There is a paper (F. Gaunet, How do guide dogs of blind owners and pet dogs of sighted owners (Canis familiaris) ask their owners for food?, Animal Cognition 2008) mentioned in a blog (here) that is billed as showing that guide dogs do not know their owners are blind. Here is the abstract:
Although there are some indications that dogs (Canis familiaris) use the eyes of humans as a cue during human-dog interactions, the exact conditions under which this holds true are unclear. Analysing whether the interactive modalities of guide dogs and pet dogs differ when they interact with their blind, and sighted owners, respectively, is one way to tackle this problem; more specifically, it allows examining the effect of the visual status of the owner. The interactive behaviours of dogs were recorded when the dogs were prevented from accessing food that they had previously learned to access. A novel audible behaviour was observed: dogs licked their mouths sonorously. Data analyses showed that the guide dogs performed this behaviour longer and more frequently than the pet dogs; seven of the nine guide dogs and two of the nine pet dogs displayed this behaviour. However, gazing at the container where the food was and gazing at the owner (with or without sonorous mouth licking), gaze alternation between the container and the owner, vocalisation and contact with the owner did not differ between groups. Together, the results suggest that there is no overall distinction between guide and pet dogs in exploratory, learning and motivational behaviours and in their understanding of their owner’s attentional state, i.e. guide dogs do not understand that their owner cannot see (them). However, results show that guide dogs are subject to incidental learning and suggest that they supplemented their way to trigger their owners’ attention with a new distal cue.
It may or may not be true that these dogs do not know that their owners are blind. This experiment indicates that but not too strongly. I could do an experiment with people talking on telephones and show that a good many of them believe that the person on the other end of the phone can see them because they would use hand gestures while talking. Or I could show that my dog has knowledge of the difference between my eyesight and my husband’s. This is because she does not move out of the way if we step over her in the daytime. She moves at night so as not to be stepped on. But if there is a lot of moonlight she moves for my husband who has poor sight in low light but not for me. She could have learned this by trial and error or she could have reasoned it out as a difference in eyesight. We don’t know. But we do know that the person on the telephone that gestures is not ignorant of what the other person can see. That person is using a habitual routine without even being aware of how silly it is.
The problem is that we treat other people differently from other animals when we try and understand their thinking. We assume animal are unintelligent as a first assumption and have to prove any instance of smarts. On the other hand we insist that humans think things out consciously and have to establish any instance of behavior being not under conscious control. We really should be using similar criteria for all animals ourselves included.