INTERVIEW: Lars Lewejohann talks about mouse intelligence
Lars Lewejohann, SCIoI PI in projects 3, 25, 40 and 46 tells us more about what’s going on in a mouse’s mind.
What is your line of research within the Cluster?
A big part of the research we are doing at SCIoI involves finding the relation between different forms of intelligence in different entities, such as robots, humans, and different species of animals. For example, we compare the intelligent behavior of mice with that of other rodents, or with that of birds such as cockatoos. One way to observe these behaviors is by having the animals use lockboxes with different mechanisms that need to be opened in a certain order. In our labs we also go a step further and investigate social interactions in mice in order to understand the connection between sociality and learning. And our projects also have further ramifications: for example, we study facial expressions in mice and communication with others, and we also examine the type of physical environments the mice are acting in. It is a varied set of experiments, currently grouped under SCIoI projects 3 and 25, though we have a further project, starting in the coming months.
More in detail, how would you describe your projects?
In projects 3 and 25, we set up tasks of varying complexities using lockboxes and a so-called IntelliCage. In both projects mice are requested to solve tasks in order to receive rewards. Both experiments have an analytical part and a synthetic part that are linked together. For example, in project 25 the analytical part is led by Benjamin Lang, and the synthetic part is led by Mark Boon. First, Benjamin observes the mice’s behavior and social interactions and tracks them with a live mouse tracker that uses both computer vision and films the mice 24/7, then Mark analyzes the data synthetically, exploring the mice’s learning and memory behaviors and modeling the outputs to find algorithms. This process helps us develop theoretical definitions of the behaviors and gives us insights on which strategies are used and which sensory signals lead to social learning behavior. It’s a pretty good example of the synthetic-analytic loop, which is an important aspect in SCIoI research.
What does a mouse’s facial expression tell us?
We know that mice can signal emotions through their facial expressions, especially if they are in pain. But there are few hints of evidence that show that this is not only evident in negative emotional states, but also in positive affective states, and that’s what we are interested in. With our researcher Katharina Hohlbaum we look at many individual pictures of mice and see features that give scientific evidence for happy faces and positive emotions. This gives us a whole new range of aspects we can work with, especially when it comes to social interaction. For example, if we can prove that a mouse can convey positive information to their conspecifics through a facial expression of satisfaction (maybe because there’s a reward,) other mice that are observing are bound to be more interested in taking part in the experiment. These emotional responses can enhance the learning process of the observing mouse.
How does mouse intelligence compare with, say, cockatoo intelligence?
One thing that stands out when we compare these two species is that while cockatoos are able to first think and then act, for example when opening a lock in the lockbox, mice are more “hoppy,” displaying a trial-and-error behavior that I sometimes call a “Homer Simpson” type of behavior, where the mouse tries something out and then seems to think “Doh! That was wrong!” But then again, cockatoos and mice are probably too far off to really allow for a comparison, and it would probably be more appropriate to compare mice with rats. A bit like cockatoos, rats tend to be slower in their exploratory behaviour, and give the impression that they think before acting. Mice will make the same mistake over and over until they get it right. Eventually they’ll learn, but I have the impression that they never fully give up their trial-and-error method. But that’s not a bad thing, and it has its advantages: it makes you more flexible, and helps find new solutions to the problem. If you always do things in the same way, then you’re kind of stuck with the one solution you’ve learned. Our exploration of mouse intelligence depends on finding out how to ask the mice the right questions, and on observing the type of decisions they make in order to find the right solution.
In one of your latest papers you talk about performing these experiments in semi-naturalistic environments. What advantages does this have?
What we observe in shoebox-sized cages is different from what happens in a natural environment, and this is why we are also performing these experiments in more naturalistic setups, where the mice’s behavioural repertoires are more varied and diverse compared to those occurring in the smaller cage. This shows us that there’s much more that these animals are able to do, and that in order to solve riddles and mechanisms that they have never seen before, they are able to use skills that have evolved and that they have learned in ontogeny. And this an adaptive behaviour.
Another big concern of ours is animal welfare, and I believe that improving a mouse’s housing conditions can really help improve their wellbeing, because it introduces variation and makes the environment more interesting. That’s also where the lockboxes came into play, because they can serve as a cognitive enrichment. They give the mice something to do, a challenge and problems to solve and to work for in order to make a living. In the recent paper you mentioned, we examined the validity of using semi-naturalistic environments in oder to understand whether this environment still ensures data quality, and we concluded that even taking quantum leaps towards improving animal welfare does not inevitably mean a setback in terms of data quality, which is great news.
Do you think the mice are intelligent according to the definition of SCIOI?
Yes, they are. They probably will not go to university and get a degree or anything like that, but their form of intelligence has evolved in order for them to succeed in their ecological niche, and the adaptive behaviours we observe in our experiments fit the basic definition of intelligence that we agreed in the initial proposal of SCIoI. But we should not forget that the definition of intelligence is also evolving, and we might come to different conclusions further down the road.