Wow! Kind of re-adjusts the weight thrown behind the mirror test. [0]
Recognizing self-recognition and sentience is truly a quagmire of problems, but I suppose it makes sense that it becomes tied to the most salient sensory input for the entity.
This concept will deeply complicate the uncanny valley of machine sentience, and the boundary between live organic life, and contrived artifacts that simulate comparable systems.
Interesting idea, that the mirror test for dogs is unfair because their vision sucks compared to their nose, but this reasoning seems to take quite a leap. From the article:
“This study confirmed the previous evidence proposed with the STSR by Dr. Cazzolla Gatti showing that "dogs distinguish between the olfactory 'image' of themselves when modified: investigating their own odour for longer when it had an additional odour accompanying it than when it did not. Such behaviour implies a recognition of the odour as being of or from 'themselves'."”
Maybe the dogs just recognize their own modified scent as very familliar? Not as a representation of themseles, but just a smelll they smell often. I love dogs, and don’t doubt for a moment they have some level of self-awareness similar to our own, but to me, this test stinks.
The mirror test works like this: you stealthily put a mark on the face or head of the being and then present the mirror. If the being then tries to reach the mark it is a strong clue of self-recognition.
In the excerpt the exact way of the sniff test is not described. I didn't read the paper, hoping for a summary here on HN. I thought maybey they controlled for confounding effects, for example by presenting other familiar smells, including the smell of self. The modified smell showing a higher level of interest compared to other smells would be a clue of self-recognition.
Anyway the approach shows clearly a bias of the mirror test. You could say the mirror test was discriminating the dogs.
This might be a leap, but I felt reminded of the hearing bias Deaf people experience every day. The majority of people assume that hearing and communicating via the audio channel is the way to go, and so I experience hurdles like needing to phone for identification for services.
Is this like the experiment where they put a dot on a bird and then showed the bird itself in a mirror, and then the bird tried to clean the dot off itself?
It seems that compared to the mirror test, this sniff test lacks an action analogous to the "wiping the mark off the face" action or (from the paper) "use the mirror to direct examination of the mark" that tends to happen in the mirror test, which would allow for differentiation between sniffing out of increased interest/familiarity and self-recognition.
> Interesting idea, that the mirror test for dogs is unfair because their vision sucks compared to their nose, but this reasoning seems to take quite a leap.
This has been a decades old complaint about the mirror test. It is a human biased test which favors the visual over other senses. Different animals use different senses. It would be like giving the mirror test to a blind person and then claiming the guy wasn't self aware.
Then there is the entire about what "self-aware" means, especially in the current environment where the "self" has taken quite a philosophical and scientific beating where the "self" like the "soul" appears to be a man-made illusion/fantasy.
> I love dogs, and don’t doubt for a moment they have some level of self-awareness similar to our own, but to me, this test stinks.
This is an age old philosophical problem. Should we expect a dog or any animal to have "self-awareness similar to our own"? If so, is it even possible to verify it either way? Considering our brains are different, is it even physically possibly for two species to be "self-aware" in the same sense of the word?
E.g. " Finally, in a second experiment, subjects spent longer with the modified stimulus than with the modified odour by itself, indicating that novelty alone does not explain the dogs' behavior."
If you read a news item about the scientific article and it appears lacking, most of the time it's that the journalist (or the editor) simply omitted the details that make the argument consistent. The news can't copy everything word-for-word, and the time that the "reporter" has to write the "news" and the time he spent on learning on the subject are typically many orders of magnitude smaller than the person who did the scientific work.
Which also doesn't mean that all scientific papers can stand the test of the time. But when seeing inconsistencies, it's worth to first look at the real thing.
I’m not sure we’d like the result of that. Even milk is produced by the terrible way of separating mothers from their children. Anyone who’s been to a dairy farm knows most of those cows are absolutely miserable for the majority of their lives.
I'm not sure this is just a problem with dog's poor visual processing. Sample size N=1 and what not, but I had a dog that would see it's reflection in the mirror, and start growling, hackles would start raising, and as the reflected "dog" started doing the same thing would start a loop that further enraged my dog. Shouldn't a dog be able to also recognize their own sound and emotions?
Counter-anecdote. My dog finds his reflection curious for a minute or so until he sniffs at it. Without any scent, he loses interest in the reflection, appearing to shrug it off as make-believe like the dogs he sees on television.
Seems kind of self-aware-ish to me, in only that he can distinguish between sight and scent of himself.
I always thought my border collie sometimes seemed stressed and panted after she investigated the clumps of hair that would sometimes fall out when being pet or groomed. I remember it happening after saying to her "it's you!" Haha :-(. Poor dog thought she was falling apart. I remember telling her "it's okay" and she licked the air in my direction. Some part of me thought I was nuts for thinking she understood or anthropomorphizing her. But as the years went on we realized she knew what was up more often than we could imagine, so many instances of her reacting to spoken words or the world in ways we never thought her capable of.
This is kind of fascinating. My first impression was I'm not sure this works as a substitute for the mirror test because I don't know how you distinguish a dog recognizing a smell as self from a dog recognizing an unfamiliar smell jumbled with a familiar smell (without associating the self-smell with actual self, just with "I smell this all the time").
But that got me thinking about the mirror test itself. If it were somehow possible to---in such a way that the person was unaware of the change---wire someone up to co-opt their vision to remote cameras, then give them control of a robot body that looked like them and mimicked their movements faithfully, then numbed their sense of touch... Could you fool a person into "failing" the mirror test by identifying the robot as themself when it is not actually their own body?
Too far down this road, of course, you run into some weird philosophy questions on "what is 'my body' anyway?" But I don't know of an experiment of this style that's been tried on people. Could be interesting.
19 comments
[ 4.4 ms ] story [ 61.0 ms ] threadRecognizing self-recognition and sentience is truly a quagmire of problems, but I suppose it makes sense that it becomes tied to the most salient sensory input for the entity.
This concept will deeply complicate the uncanny valley of machine sentience, and the boundary between live organic life, and contrived artifacts that simulate comparable systems.
[0] https://en.wikipedia.org/wiki/Mirror_test
“This study confirmed the previous evidence proposed with the STSR by Dr. Cazzolla Gatti showing that "dogs distinguish between the olfactory 'image' of themselves when modified: investigating their own odour for longer when it had an additional odour accompanying it than when it did not. Such behaviour implies a recognition of the odour as being of or from 'themselves'."”
Maybe the dogs just recognize their own modified scent as very familliar? Not as a representation of themseles, but just a smelll they smell often. I love dogs, and don’t doubt for a moment they have some level of self-awareness similar to our own, but to me, this test stinks.
In the excerpt the exact way of the sniff test is not described. I didn't read the paper, hoping for a summary here on HN. I thought maybey they controlled for confounding effects, for example by presenting other familiar smells, including the smell of self. The modified smell showing a higher level of interest compared to other smells would be a clue of self-recognition.
Anyway the approach shows clearly a bias of the mirror test. You could say the mirror test was discriminating the dogs.
This might be a leap, but I felt reminded of the hearing bias Deaf people experience every day. The majority of people assume that hearing and communicating via the audio channel is the way to go, and so I experience hurdles like needing to phone for identification for services.
I sympathise with the dogs.
This has been a decades old complaint about the mirror test. It is a human biased test which favors the visual over other senses. Different animals use different senses. It would be like giving the mirror test to a blind person and then claiming the guy wasn't self aware.
Then there is the entire about what "self-aware" means, especially in the current environment where the "self" has taken quite a philosophical and scientific beating where the "self" like the "soul" appears to be a man-made illusion/fantasy.
> I love dogs, and don’t doubt for a moment they have some level of self-awareness similar to our own, but to me, this test stinks.
This is an age old philosophical problem. Should we expect a dog or any animal to have "self-awareness similar to our own"? If so, is it even possible to verify it either way? Considering our brains are different, is it even physically possibly for two species to be "self-aware" in the same sense of the word?
https://www.sciencedirect.com/science/article/pii/S037663571...
E.g. " Finally, in a second experiment, subjects spent longer with the modified stimulus than with the modified odour by itself, indicating that novelty alone does not explain the dogs' behavior."
If you read a news item about the scientific article and it appears lacking, most of the time it's that the journalist (or the editor) simply omitted the details that make the argument consistent. The news can't copy everything word-for-word, and the time that the "reporter" has to write the "news" and the time he spent on learning on the subject are typically many orders of magnitude smaller than the person who did the scientific work.
Which also doesn't mean that all scientific papers can stand the test of the time. But when seeing inconsistencies, it's worth to first look at the real thing.
Point being that perhaps the visual markers we used for dogs were not sufficiently strong.
The other cows we eat. :p
But that got me thinking about the mirror test itself. If it were somehow possible to---in such a way that the person was unaware of the change---wire someone up to co-opt their vision to remote cameras, then give them control of a robot body that looked like them and mimicked their movements faithfully, then numbed their sense of touch... Could you fool a person into "failing" the mirror test by identifying the robot as themself when it is not actually their own body?
Too far down this road, of course, you run into some weird philosophy questions on "what is 'my body' anyway?" But I don't know of an experiment of this style that's been tried on people. Could be interesting.