09/10/2024
Une publication forte intéressante par Natalie Duberow au sujet du pairing :
Si besoin d’aide dans la traduction n’hésitez pas ! 🥰
Les méthodes utilisées depuis des années évoluent grâce aux études, c’est ce qui rend la formation enrichissante car cela n’a de cesse d’évoluer.
« You asked!
My thoughts on pairing. Based on science not on the method. Just because the “method” had been used for years, support for this method is unjustified if it lacks scientific validation.
Pairing can be problematic in scentwork because it risks creating a dependency on visual or contextual cues from the reward, rather than focusing solely on the target odor. Here are a few reasons why I believe pairing is not an ideal method:
1. Visual and Contextual Reliance
When food or a toy is consistently placed with the odor, dogs may start associating the visual presence of the reward or the context around it, rather than the scent itself. This can lead to dogs searching for the reward rather than the target odor, reducing their independence and scent discrimination skills.
2. Reduced Precision
Over time, pairing will promote “fringing” behavior, where dogs alert near the source but not directly on it. Since they anticipate the reward rather than working to the exact location of the odor, they might settle for an alert that’s in the vicinity, which can be problematic in competition or real-world applications where precision is crucial. That is exacerbated by so-called “rewarding on source” which is not only out of parameter, but quite often is out of the spectrum altogether
3. Decreased Motivation for Odor Alone
If dogs are constantly rewarded directly on the “source”, they might not develop an intrinsic motivation for finding the target odor. When the reward is removed, their drive to search may diminish. The goal in scentwork is to create a strong, odor-driven search pattern where the dog works to source, not simply to get a visible or expected reward.
4. Difficulty in Transitioning Away from Pairing
Once dogs are trained using pairing, it can be difficult to phase out the reward without causing confusion. Dogs start to hesitate or stop searching as enthusiastically when the reward is no longer paired with the odor. This can lead to a breakdown in their understanding of the scentwork task.
In advanced scentwork teaching, it is more effective to build value in the odor itself from the start, using indirect reinforcement (rewarding the dog away from the source) to encourage a strong commitment to the odor without relying on visible rewards.
To address the limitations of pairing in scentwork training, here is scientific research on scent detection, learning theory, and reward-based training. Although pairing is commonly used, studies reveal that this is not a training method that lead to more precise odor detection without the pitfalls of reliance on visible rewards.
Here is the science -
1. Overshadowing
Research on overshadowing in scent detection (Hall et al., 2016) shows that when multiple stimuli (e.g., odor and food) are presented together, the more salient stimulus (often food or a toy) can overshadow the less salient one (the target odor). Dogs may become fixated on the food or toy and learn to associate the reward with visual or environmental cues rather than the odor. This weakens their ability to focus solely on the target odor and reduces search accuracy over time.
Study Reference:
Hall NJ, Smith DW, Wynne CD. “Training domestic dogs (Canis lupus familiaris) on a novel odor detection task: The role of social and visual cues.” Animal Cognition, 2016.
2. Blocking Effect
The blocking effect (Kamin, 1969) in classical conditioning explains that when a previously learned association (reward with odor) exists, introducing a secondary reward (like pairing food with odor) can “block” the learning of the new association. This could mean that when pairing is used, the dog may fail to fully learn the odor-reward connection because it’s being blocked by the more dominant food-reward association.
Study Reference:
Kamin LJ. “Predictability, surprise, attention, and conditioning.” In B. A. Campbell & R. M. Church (Eds.), Punishment and Aversive Behavior, 1969.
3. Latent Inhibition
Latent inhibition (Lubow & Moore, 1959) occurs when a dog is exposed to a non-reinforced stimulus (e.g., odor paired with food) too frequently, reducing the likelihood that the dog will respond to the odor on its own. By pairing food and odor too early or too often, dogs may form a weaker association between the target odor and the reward when the food is phased out, leading to less reliable performance.
Study Reference:
Lubow RE, Moore AU. “Latent inhibition: The effect of non-reinforced pre-exposure to the conditioned stimulus.” Journal of Comparative and Physiological Psychology, 1959.
4. Delayed Reinforcement and Independent Odor Commitment
Research has shown that delayed reinforcement encourages dogs to develop more independent and reliable odor detection. Delayed rewards (giving the reward after the dog has moved away from the source) prevent dogs from focusing on the immediate presence of a treat or toy, reinforcing the idea that the odor itself leads to reward. Studies in canine scent detection (Gazit et al., 2005) suggest that dogs trained with delayed reinforcement exhibit stronger, more precise performance than those trained with immediate pairing of odor and reward.
Study Reference:
Gazit I, Goldblatt A, Terkel J. “The role of context specificity in learning: The effect of training context on explosives detection in dogs.” Animal Cognition, 2005.
5. Precision Through Odor-Reinforcement Separation
Finally, a study by Macpherson et al. (2008) highlights the benefits of odor-reward separation in scent detection. When dogs are taught that finding the odor results in reward (away from the odor source), they become more accurate at locating the odor itself, reducing fringing and improving precision. This method relies on building value in the odor independently of the reward, which is not typically achieved with pairing.
Study Reference:
Macpherson K, Roberts WA. “Can dogs (Canis familiaris) detect human emotional expressions? Learning from training.” Animal Cognition, 2008.
Here is Dot working on thermal diffusion. »
