Dogs play an increasingly important role in human health. Their acute sense of small is harnessed as they are trained to detect a range of medical conditions (e.g. cancer; type 1 diabetes1) and targets ranging from elicit contraband (drugs; explosives2) to invasive species. These roles require the dog to be sensitive (detect a large proportion of targets), but also specific (avoiding false alerts). The relative importance of sensitivity and specificity differs depending upon the specific task and the substance that the dog is trained to detect (e.g. for covid detection sensitivity is paramount). The detection task is essentially equivalent to making a decision about ambiguous stimuli and we have developed a now widely used ‘cognitive bias’ 3 task that quantifies responses to ambiguity in a range of species, including dogs4. Our initial studies with medical detection dogs show that dogs with a more negative cognitive bias exhibit higher specificity when trained. Such dogs may be ideally suited to detection roles such as finding explosives (for example when false positives could lead to evacuation of an airport). In this study, we will examine the potential of cognitive bias as a selection tool for detection dogs for medical and related purposes.
Aims and objectives
- Compare dogs in a range of detection roles for performance on a cognitive bias test.
- Determine relative importance of sensitivity and specificity for detection of different targets.
- Investigate whether cognitive bias tests can be used to predict variation in sensitivity and specificity within different detection contexts.
- Quantify the potential value of a cognitive bias selection test for a range of dog detection roles.
- Trial a cognitive bias test on naïve dogs ahead of training for a simple discrimination test.
- Explore the extent to which cognitive bias is modified by environment and hence can be ‘shaped’ to suit different roles.
We will use our established collaborations with a range of detection dog agencies to recruit large numbers of explosives, drug, medical and conservation dogs (e.g . trained to find invasive and protected species). We will refine the cognitive bias test and devise operational sensitivity and specificity measures specific to each discipline. The search characteristics of already trained dogs from different contexts will be measured in standardised tasks, compared to robust and validated metrics of performance2 and related to performance in cognitive bias tests. Comparisons will be made between and within detection discipline, to ascertain whether cognitive bias tests can effectively predict a dog’s sensitivity and specificity appropriate to discipline prior to investing large amounts of time and money in training. We will apply the principles of signal detection theory to ascertain optimal combinations of sensitivity and specificity for different roles. We will also investigate whether cognitive bias testing of naïve pet dogs predicts their performance after training in a simple detection task and how this relates to owner-reported measures of early rearing experience and behaviour.
The student will acquire skills in behavioural and cognitive science, learning theory, experimental design, statistics, management and analysis of large datasets, and dog handling and training.
How to apply for this project
This project will be based in Bristol Veterinary School in the Faculty of Health Sciences at the University of Bristol.
Please visit the Faculty of Health Sciences website for details of how to apply