Exercise and brain function, hedgehog hearing, and can AI change our minds? – podcast

0:00This is the Guardian. Hear that? No? Well, according to scientists, that definitively proves you are not a hedgehog. The revelation that these sweet, spiky creatures pick up sounds well into the ultrasonic range opens up the possibility of better understanding hedgehog communication and how we can make use of it to slow or reverse their drastic declines. As well as exploring new ways to converse with the hedgehogs, researchers have been investigating one of the new ways we converse with each other with the aid of AI autocomplete. It's certainly changing how we write, but could it also change how we think?

1:04A much more low-tech and simple approach for positively influencing the mind is exercise. And scientists this week have been probing the possible reasons why working out appears to be good for our body and brain. So today I sit down with my co-host and science editor, Ian Sample, to discuss the stories you need to know about this week. From the Guardian, I'm Madeline Finley, and this is Science Weekly. Ian, our first story today is about the link between exercise and the brain. There is a fair bit of research that links exercise and brain function, but the mechanisms behind all of it are still quite murky. So what has this group at UCL done to try and clear this up? You're right, there's been an awful lot of work looking at exercise and brain function,

2:05and I would say a lot of that researchers have pretty mixed results. What this group at UCL was doing was they wanted to look at exercise and brain function, but as a potential mechanism for that, they wanted to look at blood levels of a protein called brain-derived neutrophic factor. We can call that BDNF for short. It's often referred to as fertiliser for the brain. It's actually a protein that's produced in the brain, and crucially, it helps protect neurons. It helps to drive the growth of fresh neurons, and also helps in the formation of new connections between brain neurons. So really this study, it's exploratory. It's really sensitive research, but they wanted to look at is there some kind of mechanism going on here, which is linking exercise and brain function through the release of this neuron nurturing protein? Very intriguing. So what did they do? How did they look at this? So for the study, they recruited people who were unfit and sedentary. These people weren't getting much exercise. They had high BMI's, and they divided them into two groups. So first

3:09we had a control group, and for them, it's just carry on as normal. The interesting part of this is the intervention group, and they were put on an exercise program, which involved going on an exercise bike at a gym, whereby the intensity of the exercise increased over 12 weeks, and throughout the study, they're giving blood so that the researchers can measure their blood levels of BDNF, and they're also testing them for their attention, their focus, for their memory, and their reaction times, things like that. And they're also using a technique to look at their brain activity while all this is going on to see if they can see any changes in how their brains are behaving. Right, so what did they find? So it's worth saying straight up, they didn't see any improvement in how people performed on the cognitive tasks after doing the exercise. They also didn't see any rise in the resting level of BDNF in people's blood. But what was interesting, and I think this

4:10is really just one of those studies which says, okay, there might be something interesting here to go and look at properly, was the people got fitter after this 12 week exercise regime, that increase in fitness was linked to a BDNF response. So let me sort of unpack that. Once these people have finished that 12 week exercise program, if you then got them to do 15 minute bouts of really intense exercise, the amount of BDNF they produced really spiked, they produced more than they would have done when they were less fit. So what the study's really doing is it's saying that if you're fitter, then you seem to produce more of this BDNF when you have a sort of sharp, sharp bout of exercise. And they also see that even though people weren't doing better on the cognitive tasks, some of the brain imaging they were doing suggests that the prefrontal cortex, that part of the brain where you're making decisions and all sorts of things like that, they thought the activity there looked as if they had seen some changes that were beneficial. So Ian, if we were going to try and interpret this, as you said it's exploratory,

5:13but it sounds like the idea would be, once you get fit, if you keep exercising, you might potentially have some kind of brain benefit. I think that is the hope. This is all part of this broader feel to try and understand whether exercise is really benefiting the brain in terms of cognition, and if it is how it's doing that. And I guess a good thing is the participants in this got fit, the very least. Absolutely, it's only three months and there's a pretty good commitment, you know, four times a week, but their improvement in their fitness was pretty impressive. So I think that should at least give everybody hope. Maddie, you've got a story about hedgehogs and what they can hear. What's all this about? Yeah, so researchers have found that hedgehogs can hear in the ultrasound and they hope that this might lead to more effective hedgehog deterrence. Now why would

6:16you want to deter a hedgehog because they're the most wonderful creatures? It's because, unfortunately, in Europe, particularly there has been a drastic decline in hedgehogs. We'll probably both be old enough to remember the road safety TV ads with the parent and child hedgehog giving advice about how to cross the road. That was very pertinent because actually it's been estimated that one in three hedgehogs are killed by vehicles. I've seen numbers around 335,000 hedgehogs dying on the road every year and that isn't even their only threat. The use of pesticides and the huge drop in insect numbers we've seen habitat loss and fragmentation and then there's also them getting other injuries from things like streamers or robotic lawn mowers. In the UK, it's been estimated that there might be up to a 75% decline in their numbers in rural areas since the start of the 2000s.

7:19So as soon as I heard about this story, I got in touch with the lead researcher Dr. Sophie Lund Rasmussen. She's from the University of Oxford and I wanted to ask her, was there any reason why she thought that hedgehogs might hear in the ultrasound in the first place? So I've often been wondering about hedgehog communication because they don't have that many sounds and I think it's a bit strange because you can tell when they meet each other in the garden that they are communicating, they are interacting somehow but no noise is coming out of them. So to actually investigate their ability to hear ultrasound was pretty obvious I think because I was just sure that something was going on. But Maddie, how did they test the hedgehogs hearing to find out that they can hear in the ultrasonic? So there were two parts to this study. One part was they did CT scans of a euthanised hedgehog who had been injured. They built a 3D model of the middle

8:23and inner ear and in doing that they found structures that basically showed them that hedgehogs can process high pitch sounds very efficiently and they found structures that very similar to what you might see in other animals that use ultrasound like echolocating bats. And then in the other part of this study they took 20 rehabilitated hedgehogs from wildlife rescue centres and they gave them a small amount of anesthesia and they put electrodes just under the skin to measure the brain activity and then they played them different frequencies of ultrasound. Now when I was reading about this I couldn't stop thinking that this is like you know when people describe being abducted by aliens what's happened to them. I feel like these poor hedgehogs you know when they came round must have felt the same way anyway. What these scientists were doing is they were looking for the auditory brain stem response. So electrical signals that are travelling between the inner ear and

9:25the brain in response to these different sound frequencies. What they found was that hedgehogs hear sounds from around four kilohertz up to at least 85 kilohertz and they seem to have peak sensitivity around 40 kilohertz so this is well into the ultrasonic range it's into a range that humans can't hear and actually Sophie told me that now they're trying to understand a bit more about how hedgehogs use ultrasound themselves to communicate. If they hear best at 40 kilohertz there's a great likelihood that they also communicate around 40 kilohertz. So we've already started testing this and we've been recording a lot of interactions between hedgehogs to detect whether they are communicating it ultrasound and I'll be happy to share the results when we know what we found. And you said that the researchers thought this discovery could be useful for designing more effective hedgehog deterrence. Yeah one thing that occurred to me when I read this study was

10:31cars and garden streamers are really loud and there's lots of vibrations but clearly that's not a deterrent right because so many are dying on our roads so maybe the hedgehogs would react more to the peak of where they hear. Scientists and conservationists might be able to design something that works better that perhaps you put on the front of your car you put on your lawn mower and it was something that Sophie was really really keen on it. I mean ultimately to her this was the point of the research. How does one design a sound repellent that is efficient? Which sounds actually scare hedgehogs? Perhaps it's a pulsating sound perhaps it's a deep vibration we don't know that yet but we also have to make sure that the sound repellents are only efficient when they're running. I don't want to risk evicting hedgehogs from gardens causing the hedgehogs to

11:32not cross the roads altogether. We also have to be aware that bats and mice can hear in the same frequency range and what happens if they're scared it takes a lot of thorough solid research to develop these sound repellents and I'm hoping that the kind industry would want to support this research and save thousands of hedgehog lives and hopefully turn this massive and worrying decline around and ensure that our grandchildren get this wonderful nature experience it is to meet a hedgehog in the garden at night. I don't suppose she had any thoughts on ways that you might use ultrasound to attract hedgehogs into your garden for anyone who might be interested in doing that. Do you know what? She was just talking about deterrence but Ian I think you've come up with a great idea there you've got a business plan on your hands because I can't imagine anyone who wouldn't want hedgehogs in your garden I mean that they eat slugs and snails that's so handy.

12:34Coming up, can AI assistance change our minds? Chicago 2011 a cop is murdered police and prosecutors swear they have the trigger man he swears he didn't do it how far will each side go to prove their right like it's just one bombshell after another you know where you're like what what the story of a playstation a brain eating amoeba and the relentless pursuit of justice off duty coming soon listen wherever you get your podcast. So Ian our third and final story for today is about AI and how it can influence what we think we've heard about how chat bots can be persuasive but this research is

13:39looking at something a little more subtle isn't it? That's right so I'm sure pretty much everyone by now has come across these sort of AI writing assistance where whether you want them to or not they will start completing your sentence for you before you've managed to complete it yourself often with some sort of banal suggestion of how you might want to communicate with someone else and what these researchers from Cornell University in New York wanted to do was to see whether those suggestions which obviously cause they're coming from an AI can easily have inherent biases actually shifted people's attitudes about issues so it's not only would they have an influence on obviously what they're writing but does it actually affect how they do things afterwards? This is so interesting because when I see auto complete I immediately am thinking to myself well I'm not going to say that because you've suggested it but how did the researchers actually go about testing this? So they had a couple of really large-scale experiments they involved more than two

14:40and a half thousand people and in the first one people were asked to write a short essay four or a gates standardized testing and education and as they went about the task some of the people taking part saw these auto complete suggestions so these are AI powered suggestions of how they might want to write that sentence but these suggestions came from an AI assistant that they had engineered to give a sort of slightly biased view on the topic. In the second experiment they had a similar number of people but they were asked to write about things like the death penalty fracking genetically modified organisms and voting rights for criminals so all really quite sort of sensitive public issues and again the AI was prompted to give slightly biased suggestions in its corrections and what they wanted to do in that second study in particular was they had got people to fill out a quiz about their views beforehand and then they got them to fill out the same quiz afterwards to see if there was any shift in their attitudes. I would imagine that no

15:45one would want to think that they could be biased by an AI but what did the researchers find? What's really interesting in the results is that not only did their actual attitudes to issues change in the same direction that the AI was biased they were unaware of this that they didn't appreciate that that is what happened and I think even more sort of devastating is that even when they were told beforehand that the AI would be biased or afterwards it didn't mitigate the effect they still express this shift in attitudes so by being exposed to these auto completions it wasn't just changing what they wrote it was shifting their attitudes about the issues that they were writing about. I mean I find these results quite troubling because we know that AI can be biased and you look at a system like Grock and what that's come out with why would this be happening though

16:46why is it making us sort of think differently? Well one question that came to my mind when I was reading this paper was okay you're seeing a shift in people's expressed attitudes immediately after they've done this task but I'm kind of interested in whether that effect would really hold if you gave them the same questionnaire a week two weeks three weeks down the line is this just the kind of effect where you might have a conversation with someone who has slightly different views to you you hear the logic of what they're saying it might seem reasonable for about five minutes until you go for walk and realise actually no it's not what I believe but I think it's really interesting why this particular sort of mode is affecting people's attitudes that it's affecting them when it's completing their sentences in another part of this study they actually gave people instead of having the AI auto complete their sentences they were given some similarly biased sentences that have been generated by the AI but just as text to have while they wrote if you have that same biased information but not presented to you as sort of

17:48completing your sentences the shift in your attitude was much less which suggests that there is something particularly going on when the AI is trying to complete a sentence as you're writing it and I don't know why that is and I don't think the scientist know why that is but you can imagine if you're writing something you're thinking it through and these answers are pouring out on the screen in front of you as you're typing it's probably quite easy for you to be fooled that this is how you're thinking again maybe until you've had time to reflect and think about whether this really gels with your core values. Gosh the implementation of all these systems is just really a big experiment on us isn't it? I think it really is and Lord knows how many emails and documents are written using these auto complete writing assistance but I'm sure there is another bunch of people I think journalists are some of them where as soon as they see these suggestions they bristle and will go out of their way to write anything else. And for those that aren't resistant Ian I've got a business idea for you a Guardian AI auto complete bot so that we can really sell our pro hedgehog

18:55pro fermented food pro wild swimming agenda. My thanks again to Ian and to Dr Sophie Rasmussen you can find the article about her research on the Guardian.com and before you go I'd like to recommend the latest episode of the Guardian Long Read where writer and academic Ariane Chavezi describes her happy childhood holiday spent in Iran. She paints a portrait of everyday life in the country behind today's headlines. Just search for audio long read wherever you're listening to this. And that's it for today. This episode was produced by Ellie Sands. It was sound designed by Joel Cox and the executive producer is Ellie Beary. We'll be back on Tuesday. See you then.

19:56This is the Guardian