Explainable automated recognition of emotional states from canine facial expressions: the case of positive anticipation and frustration.

In animal research, automation of affective states recognition has so far mainly addressed pain in a few species. Emotional states remain uncharted territories, especially in dogs, due to the complexity of their facial morphology and expressions. This study contributes to fill this gap in two aspects. First, it is the first to address dog emotional states using a dataset obtained in a controlled experimental setting, including videos from (n = 29) Labrador Retrievers assumed to be in two experimentally induced emotional states: negative (frustration) and positive (anticipation). The dogs' facial expressions were measured using the Dogs Facial Action Coding System (DogFACS). Two different approaches are compared in relation to our aim: (1) a DogFACS-based approach with a two-step pipeline consisting of (i) a DogFACS variable detector and (ii) a positive/negative state Decision Tree classifier; (2) An approach using deep learning techniques with no intermediate representation. The approaches reach accuracy of above 71% and 89%, respectively, with the deep learning approach performing better. Secondly, this study is also the first to study explainability of AI models in the context of emotion in animals. The DogFACS-based approach provides decision trees, that is a mathematical representation which reflects previous findings by human experts in relation to certain facial expressions (DogFACS variables) being correlates of specific emotional states. The deep learning approach offers a different, visual form of explainability in the form of heatmaps reflecting regions of focus of the network's attention, which in some cases show focus clearly related to the nature of particular DogFACS variables. These heatmaps may hold the key to novel insights on the sensitivity of the network to nuanced pixel patterns reflecting information invisible to the human eye.

'Puppy Dog Eyes' Are Associated With Eye Movements, Not Communication.

The inner brow raiser is a muscle movement that increases the size of the orbital cavity, leading to the appearance of so-called 'puppy dog eyes'. In domestic dogs, this expression was suggested to be enhanced by artificial selection and to play an important role in the dog-human relationship. Production of the inner brow raiser has been shown to be sensitive to the attentive stance of a human, suggesting a possible communicative function. However, it has not yet been examined whether it is sensitive to human presence. In the current study, we aimed to test whether the inner brow raiser differs depending on the presence or absence of an observer. We used two versions of a paradigm in an equivalent experimental setting in which dogs were trained to expect a reward; however, the presence/absence of a person in the test apparatus was varied. In the social context, a human facing the dog delivered the reward; in the non-social context, reward delivery was automatized. If the inner brow raiser has a communicative function and dogs adjust its expression to an audience, we expect it to be shown more frequently in the social context (when facing a person in the apparatus) than in the non-social context (when facing the apparatus without a person inside). The frequency of the inner brow raiser differed between the two contexts, but contrary to the prediction, it was shown more frequently in the non-social context. We further demonstrate that the inner brow raiser is strongly associated with eye movements and occurs independently in only 6% of cases. This result challenges the hypothesis that the inner brow raiser has a communicative function in dog-human interactions and suggests a lower-level explanation for its production, namely an association with eye movements.

Fear expressions of dogs during New Year fireworks: a video analysis.

A high proportion of pet dogs show fear-related behavioural problems, with noise fears being most prevalent. Nonetheless, few studies have objectively evaluated fear expression in this species. Using owner-provided video recordings, we coded behavioural expressions of pet dogs during a real-life firework situation at New Year's Eve and compared them to behaviour of the same dogs on a different evening without fireworks (control condition), using Wilcoxon signed ranks tests. A backwards-directed ear position, measured at the base of the ear, was most strongly associated with the fireworks condition (effect size: Cohen's d = 0.69). Durations of locomotion (d = 0.54) and panting (d = 0.45) were also higher during fireworks than during the control condition. Vocalisations (d = 0.40), blinking (d = 0.37), and hiding (d = 0.37) were increased during fireworks, but this was not significant after sequential Bonferroni correction. This could possibly be attributed to the high inter-individual variability in the frequency of blinking and the majority of subjects not vocalising or hiding at all. Thus, individual differences must be taken into account when aiming to assess an individual's level of fear, as relevant measures may not be the same for all individuals. Firework exposure was not associated with an elevated rate of other so-called 'stress signals', lip licking and yawning.

Differences in facial expressions during positive anticipation and frustration in dogs awaiting a reward.

Facial expressions are considered sensitive indicators of emotional states in humans and many animals. Identifying facial indicators of emotion is a major challenge and little systematic research has been done in non-primate species. In dogs, such research is important not only to address fundamental and applied scientific questions but also for practical reasons, since many problem behaviours are assumed to have an emotional basis, e.g. aggression based on frustration. Frustration responses can occur in superficially similar contexts as the emotional state of positive anticipation. For instance, the anticipated delivery of a food reward may induce the state of positive anticipation, but over time, if the food is not delivered, this will be replaced by frustration. We examined dogs' facial expressions in contexts presumed to induce both positive anticipation and frustration, respectively, within a single controlled experimental setting. Using DogFACS, an anatomically-based method for coding facial expressions of dogs, we found that the "Ears adductor" action was more common in the positive condition and "Blink", "Lips part", "Jaw drop", "Nose lick", and "Ears flattener" were more common in the negative condition. This study demonstrates how differences in facial expression in emotionally ambiguous contexts may be used to help infer emotional states of different valence.

Spotlight on Assistance Dogs-Legislation, Welfare and Research.

Assistance dogs are a very diverse group of working dogs that are trained to assist humans with different types of disabilities in their daily lives. Despite these dogs' value for humankind, research on their welfare status, cognitive and behavioural capacities, selection criteria for the best fitting individuals, effective training and management practices, and genetic issues are so far lacking. This review highlights the need to address these topics and to promote progress in legal issues around assistance dogs. The topic of assistance dogs is approached comprehensively by outlining the current status of knowledge in three different dimensions: (1) the legal dimension, outlining important legal issues in the EU and Australia; (2) the welfare dimension; and (3) the dimension of research, covering assistance dog selection and training. For each of these three dimensions, we discuss potential approaches that can be implemented in the future in order to support assistance dog working performance, to protect the dogs' welfare, and to improve our knowledge about them. Additionally, there remain many legal issues, such as the presence of assistance dogs in public areas, the resolution of which would benefit both the assistance dog and the owner with disability.

Incentive motivation in pet dogs - preference for constant vs varied food rewards.

Recently, there has been a move towards positive reinforcement using food rewards in animal training. By definition, rewards function as reinforcers if they increase or maintain the frequency of behaviour that they follow. However, in operant conditioning tasks animals frequently show systematic changes in performance - in particular a reduction in responding over time. One suggested strategy to avoid such performance decrements is to provide a variety of food rewards, rather than the same food reward in all trials. The enhancement of appetitive behaviour and consumption by reward variation is referred to as 'variety effect'. We investigated whether dogs preferred a variable or a constant food reward in a concurrent two-choice test. Of 16 dogs, six subjects showed a significant preference for the varied food reward and six for the constant food reward, while four dogs exhibited no significant preference for either option. At the group level, there was a significant effect of block: preference for the varied food reward increased across six blocks of ten trials each. Thus, although some individuals may prefer a single, favourite food reward in the short term, introducing variation in reward types may maintain dogs' motivation in operant tasks over a longer time period.

Animal Morality: What It Means and Why It Matters.

It has been argued that some animals are moral subjects, that is, beings who are capable of behaving on the basis of moral motivations (Rowlands 2011, 2012, 2017). In this paper, we do not challenge this claim. Instead, we presuppose its plausibility in order to explore what ethical consequences follow from it. Using the capabilities approach (Nussbaum 2004, 2007), we argue that beings who are moral subjects are entitled to enjoy positive opportunities for the flourishing of their moral capabilities, and that the thwarting of these capabilities entails a harm that cannot be fully explained in terms of hedonistic welfare. We explore the implications of this idea for the assessment of current practices involving animals.