Digitally-enhanced dog behavioral testing.

Area of Science:

• Animal Behavior
• Artificial Intelligence
• Computational Biology

Background:

• Assessing dog behavioral traits is crucial for breeding, adoption, and working aptitude.
• Current methods rely on subjective and time-intensive questionnaires or expert observations.
• These traditional methods can be prone to bias and lack reliability.

Purpose of the Study:

• To develop an automated, objective, and resource-efficient computational approach for dog behavior assessment.
• To provide a reliable alternative to subjective and labor-intensive traditional methods.
• To explore the application of machine learning in analyzing dog behavior patterns.

Main Methods:

• Utilized a 'Stranger Test' protocol with 53 dogs responding to a stranger's presence and actions.
• Collected expert scoring of dog coping styles and owner data via the Canine Behavioral Assessment and Research Questionnaire (C-BARQ).
• Applied unsupervised clustering to dog movement trajectories and developed machine learning classifiers and regression models.

Main Results:

• Unsupervised clustering identified two distinct dog groups with significant differences in stranger-directed fear (C-BARQ) and expert-scored coping styles.
• A machine learning classifier achieved 78% accuracy in predicting expert scoring of coping styles towards strangers.
• Regression models demonstrated predictive capabilities for C-BARQ scores, notably for Owner-Directed Aggression and Excitability.

Conclusions:

• This study presents a novel 'machine-based' paradigm for dog behavioral assessment.
• The findings highlight the significant potential of Artificial Intelligence (AI) in creating objective and efficient behavioral evaluations for dogs.
• The developed computational approach offers a promising, reliable, and scalable solution for canine behavior analysis.