#16 — Theory of Mind in Animals
Ever wondered if animals can guess what’s going on in our heads? The concept of Theory of Mind (ToM) — allowing us to infer what’s on other people’s mind — is thought be differentiate humans from other animals. However, is that really the case? Maybe we are just not able to measure it properly just yet?
Theory of Mind in Humans
Theory of Mind refers to the ability to infer the mental model of others. We also call this mentalizing or mind reading if you will. It is something that humans make use of all the time in their communication, without even thinking about it too much. If you want to know more about how this works in humans, you could check out this other blogpost I wrote about it.
Some studies state that this is one of the capabilities that differentiates humans from animals. However, there are some highly intelligent beings that make that distinction less… distinctive. We ask ourselves the questions:
- Are they able to attribute mental states to themselves and others?
- Are they able to differentiate between their own mental states and the mental states of others?
Primates
It’s only logical to start with animals closely related to humans that we know to be very intelligent: primates. However, in general ToM research in animals brings in an extra complexity because we are not able to just ask the primates about their thoughts. We will need to infer these from other measurables. In the experiment below the chosen metric is gaze detection.
Gaze tracking
What do we know about the gaze of the primate?
- We know that the primate looks at (1) the initial location of the object, which is also (2) the location the human believes the object to be.
- We know that the primate does not look at (1) the last location of the object.
The real location of the object was not in the mix. I guess because that would cause the primate to just do object following so that would not tell us very much about their ToM-capabilties. However, to make sure that the primate looks at the location the human believes the object to be, there would need to follow a few more variations. Although, in that case, we still rely on how we interpret the gaze. We are the ones that connect a certain meaning to that. The fact that the primates are looking at a location does not necessarily mean that they do so for the reasons we expect them to. It is still a cool experiment though and it does provide some evidence for the hypothesis of ToM-capabilties being present in primates.
Opaque vs see-through experiment
Other experiments make use of opaque vs see-through surfaces that require the primate to infer whether the other test participant can see through it. There have been some positive results for ToM-capabilties in chimpanzees, bonobos, orangutans, and rhesus monkeys. However, this is to a certain extent. For example, they don’t seem to understand human communicative gestures too much. This makes sense because they themselves probably use different communicative gestures. I guess you could train them on this.
Training for ToM
That raises the question: if you train them to improve ToM-capabilities, do they really have those capabilities or are they just playing our heuristics? What does it even mean to have those capabilities? Aren’t we in some way all playing the heuristics? There is no way that the assessment method is perfectly predicting what we need it to predict. In some way, it’s all still an approximation.
Ravens
There are some highly intelligent birds, such as ravens, that might make you wonder whether they have some ToM-capabilities as well. There have been studies on this. Ravens appear to understand the difference between transparent and opaque surfaces and how that influences what can be seen by other test participants. Also, they are able to predict other raven’s behavior as a reaction to humans hiding food. It’s by far not on a human ToM-level, but in a minimal form there are some ToM-capabilties present. They smart. 🦜
Testing these things in ravens also brings in extra complexities because you can not use gaze tracking in the same way as we did with the testing on primates.
Dogs and other pets
If you have a pet then you might be thinking: wait a minute, my dog is perfectly capable of predicting when I am going to collect his leash. Does that count as ToM? That’s a tricky one, because recognizing certain specific patterns is a bit more basic than really understanding the humans intentions. Would the dog still be able to infer you intentions if it were not a specific situation they had seen a 1000 times before?
Research indicates that they do not have human-like ToM, but they do have the ability to rapidly learn to find food locations based on human behavior. Priorities are clear. 😋
Why do we care?
A valid question to ask would be as to why we care so much about these ToM-capabilities. These capabilities are key in human communication and maybe that’s one of the reasons that humanity flourished more than other species? Does humanity really though, or is that just our perception? Maybe ant colonies feel like they are the ones ruling the world? Or some deep sea creatures that we haven’t even seen before? I’m going a bit off-road here, but I believe that everything we know is hugely colored by perception.
But to get back on track: ToM is useful for communication. That’s for sure. That means that being able to simulate that in a robot would allow for more intuitive interaction between humans and robots. Useful tool. ✅
What would be the use of testing for ToM in animals? If ToM is something that developed over evolution, figuring out in what animals and to what extent ToM-capabilties are present could teach us more about evolution. That just sounds generally cool and super useful. ✅
