Developing a translational and computational approach to studying animal affect and welfare

Having good measures of emotion and mood in animals is highly important to science and society. There are billions of animals globally under the care of humans, for example in farms, in zoos, and in laboratories, and there is an increasing societal drive to ensure that these animals have a good quality of life. Additionally, research into the aetiology and treatment of mood disorders relies heavily on animal models. Despite this, existing measures have various limitations and there is scope for the development of new measures.

In recent years, computational approaches have been developed to study mood disorders in humans - these involve mathematically describing the cognitive processes underlying behaviour and how these differ in patients with mood disorders. This theory-driven field of computational psychiatry has been highly successful in furthering our understanding of emotion and mood in humans. My research has shown that computational analyses can also be valuable in better understanding the influence of emotion and mood on behaviour in rats, but a translational and computational approach has yet to be fully explored and exploited.

I propose assessing the validity of two phenomena that have been studied in computational psychiatry as potential novel measures of mood (and hence welfare, given that minimising experience of negative moods and maximising experience of positive moods is crucial to ensuring good welfare) in rats: Pavlovian interference and goal-directed vs. habitual learning. Pavlovian interference describes the influence of hard-wired tendencies on behaviour - it is much harder for humans to press a button, than to avoid pressing a button, to get a reward. Computational psychiatry research has shown that humans experiencing mood disorders are more susceptible to Pavlovian interference; it's even harder for patients with mood disorders to go against hard-wired tendencies. Goal-directed vs. habitual learning refers to the extent to which an individual makes decisions based on a complete understanding of the consequences of their actions, as opposed to making decisions based on which actions were successful in the past. Studies using computational methods have demonstrated that individuals with mood disorders are more prone to relying on the latter form of decision-making. Behavioural tasks have been developed to study both Pavlovian interference and goal-directed vs habitual learning in rats, but these methods have yet to be combined with computational approaches and manipulations to assess their potential as measures of welfare. I also propose developing behavioural tasks to study these phenomena in rodents using Raspberry Pi based equipment, so that they can more easily be scaled up or down for different species and studies can be conducted at lower cost hence aiding uptake of these methods, and ultimately also conducted within the home-cage to minimise disturbance to the animals.

Ultimately, this research has the potential to drive a shift towards more translatable and affordable methods to assess animal emotion, mood, and welfare, that may help us to improve the welfare of captive animals and develop more effective treatments for mood disorders.

The field of computational psychiatry has been highly successful in furthering our understanding of human mood disorders. However, translational computational psychiatry studies remain few and far between. This is surprising given the importance of the study of affect in non-human animals both to understand human mood disorders, and because affect is core to animal welfare; concerns about animal welfare ultimately stem from the belief that animals can subjectively experience negative affect. Computational analyses could play a pivotal role in developing improved measures of animal affect, furthering our understanding of the aetiology of mood disorders and the likely underlying processes involved, and answering functional and evolutionary questions regarding affect. Indeed, my research has highlighted the fruitfulness of applying a computational approach to better understand affect and decision-making in rats.

To further capitalise on the success of the field of computational psychiatry, and establish myself as an independent researcher at the forefront of the novel field of translational computational psychiatry, I propose assessing the validity of two phenomena that have been studied in human computational psychiatry as a measure of mood: Pavlovian interference and goal-directed vs. habitual learning. Both phenomena have been studied in animals by ethologists and neuroscientists, but not using computational approaches and not in the context of investigating affective states. I plan to fill this gap by carrying out novel studies that will involve training rats on two different tasks: an orthogonalized go/no-go task and a contingency degradation task, with testing conducted while rats undergo an affect manipulation. Resulting data will be analysed using computational modelling within a reinforcement learning framework. Then, to aid uptake of these measures, I will develop behavioural tasks using Raspberry Pi based equipment to study these phenomena.