Animal emotion and welfare: a decision-making and computational approach
Lead Research Organisation:
University of Bristol
Department Name: Clinical Veterinary Science
Abstract
An animal's welfare is strongly dependent on its emotional (affective) state. These states can be
operationally defined as being elicited by rewarding and punishing events, allowing empirical study
even though the conscious experience of such states in other species remains unknown. Long-term
'moods' (e.g. depressed states) are particularly important determinants of animal welfare and may
also play a key role in guiding decision-making by biasing an individual's expectations of positive or
negative outcomes, especially in ambiguous situations (Mendl et al. 2010. Proc. Roy. Soc. B. 277, 2895-
2904). We have developed a 'judgement bias' (JB) assay of decision-making under ambiguity to test
this idea and to provide an objective 'cognitive' measure of animal affect (Harding et al. 2004. Nature
427, 312). There are now over 100 studies using this assay, many supporting the hypothesis that
individuals in a positive affective state behave as if anticipating positive outcomes under ambiguity,
and vice versa for those in a negative state. However, there are also null and opposite results. One
potentially important reason for these is that affective states have a variety of different influences on
decision-making, leading to a variety of different responses.
This project will explore exactly how affective states alter decision-making, by combining operant
studies of laboratory rodent decision-making behaviour with computational modelling of the resulting
data. Using our JB assay and other operant tasks, data will be generated that can be analysed using
conventional statistical approaches, and also modelled computationally to identify underlying
parameters, such as response bias and reward sensitivity, that influence decision-making.
Computational modelling will allow specific hypotheses to be tested, for example that short-term
negative states generate negative biases about the outcomes of ambiguity but also increase reward
sensitivity, and hence to clarify findings in the literature and advance theory on the relationship
between affect and decision-making. There will also be opportunity to develop theoretical
computational models to investigate our predictions, including that experience of rewards and
punishments in the environment generates adaptive decision-making profiles.
The student will receive training in animal learning and behaviour, perceptual and affective
psychology, and computational theory and modelling from a supervisory team with expertise in all
areas. They will learn to design perceptual discrimination tasks, to programme and use automated
operant equipment, to implement computational, statistical, and trial-by-trial analysis of complex
datasets, and to build theoretical computational models. Such skills will be invaluable within the
increasingly mathematical context of modern biology.
operationally defined as being elicited by rewarding and punishing events, allowing empirical study
even though the conscious experience of such states in other species remains unknown. Long-term
'moods' (e.g. depressed states) are particularly important determinants of animal welfare and may
also play a key role in guiding decision-making by biasing an individual's expectations of positive or
negative outcomes, especially in ambiguous situations (Mendl et al. 2010. Proc. Roy. Soc. B. 277, 2895-
2904). We have developed a 'judgement bias' (JB) assay of decision-making under ambiguity to test
this idea and to provide an objective 'cognitive' measure of animal affect (Harding et al. 2004. Nature
427, 312). There are now over 100 studies using this assay, many supporting the hypothesis that
individuals in a positive affective state behave as if anticipating positive outcomes under ambiguity,
and vice versa for those in a negative state. However, there are also null and opposite results. One
potentially important reason for these is that affective states have a variety of different influences on
decision-making, leading to a variety of different responses.
This project will explore exactly how affective states alter decision-making, by combining operant
studies of laboratory rodent decision-making behaviour with computational modelling of the resulting
data. Using our JB assay and other operant tasks, data will be generated that can be analysed using
conventional statistical approaches, and also modelled computationally to identify underlying
parameters, such as response bias and reward sensitivity, that influence decision-making.
Computational modelling will allow specific hypotheses to be tested, for example that short-term
negative states generate negative biases about the outcomes of ambiguity but also increase reward
sensitivity, and hence to clarify findings in the literature and advance theory on the relationship
between affect and decision-making. There will also be opportunity to develop theoretical
computational models to investigate our predictions, including that experience of rewards and
punishments in the environment generates adaptive decision-making profiles.
The student will receive training in animal learning and behaviour, perceptual and affective
psychology, and computational theory and modelling from a supervisory team with expertise in all
areas. They will learn to design perceptual discrimination tasks, to programme and use automated
operant equipment, to implement computational, statistical, and trial-by-trial analysis of complex
datasets, and to build theoretical computational models. Such skills will be invaluable within the
increasingly mathematical context of modern biology.
People |
ORCID iD |
| Michael Mendl (Primary Supervisor) | |
| Molly Davidson (Student) |
| Description | We ran an online experiment to look at the effects of negative trait affect on decision-making under ambiguity. A total of 540 participants (final sample of 509; 273 female, aged 18-73, median age = 36) were recruited to an experiment where they completed questionnaires assessing trait anxiety, depression and pessimism, as well as taking part in a judgement bias task. The judgement bias task is a cognitive task designed to look at decision-making under ambiguity. This study was preregistered here: https://osf.io/shvfr, publication in progress. Initial findings were presented at the 'Cognitive affective biases: from mechanisms to disease symptoms' conference in Krakow, Poland in October 2022. New knowledge: We found that higher trait anxiety and depression scores were significantly associated with a lower proportion of 'optimistic' responses to ambiguous stimuli, i.e. a 'pessimistic' judgement bias. Interestingly, this was not true for trait pessimism scores. We also found significant effects of age and sex on judgement bias; males were significantly more 'optimistic' in the task than females and 'optimistic' judgement bias responses significantly decreased with age. In conclusion, self-reported trait anxiety and depression, but not trait pessimism, are associated with 'pessimistic' decision-making under ambiguity. New or improved research methods: We developed a novel version of the judgement bias task that can be completed by participants online. This allowed for the collection of a significantly larger sample size than would have been possible if the data was collected in-person. It is also likely that online experiments allow for a more diverse sample than traditional lab experiments. This task has already been adapted for use by our collaborators (Danny Longman et al.) and is being used for a further experiment within our group. |
| Exploitation Route | Alterations of the online judgement bias task we developed have already been used by our collaborators (Danny Longman et al.) as well as within our group. Once published, we plan to share our task openly on Gorilla (https://gorilla.sc/) for use by others. |
| Sectors | Other |
| Description | Collaboration with Danny Longman et al. The effects of environmental manipulations on human judgement bias. |
| Organisation | Loughborough University |
| Department | School of Sport, Exercise and Health Sciences |
| Country | United Kingdom |
| Sector | Academic/University |
| PI Contribution | The collaborators were interested in using a judgement bias task as a cognitive measure of affective state. Judgement bias would be one of many psychological and physiological measures of stress collected as part of a larger experiment looking at the effects of natural versus urban environments on human stress. We adapted a judgement bias task we had previously developed to meet their requirements. This task was shared along with training on how to collect the data. We have processed and begun to analyse the judgement bias data collected by the collaborators. We have also provided expertise on interpreting the findings. |
| Collaborator Contribution | Our collaborators designed the experiment, collected all data, shared data with us, and will be analysing the judgement bias data in context with all the other psychological and physiological stress variables they measured. |
| Impact | Data has been collected and analysis of the data has begun, however it is too early to report on any specific outcomes. This collaboration is multi-disciplinary combining psychology and human physiology. |
| Start Year | 2022 |
| Description | Collaboration with Danny Longman et al. The effects of environmental manipulations on human judgement bias. |
| Organisation | Loughborough University |
| Department | School of Sport, Exercise and Health Sciences |
| Country | United Kingdom |
| Sector | Academic/University |
| PI Contribution | The collaborators were interested in using a judgement bias task as a cognitive measure of affective state. Judgement bias would be one of many psychological and physiological measures of stress collected as part of a larger experiment looking at the effects of natural versus urban environments on human stress. We adapted a judgement bias task we had previously developed to meet their requirements. This task was shared along with training on how to collect the data. We have processed and begun to analyse the judgement bias data collected by the collaborators. We have also provided expertise on interpreting the findings. |
| Collaborator Contribution | Our collaborators designed the experiment, collected all data, shared data with us, and will be analysing the judgement bias data in context with all the other psychological and physiological stress variables they measured. |
| Impact | Data has been collected and analysis of the data has begun, however it is too early to report on any specific outcomes. This collaboration is multi-disciplinary combining psychology and human physiology. |
| Start Year | 2022 |
| Description | SWBio stand at Bristol Neuroscience Festival 2023 |
| Form Of Engagement Activity | Participation in an activity, workshop or similar |
| Part Of Official Scheme? | No |
| Geographic Reach | Regional |
| Primary Audience | Public/other audiences |
| Results and Impact | The Bristol Neuroscience Festival is a large, three-day public engagement event celebrating all things neuroscience. This year, the event was attended by around 3,000 visitors. The first day was open to primary schools, the second to secondary schools and the third to the public. I, along with seven other SWBio PhD students, ran a stand showcasing the breadth of neuroscience research within our PhD programme. This included a poster display of 'Postcards from Around the Brain', with each postcard giving information about a student's research focus. We also ran three interactive games focused on reward, motivation and cognitive biases. The cognitive biases game was a shortened, demo version of a judgement bias task we developed for a previous experiment. By taking part on a laptop, visitors were able to see what a real human psychology task looks like. They were also able to learn about the theory behind the task and why we use these kinds of experiments, both in humans and animals. |
| Year(s) Of Engagement Activity | 2023 |
| URL | http://www.bristol.ac.uk/neuroscience/news/2023/bnf-2023.html |