Digitally-enhanced dog behavioral testing

Abstract

Behavioral traits in dogs are assessed for a wide range of purposes such as determining selection for breeding, chance of being adopted or prediction of working aptitude. Most methods for assessing behavioral traits are questionnaire or observation-based, requiring significant amounts of time, effort and expertise. In addition, these methods might be also susceptible to subjectivity and bias, negatively impacting their reliability. In this study, we proposed an automated computational approach that may provide a more objective, robust and resource-efficient alternative to current solutions. Using part of a ‘Stranger Test’ protocol, we tested n = 53 dogs for their response to the presence and neutral actions of a stranger. Dog coping styles were scored by three dog behavior experts. Moreover, data were collected from their owners/trainers using the Canine Behavioral Assessment and Research Questionnaire (C-BARQ). An unsupervised clustering of the dogs’ trajectories revealed two main clusters showing a significant difference in the stranger-directed fear C-BARQ category, as well as a good separation between (sufficiently) relaxed dogs and dogs with excessive behaviors towards strangers based on expert scoring. Based on the clustering, we obtained a machine learning classifier for expert scoring of coping styles towards strangers, which reached an accuracy of 78%. We also obtained a regression model predicting C-BARQ scores with varying performance, the best being Owner-Directed Aggression (with a mean average error of 0.108) and Excitability (with a mean square error of 0.032). This case study demonstrates a novel paradigm of ‘machine-based’ dog behavioral assessment, highlighting the value and great promise of AI in this context.

Automation in canine science: enhancing human capabilities and overcoming adoption barriers

“…The most common representation of behavior used in the majority of these works is by tracking the dog's body (either in two or three dimensions). Bleuer-Elsner et al ( 41 ), Karl et al ( 36 ), Byosiere et al ( 50 ), Fux et al ( 40 ), Menaker et al ( 51 ), Farhat et al ( 46 ), Tsiourti et al ( 47 ), and Watanangura et al ( 42 ) use a convolutional neural network for object detection, producing a time series representing a trajectory of the dog from an above view, see examples in Figure 1 . Similarly, Völter et al ( 37 , 38 ) and Ren et al ( 39 ) use a convolutional neural network on multiple cameras producing a 3D time series representation of multiple key points on the dogs' bodies.…”

“…Some of them extract from these trajectories some high-level meaningful features, such as average speed, residence in areas of interest, distance from and interaction with certain objects or people, studies that used time series of key points additionally extracted specific limb movement such as head angle, tail angle, velocity and amplitude. Two studies ( 46 , 48 ) used deep learning to extract features automatically from the computational representation of the behavior.…”

“…Farhat et al ( 46 ) investigated a computational approach to assess behavioral traits. The participated dogs N=53 were recorded in a behavioral test reacting to the presence of a strange, their coping styles were categorized into neutral, negative (reacting away from the stressor), and positive (reacting toward the stressor) reactions.…”

“…Specifically, the narrow scope of these hand-picked features might not provide the comprehensive data necessary to fully address the research questions at hand. Using alternative approaches in machine learning, such as deep learning, automated feature selection and unsupervised learning may be beneficial in this domain [see, e.g., the approaches used in ( 46 , 48 )] and should be further explored.…”

“…The data collection instrument was designed as a questionnaire incorporating both Likert-like scale and open-ended questions, designed to gather comprehensive insights into participants' experiences and perceptions. The formulation of questions drew from the authors prior involvement in collaborations with animal researcher ( 36 , 46 , 50 ). Furthermore, the questionnaire underwent a pilot phase, involving a behavioral veterinarian who had experience with an automated tool, and an animal behavior researcher who had not previously utilized an automated analysis tool.…”