Interview Chris Lintott
Citizen science is the backbone of astronomical discoveries. Professor Chris Lintott speaks to All About Space about his popular project, the Zooniverse, and how it’s continuing the tradition with great success
The astronomer reveals how you can get involved in real-life research – without the need for a science degree!
How would you explain ‘citizen science’ to someone that isn't aware of it?
I think to astronomers in some sense it's old news, because astronomers – of all stripes – have been contributing to science for centuries. You can go right the way back to the first studies of variable stars, or the people who kept an eye on storms on Jupiter in the 19th century, and you discover that there were plenty of people who were well resourced and had telescopes in their backyard.
I remember growing up, realising that I was sort of getting interested in astronomy at the time and that amateur [astronomers] were getting serious about supernova hunting. People like Tom Boles and Mark Armstrong were discovering the first British supernovae. Citizen science is as simple as the idea that anyone can do science, whereas what has happened now is that we all have access to large data sets produced by professional astronomers and professionalised observatories. We need help sorting through the data, and that's really the new thing that's happened in the last few years. We've got professionals who have got so good at collecting data that they are overwhelmed – that’s where we need help.
A chapter in your book talks about the history of crowdsourcing. Do you think it's evolved since the days of Edmond Halley crowdsourcing observation of a solar eclipse in 1715, or do you think it's been as strong as it ever was?
There's some evidence in the 20th century where science becomes professionalised. Towards the end of the 19th century, people started getting paid to be astronomers for the first time. I think in the 20th century, cutting-edge science was a bit more – not entirely, of course – but was a bit more centralised towards professionals. In the 21st century, because we built these communication and digital technologies, we’re back to the point where everyone could communicate. I think it's an old idea, but with a very 21st-century spin on it.
A term you use a lot in the book is ‘proper science’. How would you define proper science?
What I'm trying to say is that anything that contributes is proper science. I think we’re very bad at teaching what science is actually like. I think people have this idea of proper science which probably involves somebody in a lab coat, probably with some purple, bubbling liquid. Or maybe it's a professor in front of a blackboard with some chalk and the word ‘eureka’. But the people who build cameras for observatories are doing science. The people who literally grease the wheels of the telescopes are doing science. The people who are online in the Zooniverse projects and looking at the data for the first time are doing science.
Part of it is that science is this great collective effort, and a lot of us can play a part in it. We don't have to be the person with the bubbling liquid or the person solving the equations to be a scientist. I also think there's a promise in that when we ask people to look at galaxies on Galaxy Zoo or to look for planets on Planet Hunters, or whatever it is, I think there's a guarantee that from us, we're going to use those results.
Could you tell us a bit about the variety of the projects that are on the Zooniverse website?
I usually describe myself as the world's most distracted astronomer. We didn't set out to do this deliberately. We were just trying to solve an astronomy problem. But it turns out that we're not the only ones who have had this problem of creating too much data. Pretty much as soon as we launched the project, we started to get other people approaching us. Now we've done all sorts of things, from working with amazing microscopes to historical texts and other things.
I think my favourites, or at least the area that is instinctively interesting, are the ecology projects. Ecology is fascinating, partly because the science is interesting, but I hadn't really realised it's a bit like astronomy in that it's a science where it's hard to do experiments. If I want to understand how galaxies merge, I can simulate them on a computer, but I can't really crash Andromeda and Triangulum together tomorrow and see what happens – probably just as well. If I want to understand cheetah’s behaviour, yes, I can simulate them on a computer, but I'm not allowed to put a cheetah in a park with seven antelope and see what happens.
So you've got the same sort of problem. You've got a set of information that you've observed. There
“Part of it is that science is this great collective effort, and a lot of us can play a part in it”
are all sorts of biases and so on, and the scientific trick is to make use of that information.
When you look back over the last decade or so since Galaxy Zoo’s inception, what have been some major milestones in the project?
Day one of Galaxy Zoo was the most surprising day of my life. We thought we were launching a tiny project that a few thousand amateur astronomers might help us with. I just remember sitting looking at the email inbox and realising that in the first half an hour of a project, we had got 20,000 emails. Now, most of those emails said ‘your site isn't working’ because it was buckling under the pressure, so day one was very important.
Getting the first Galaxy Zoo paper accepted was a big deal, because we had this idea that this had to be real science – it had to be authentic science. I had a quarter of a million collaborators whom I promised we would make use of this data, so being able to see other astronomers take on that data and use it was good.
I think the moment I really realised that this wasn't going to be a flash in the pan was when Planet Hunters launched. I have a long track record of being wrong about many things. This was one of the things I was wrong about. I remember sending an email to the team saying that I was not sure we should be launching this project. I'm not sure that we're going to find anything. But I'd talked myself into launching it because if we fail, we'll learn something about what we can and can't do. Then pretty quickly after launch we had a couple of candidate exoplanets, and Planet Hunters is still an enormous success.
Just as a personal high point, the projects we did with Stargazing Live, where we set ourselves the challenge of producing results in two days. I don't know how it comes across on camera. Dara [O’Briain] and Brian [Cox] are great fun, and it has been good to do those things. But I didn't sleep for those two days.
What are the advantages of having citizen scientists over using artificial intelligence?
There are a few. One of them is that there are still many places where humans are better than computers. I just had a call with my PhD student Mike Walmsley, who's briefly in the book and who's our artificial-intelligence expert. We were arguing about why humans are better at finding bars in spiral galaxies than his machine. And one of the reasons is that our visual system is very good at taking in information on different scales at once.
Another answer is that people are distractible, which is sort of the theme of the book, I guess, that people can be distracted by the unusual and the unexpected. Lots of the great discoveries have come from when people have found things that they weren't expecting, or that we didn't ask them to look for. And that goes all the way back to Galaxy
Zoo, but it's still happening today. The ability to be doing one task, and then go “hang on. There’s something weird here” is very natural and very human. It's actually something that volunteers are better at than experts. Experts have a hard-coded understanding of what they should be looking at.
Do you think artificial intelligence will ever get to a point where it is up to that standard?
I think that's an open question. My flippant answer is that I’ve spent ten years with machine learning people telling me that if they had an afternoon, they could solve these problems. And it always turns out they can get 80 per cent of the way there. What we're really finding though is that because the data sets keep growing, a combination of human and machine is really powerful.
For many of the problems that we're dealing with, as the data sets get larger and larger, we definitely need machine learning to do most of the work for us. There's no way we're going to cope with the information that's coming from future telescopes without machines doing the bulk of the work. But, if you take humans and put them in the loop, not only do you get a better result by combining them, your machine can also do better. My bet is that we are going to have hybrid systems.
A telescope you mention towards the end of the book is the Large Synoptic Survey Telescope (LSST), which is going to produce ridiculous amounts of data in just a single night. Do you have plans to collaborate and use this data for citizen science projects?
The LSST is an amazing project. I've been involved with it for most of the time I've been doing Zooniverse, and it's the first observatory that's planning, from the start, to have citizen science.
For astronomers using LSST data it will be really simple for them to connect to Zooniverse and talk to our volunteers.
There is an army of postdocs around the world building routines for their special subject, so if you study Type II supernovae and you write good code, you will be able to get the LSST to give you more Type II supernovae than you've ever dreamed of. And that's cool. The problem is if you want to be surprised.
We have citizen science in the loop. We will take the weirdest stuff that the machine doesn't understand and we will pass those to people. There's some urgency, because one of the things that's happening in astronomy is we're realising things in the sky change faster than we expect.
I think we’re ready. We've got a solution. And then I think the task will be to be as flexible as possible. I haven't, for example, tried to bet on which specific science cases will need citizen science in five years time with the LSST. I'm just pretty sure there will be some.
In the book you talk about how astute you are in assessing your contributors, and especially with the Space Warps project. How much effort goes into ensuring that valuable data is assessed correctly?
It's a big effort, actually. I don't quite like thinking about the Zooniverse like this, because it sounds a little callous, but one way of testing what we're doing with the Zooniverse is to think of it as a black box.
The Zooniverse is a black box that analyses astronomical images. And if I brought that black box and handed it to my colleagues and said “this will solve all your problems,” they would spend six to nine months testing it to make sure it did what they expected and so on. We have to do the same with the Zooniverse, even though the box contains 1.6 million people. It's the same sort of tasks that we have to do when we are building a new machine learning algorithm and so on.
It's fascinating. We've looked at all sorts of things. At one point we looked at whether there was a time of day, or time of the week, that people were better at classifying. I was slightly worried that it would come out that people were great at 11pm on a Friday, after a few beers. It turned out that it was pretty flat.
The truth is that for most of our projects, most people are pretty good or become pretty good with experience. That’s useful, and we test that. Partly we're doing it to make sure that we're living up to this promise that people really are making a useful contribution.
What are your thoughts on making, for example Planet Hunters, into a game interface to try and make it more interactive?
I think that it's an interesting route. There are people who've done this, but not particularly in astronomy. But my sense is partly fear. Writing a good computer game is really hard. I'd say writing a good game that also happens to find planets is really, really, really hard, so partly we've avoided it for that. Also we found that people take part in these projects because they want to help. That was the big surprise early on. When we tried this sort of thing, we found that people changed how they talked about the project. Instead, when you don't make it a game, and specifically when you don't make it a competition, people talk about the project as something joyful and inspiring. Then the moment you make it a competition, they talk about it as if it's work. It becomes stressful. “I had to log on this morning” as opposed to “I was excited to”.
Games are really powerful. As a species, we like playing games and competitions. I try and keep the two a bit separate and make sure that people know that they're doing science. We're not seriously sitting in ranks all during our science work, but there is a seriousness of purpose there that I think is really, really important.