Service Dogs and Failure Rates
My wife is a certified canine behavior consultant and trainer who specializes in fearful dogs. Recently she began receiving calls for help with service dog training, which led us to a conversation about whether service robots could someday replace service animals. (Don’t worry; I’m not going to suggest that any “replaced” dogs go away. Perhaps they could provide blanket warming service instead?) Per the ADA, a service animal is a dog that has been individually trained to do work or perform tasks for an individual with a disability. The task(s) performed by the dog must be directly related to the person’s disability–specific actions needed to assist the person.
Apparently, the ‘fail’ rate of service animals is pretty high; 70% seems a commonly cited number.+number seems re rate. There may be several reasons for this: Temperament is one commonly cited reason, but training style is another. Positive reinforcement trainers, if asked, often say that aversive training methods result in emotional and/or behavioral problems in an animal, and not uncommonly mutter that many service dogs are still trained with aversive methods.
One private service dog trainer in the eastern USA, who also provides consultations to owner-trainers (people who wish to train their own service dogs) kindly shared her data and observations. Here’s what she said:
15 % of dogs appear to have a good likelihood of success. They have “YES” potential.
64% of dogs definitely cannot be service dogs. They have “NO” potential 21% of dogs might work out (usually these are puppies who show no red flags at time of evaluation).
They have “MAYBE” potential.
Of these owner-trained dogs, most of the MAYBEs become NOs once they mature, and many of the YES dogs will also not succeed because their owners are not good enough or dedicated enough trainers, or because something else goes wrong—for example, dogs that seemed solid YESes
at 1 year old began resource guarding or practicing other disqualifying behavior at 15 months old.
This particular service dog trainer, who is a certified and skilled, seasoned professional, drove home a point I’ve been hedging around:
“I’ve worked with over 100 service dog clients and so far I can count on the fingers of one hand the dogs that have or will very likely work out as service dogs.”
Additionally, I wonder: What about the stress that can come with doing a job we didn’t choose for ourselves? It is common knowledge that being a service animal is stressful. Both the training and the day to day expectations are much more than those placed on a well-trained pet dog—perhaps even than that of many competition dogs. Even if service animals love their jobs and their owners, anyone who has been in love knows that love doesn’t take the stress away. Indeed, sometimes love can intensify it.
Properly training a service dog can take years and cost $25,000 or more, not including the cost of the dog, which can also be high. Because those relatively few dogs who do make it as a service animal work 6 to 8 years, that could equate to one person being paired with as many as 8 to 10 service dogs over his/her life, and that adds up to $250,000 in training or more–plus whatever a breeder charges for each of those 8 to 10 dogs–not to mention the emotional drain. What if there were another way? Just as we have come to embrace technology in other areas of life, so may we come to embrace the potential that technology offers to those with disabilities. AI and Robotics will soon be able to provide what service animals currently provide, without the oppressive expense, stress, and emotional drain.
One peek into this freer future resides with Boston Dynamics and its robot dog, Spot Mini, who handles objects, climbs stairs, and has the ability to operate in offices, homes, and outdoors. It’s equipped with an array of sensors, a navigation system with stereo cameras for depth perception, and of course the requisite equipment for physical aptitude, and can calculate distance and force. Sony provides another peek with its Aibo robot dog, currently a consumer device. Aibo offers a range of preset commands, behaviors, and personality traits, as well as voice recognition and an array of sensors. It sells for $3,000. It’s not hard to imagine a cross between the friendly exterior of an Aibo and the functional framework of Spot Mini: A Service Robot capable of myriad actions paired with affiliative, social behaviors. The Spot Mini is scheduled to go on sale this year, but no price has been set and no one has suggested a market. #BostonDynamics : I’m suggesting a market.
Robots can be specialized to do all the things service dogs do. For example, for disease detection, there are sniffing sensors with nanoscrolls 25 times more sensitive than a dog’s nose. With a camera sensor, a robot can warn of a coming psychological event or seizure by detecting a person’s microexpressions, and facial or audio tics. Robots may be not just conduits between patients and caregivers, but possibly the difference between life and death, as when a medical emergency occurs in a remote area. And while we may be quick to bark our retorts that robots will never replace human social interaction, considering the data on age discrimination and culturally engrained disrespect for the elderly, it is really no surprise that robots are proving themselves so effective in nurturing and lifting these patients, literally and figuratively.
Individual service robots would require a little personalization, but the basic training could be easily copied and downloaded, thereby adding value in terms of convenience not to mention all the training dollars and hours saved. There would be no temperament issues, no debates about training style nor potential for negative fallout from aversive training; these would be moot points, made irrelevant by AI. The chance of a service robot failure would be very low, and if one did fail, it would be readily repairable or replaceable, with no heartbreak, loss of life, or risk of human or animal harm–emotional or otherwise.
Any way you look at it, whether Boston Dynamics, Sony, or another company leaps into the robotic service dog market, there is a clear need for a Service Robot able to perform the tasks of current service animals. Furthermore, since there are no welfare issues for robots, a service robot program in all likelihood would soon progress to levels never before achieved or even hoped for. Last but not least, service robot tech would emancipate dogs from the stressful lives of service work and from whatever happens to them if they fail. (I know some will say nothing bad happens to them and I hope that’s true and I won’t argue, but I also figure it isn’t always true. What about all the dogs that bond with their owners, only to fail and be sent away? What about the broken hearts of the owners?)
Epilogue: Service for More
There’s a push for more human-like androids as well. Boston
Dynamics also has the Atlas unit, which combines running and jumping behaviors
to perform Parkour maneuvers. While folks seeking service tech might not have
Parkour services as a high priority, never fear. There’s a wide variety of
android robots steadily flowing into the consumer markets, and these will no
doubt grow in sophistication. Japan, Germany, and other countries faced with
growing elderly populations, for example, are doing research into robots as a
way of caring for the elderly, with an eye toward services like 24-hour nursing
care. As robotics technology expands, life will become better for all. It
might be hard to imagine, but we have nothing to lose in giving it a go. I
wonder if we might, instead, become more humane as we watch robots bring joy
back into the lives of slighted elders before they fade away.