment with dynamic effects on children's behavior. For example, high levels of noise caused increased heart rate in infants (Bremmer, Byers, & Kiehl, 2003) and children (Evan et. al, 2001). Additionally, the volume of sounds overheard by infants was found to be positively related to their autonomic arousal (Wass et. al, 2019).

While volume is one aspect of chaos, analyses conducted in our lab determined that volume is only moderately predictive (r = 0.42) of gold-standard human-annotated auditory chaos. Thus, in the current study, we test whether a gold-standard auditory chaos measure, automatically detected using a machine learning model developed in our lab, shows stronger associations with physiological arousal in infants than simple volume measures. Specifically, we test the hypothesis that within-person, higher levels of auditory chaos will be associated with increases in physiological arousal in infants, and that these relations will be stronger than between volume and arousal. Additionally, we hypothesize that infants high in negative affectivity and low in orienting will show stronger relations between chaos and arousal. Finally, we use our model to generate descriptive data of within- and between-person differences in exposure to auditory chaos over the course of a day in a cross-sectional sample of infants.

N=87 families collected 72h+ everyday audio recordings and heart rate data using infant-worn LENA and Movisens respectively. We will analyze a subset of n=30 infants (R = 0.87-7.10mo, 60% female) with synchronized audio and motion datastreams. Auditory Chaos Model. We developed algorithms to automatically classify 5s segments of infant-worn audio into four levels of auditory chaos (none, low, medium, and high) (Authors, Submitted). Our model achieved an F1 score of 0.587 (Precision: 0.644, Recall: 0.585), matching the accuracy of real-world cry and speech models used by the developmental community (Yao et. al, 2022; Cristia et. al, 2021). Volume. To serve as a comparison with our chaos model, we compute the average volume for all 5s audio bins. Infant physiological arousal. We calculate the average infant heart rate using the moving window technique for every 5s bin. Infant temperament. Every family filled out an IBQ-RF-VSF (Gartstein & Rothbart, 2013). Analyses include super factors, orienting and negative affectivity to examine possible moderation of chaos-arousal relations by temperament.

We will use time series correlations to assess the relations between auditory chaos, volume, and infant heart rate for the whole sample and groups high/low in orienting and negative affectivity. Infant crying algorithms developed in our lab (Authors, 2022) will be used to automatically detect and remove crying to evaluate if the relations change. These analyses will serve as an important step in validating our chaos model. If validated, this model could be valuable in answering many relevant questions in developmental psychology.

Authors

• Priyanka Khante, University of Texas at Austin
  Presenting Author

• Brandon Thint

• Kaya de Barbaro, University of Texas at Austin