Recently, I had a conversation with my friend and fellow researcher, Will Pearson about the amount of semantic information transmitted by a guide dog to its handler through the harness it wears. We wondered if anyone had studied this question (I haven’t found anything in a handful of online searches) and if the subtle movements made by a guide dog that a handler can understand is unique to each individual team or if there is a more generic component to it. Finally, we wondered if these subtleties can be quantified and used in an advanced guide dog training system that would include a far greater number of things that the handler can communicate to the dog and vice versa.
X-celerator and I have been working together for almost a year now. As time has gone on, the tactile vocabulary we use to communicate has expanded greatly. This morning, while on our exercise walk, I started counting bits of information that he communicates to me that the trainers didn’t teach us about.
A properly trained guide dog stops walking when one reaches an obstacle. The handler then “clears his space” by feeling around with their foot and by reaching their hand out to feel for things higher up. Today, I noticed that when X-celerator stops at a crack in the sidewalk, he points to it with his nose and, as a consequence, his harness points upward a bit. I noticed that he does this consistently on curbs, broken sidewalk bits and other things I might trip over.
Conversely, when X-celerator wants to indicate that I’m about to walk into a head high obstacle he stops and points his nose upward, thus lowering the handle of the harness. Throughout our walk through the neighborhood, he did this same thing every time a tree branch or bush hung out over the sidewalk.
No one taught us that we could feel the handle move to indicate where an obstacle obstructed our path. X-celerator, sometime in the past year, developed this behavior and I realized today that I had already intuited his meaning and acted accordingly before I grew conscious that this action joined our tactile vocabulary.
Recently, I walked with a friend of mine who trains dogs for a living. He does obedience training and had no experience with guide dogs prior to our walk together. He asked me how I could tell where to stop for a curb. I said that X-celerator stops and I stop when he stops. My friend then informed me that the stop the dog makes can hardly be detected visually, that the stopping process is not sudden but, rather, a very subtle slow down at the end of each block. I remembered that, when the dog and I were new to each other that stops and starts were far more sudden. Thus, a slowing “glide” approaching a stop has entered our vocabulary and works very well as a technique.
I notice all sorts of other things through the harness that I can’t quite quantify yet but will, through observation, try to define, write down and report in BC new things I learn if I can show a consistent pattern. I can usually tell when the dog wants to tell me that he is confused and, through very subtle actions, is asking, “Is this a good idea?” I can’t quite describe the action yet but I’ll watch out for it and see if he does something consistent in that case. I can easily tell when he feels anxiety but, again, I can’t quite describe exactly what he does to tell me. I also know when he feels my anxiety and will try to quantify the action he takes to say so.
A tactile vocabulary across species is pretty interesting. I wonder what Chomsky would say about such?
Afterward
If you have any experience with a tactile method of communicating with your guide dog, I’d be happy to hear your stories. Maybe we can find out consistent patterns over a variety of dog and handler teams to see if, somehow, a similar vocabulary develops during the relationship between human and service animal.
To the person who posted the comment asking how I dealt with the boredom on a non-stop flight from New Delhi to Newark, I have an easy answer – sleeping pills and Bose noise reduction headphones. Ambien works well and the headphones are a must for long haul travel.
— End
I have an interesting story about my uncle’s guide dog, when he was a child. As you may know, Richard Wells is the youngest of three brothers (all born blind). My uncle Talmadge had a guide dog named Lady. All three of the boys used to love to ride their bicycles. What is so interesting to me (and it may just be me because I am not blind) about this is, Lady would nudge the boys away from the rode with her body. And if they got too close, she would knock them completely off of the bike. I don’t know anything about the process of training a guide dog, but it just fascinates me that she knew to keep all three boys out of the road, although she was specifically trained for, and worked with, Talmadge. Anyway, I just thought I’d share that with you.
A common definition of communication is any transmission of information, and so what you describe does appear to be a valid communications channel. Communication can be described by models and one commonly used model is the Shannon-Weaver five stage transmission model of communication. Shannon, who developed the origianl model, and Weaver, who broadened its scope and popularised the model, used the following five stages:
1) Sender – generates the information to be transmitted
2) Encoding – converts the information into some physical form that can be transmitted
3) Transmission – sends the encoded information from A to B
4) Decoding – converts the information from its physical form back to raw information
5 – Receiver – receives and processes the information
So, in your communication system the X-Dog is the sender, the harness is the transmission channel, and you are the receiver.
One critically important aspect of communication is that the associations between the information and its physical encoding can be variable. This view of encoding is supported by the fields of Semiotics and Psychology, particularly work around generic memory within Psychology. if the associations between information and encoding weren’t variable then we wouldn’t have different languages and we couldn’t describe images auditorially, to name a few things that the variable encoding of communication is responsible for.
There are two sets of encodings at work in a communications system: those that the sender uses and those that the receiver uses. When the sender and receiver use the same encodings then you get fault free communication.
I think that there is a very small set of encodings that guide dogs are unintentionally taught. Guide dog training is a form of behaviour modification, and often follows Skinner’s Behaviourist views on instruction. If we know how guide dogs behave in a given situation then we can map information about the situation onto the physically detectable aspects of their behaviour. If a group of guide dogs behave in a common way to a given situation or piece of sensory stimuli then the communication encodings can be viewed as being the same on the part of the dog.
Most of the encodings that a guide dog owner learns are built up through heuristics. They observe how their guide dog behaves in a given situation and when their dog behaves in the same manner again they associate the behaviour that the dog is currently exhibiting with the previously learnt situation. This is an ecample of both deductive and inductive logic on the part of the guide dog owner. This enables the guide dog owner to learn more encodings than just those demonstrated by the behaviour modification of guide dog training. For example, if a dog tries to place a toy in the hand of its owner and then runs away the behavior is often interpreted to mean that the dog wants to play. This use of logic by the guide dog owner meansthat the range of associations are going to be greater than the range of associations common to a guide dog group that has undergone the same behaviour modification.
One interesting question is how much of a guide dog’s behaviour is inate. If it is inate to a species then there is going to be intraspecies commonalities, and this can be used as a common set of encoding associations. It’s not a case of all behaviour less that which a guide dog has been taught is inate, as dogs can use logic to learn as well.
I think teaching guide dogs to communicate more with their owners is a very interesting and useful topic. Safe mobility is about sensing the surroundings and adapting behaviour accordingly. if we can turn guide dogs into remote sensors for blind people and then communicate the sensed information back to the guide dog owner then it should lead to safer mobility as the guide dog owner can make more informed decisions. When I was training with guide dog 2.0 last year the training team manager, who as a masters in social research, and myself were discussing the possibility of some joint research. I think this would be an interesting topic, and so I might propose a study to investigate whether we can teach dogs and owners a common set of associations and whether this does lead to safer mobility. A follow up could be to look at factors that affect communication whilst navigating around an environment, such as dual task attention, etc.