The Crowd and the Cosmos: Adventures in the Zooniverse

by Lintott, Chris

  fun with paper titles.)

  The invisible gorilla teaches us not only that experts make

  mistakes, but that they’re more likely to do so than the rest of us

  in some circumstances. The gorilla in question is a small cartoon

  figure, posed with one fist in the air for reasons known only to

  itself. It was placed by Drew’s team into images produced by CT

  scans of patients’ lungs, grainy black and white images studied by

  surgeons to look for signs of cancer. The participants were

  medics on the look-out for anomalies; an ideal, expert crowd for

  gorilla spotting. To make the task easier, the researchers made

  the gorilla larger, by a factor of nearly fifty, than the cancerous

  nodules the researchers were supposed to be looking for. Frankly,

  unless they’d equipped the beast with a party hat and balloons

  it’s hard to imagine how they could have made it more obvious.

  The results are shocking. Of the twenty-four experts who took

  part in the challenge, twenty of them missed the gorilla com-

  pletely (Figure 23). They didn’t see it. They didn’t mistake it for

  anything else—how could they?—but their brains just didn’t

  register something they weren’t expecting. When I first heard

  about this, I assumed they weren’t trying very hard, but eye-

  tracking equipment used in the lab showed that most of those

  who missed the simian interloper looked straight at it. Not an

  absence of effort, then, or a sloppy inspection, but an absence of

  conscious attention.

  Surprising though it is, that’s the result that the researchers

  expected. A previous result had invoked the invisible gorilla, this

  time wandering among players on a basketball court. If you

  watch the video, the figure in a party-store gorilla suit couldn’t be more obvious, but an audience told to count the number of

  passes will miss him, even as he pauses to wave to the camera

  and hence to the inattentive viewer, who remains oblivious.

  188 Serendipity

  Figure 23 The invisible gorilla (top right) as presented to surgeons for classification. It wasn’t noticed by most of the experts looking for

  tumours in these images.

  Once you know there’s a gorilla in shot, it’s literally impossible to miss him. So famous has the experiment become that for years it

  ran in cinemas as a road safety advert, preaching the need for

  careful attention. Yet not everyone is fooled to keeping their eye

  on the ball; expert basketball players are much more likely to

  notice the gorilla in their midst.

  It’s not hard to explain why experts perform better. If you’re

  more used to following the movements of an orange ball whanged

  around a court by a bunch of players wearing vests, I’m willing to

  bet you’re also looking for different things from the rest of us.

  Those who know basketball will, I reckon, be looking not at the

  ball but for people to pass to, and will therefore instantly be

  Serendipity 189

  aware of the offensive threat posed by a gorilla near the three-

  point line. Expertise here involves carrying out the simple task of

  counting passes automatically, such that it ceases to consume

  effort, freeing you to look beyond the simple task and see the

  whole game.

  That’s why the CT scan study is so surprising to me. Here, even

  experts can’t be trusted; what the study is recording is a phenom-

  enon known as inattentional bias, and it afflicts us worst when

  the object being searched for—cancer nodules in this case—is

  very different from the interloper. It’s harder to spot things the

  more different they are from the things you’re looking for. The

  gorilla being obviously not a nodule doesn’t help, but rather

  ensures that the nodule-searching brain dismisses it before the

  conscious brain can be surprised by it. Another problem is what

  people who study this stuff call ‘satisfaction of search’, the human

  tendency to stop looking once we’ve found something. Gorillas

  close to nodules were spotted slightly more often, but still missed

  more than two-thirds of the time.

  So finding what you’re not looking for turns out to be extremely

  hard, and that has consequences in the real world. Kenny Conley

  was a police officer in Boston, and at two in the morning on 25

  January 1995 he was in hot pursuit of a suspect. An undercover

  officer was also present, but when they got to the scene Conley’s

  fellow officers mistook the disguised cop for the suspect. They

  proceeded to badly beat him up, which eventually ended them in

  deep (and, I reckon, deserved) trouble.

  Conley, chasing the real suspect, had run straight past the

  place where the assault of the undercover officer was taking

  place, but claimed that he hadn’t seen anything at all. No one was

  able to believe he could have missed what was happening, and he

  was found guilty of perjury in lying to protect his fellow officers

  and sentenced to nearly three years in jail as a result.

  190 Serendipity

  This case attracted attention, and inspired an experiment.

  Participants were asked to run after someone, passing on the

  way a group of brawling actors. At night, only a third of them

  noticed the fight, and even in broad daylight a third missed

  noticing that anything was happening. The act of concentrating

  on chasing someone decreases the attention one pays to the sur-

  rounding world. As the researchers put it in the title of their

  paper, ‘You do not talk about Fight Club if you do not notice

  Fight Club’. (Psychologists really, really have more fun with their

  paper titles than astronomers do.)

  Conley eventually won an appeal, though not because of the

  research described above. Nonetheless, once you start thinking

  like this it becomes obvious why Hanny, and not a bunch of

  astronomers, found the Voorwerp. Expert—and especially pro-

  fessional—classifiers know what a galaxy looks like, and so aren’t

  likely to be distracted by the appearance of something else.

  Newer volunteers, those with less knowledge, are likely to be

  conscious of all sorts of things in the data, some interesting and

  novel and some not.

  Hanny was also an effective advocate for her discovery. It was

  her Voorwerp, after all, and she wanted to know what it was. Her

  desire to understand pushed us on the Galaxy Zoo team to look

  into the matter, but it wasn’t easy at first. A leading hypothesis at first was that it might have been an interloper, a nebula or cloud

  of gas belonging to our own Milky Way. In order to test this idea,

  we needed a spectrum of the object, but the Sloan Digital Sky

  Survey that provided the image in which it was found hadn’t

  targeted it for spectrographic follow-up. The survey’s algorithms

  just hadn’t anticipated anyone finding it interesting. Worse, the

  thing was faint enough that we needed a large telescope to get

  enough light, but time on those is won by writing convincing

  pitches describing future scientific bounty that will inevitably

  Serendipity 191

  flow from a particular set of ob
servations, not from following

  some wild goose chase inspired by a single image of a weird

  object.

  If it were up to me, we’d be able to write ‘We found an unusual

  thing and want to look at it’ and send that off to the Time

  Allocation Committee (TAC). A TAC is not some sort of commit-

  tee of advanced aliens, something from the Doctor Who cutting

  room, but rather the group convened by an observatory that

  decides who gets time using the telescopes. Faced with more

  astronomers with more ideas than anyone should have to deal

  with, they tend to look askance at speculative proposals.

  Luckily, astronomy is a small world. I found out that Matt

  Jarvis—both then and now again a colleague in Oxford—

  happened to be observing for his own purposes at a telescope

  in the Canary Islands. With a little nudge, Matt pointed the tele-

  scope in the right direction, and emailed back a spectrum.

  Whether Matt had sacrificed his own observing time, or ‘acci-

  dentally’ pointed the telescope at the Voorwerp while setting up

  for the night, I’ve never dared ask.

  There’s always been a somewhat informal barter economy

  around telescope time; emails soliciting objects that would be

  worth observing after primary targets were set, or phone calls

  requesting emergency—or risky—observations, used to be

  common. As scheduling has become automated and efficient,

  these loopholes are closing. I worry about how we’re going to

  take risks, and think most observatories should set aside a small

  amount of time for observations which might be a bit unusual

  but which might pay off spectacularly.

  Anyway, we got our data and the spectrum was a revelation.

  Even a quick glance at it told me that the Voorwerp—whatever it

  might be—was at more or less the same distance as the neigh-

  bouring galaxy. It was therefore huge—almost galaxy-sized itself.

  192 Serendipity

  Looking back, I think that was the first moment I realized this

  was more than a curiosity; that the Voorwerp wasn’t just some-

  thing that had caught Hanny’s eye, but was genuinely interesting

  in itself.

  It clearly needed more than a casual glance. Luckily, I was sitting

  in the conference centre in the middle of Austin, Texas, at the win-

  ter meeting of the American Astronomical Society—the largest

  annual gathering of astronomers. Among them was Alabama

  professor and Galaxy Zoo observing guru Bill Keel, who quite

  literally wrote the book on how to study galaxies. Dealing calmly

  with me waving a laptop in his face, Bill immediately noticed what

  I had not; there were features in the spectrum which suggested

  the presence of elements such as sulfur,* in conditions which led

  their atoms to be highly excited. In other words, the gas in the

  Voorwerp was hot. Very hot. About 50,000 degrees Celsius in

  fact, or nearly ten times the temperature of the surface of the Sun.

  That’s not unprecedented, but it does need explanation, all

  the more so because there was nothing in the spectrum which

  suggested the presence of stars embedded in the gas. If they’d been

  there, they would have contributed what is called continuum

  light, shining at all wavelengths, but the absence of a significant

  continuum meant that very few bright stars could be present.

  There certainly weren’t enough to heat the rest of the gas. Sitting

  in the corner of a corridor in a large and almost completely

  soulless convention centre, Bill and I realized the Voorwerp was

  a real mystery. Understanding why this blob of gas was excited,

  and identifying the source of its excitement, was a proper

  scientific question, and the spectrum Matt and colleagues

  * I’m using the American ‘sulfur’ not the English ‘sulphur’ because that’s what the International Union of Pure and Applied Chemistry says we should do. They adopted English spellings of aluminium and caesium, so it’s not as if the Americans got everything their own way.

  Serendipity 193

  provided was not the end but the beginning of a scientific detective

  story. With an unusual spectrum in hand, we had the ammuni-

  tion to go and chase down more clues.

  First, though, there was a chance for some good old-fashioned

  speculation as we tried to work out what sort of thing the

  Voorwerp might be. One obvious possibility was that it was the

  remnant of a supernova; many of the explosions discussed earlier

  in the book will produce not only a dense remnant—the neutron

  star or black hole that forms from the star’s core—but also a sur-

  rounding cloud of gas. These supernova remnants don’t last for

  long—we’re watching the one produced by the 1987 explosion in

  the Large Magellanic Cloud change before our eyes—but they do

  shine brightly due to gas excited by the explosion. The shock wave

  from such an explosion might, if powerful enough, excite sur-

  rounding gas to the degree seen in the Voorwerp. A careful look

  at the Voorwerp itself supported this nascent theory; there’s a

  roughly circular ‘hole’ in the gas which would easily be explained

  if it was centred on the site of an explosion, with the gas closest to the action having been destroyed or ejected completely.

  It’s a simple, neat explanation of the object, which is utterly

  confounded by the facts. The biggest and most immediate prob-

  lem is the sheer size of the Voorwerp itself, much, much larger

  than any supernova remnant in the Milky Way. Any explosion

  capable of exciting gas over such a large volume of space must

  have been quite something, but in the early speculative phase of

  thinking about things we weren’t too discouraged, given free-

  dom to imagine the unlikely because the Voorwerp was, as far as

  we knew, one of a kind.

  The discovery of many such objects would mean making a

  claim about how frequent supernovae capable of producing such

  massive remnants are, a calculation that could be quickly tested

  by observation. With only one example, who’s to say we hadn’t

  194 Serendipity

  stumbled across the remains of the most super of supernovae?

  The next stage though, is to calculate, or at least guess, how long

  the unique thing you’re observing will survive in something like

  its current state. If it’s short lived—a nova that will come and go

  in a matter of weeks or a planetary nebula which will last for only

  a few tens of thousands of years—then an argument which relies

  on scarcity is dead in the water; you may only see one example

  now, but another will be along in a little while. If it’s long-lived—

  and the sheer size of the Voorwerp, closer to the scale of a dwarf

  galaxy than any normal remnant, suggested it wasn’t going any-

  where in a hurry—then the argument that you might be dealing

  with an exceptional specimen has more weight.

  I’m labouring the point perhaps, but this sort of argument lies

  right at the heart of the kind of astronomy I like to do, and what,

  when whiling away nights back in the school observatory,

  I thought astronomers mostly did. We
found a weird thing.

  Great! How weird is it? Is it especially close, or far away? How

  bright does it look? Is it changing? How does it compare to other

  things? These are all simple observations, but they’re as much

  part of attempting to understand the Voorwerp as writing down

  equations that convert features observed in a spectrum to physical

  properties like temperature.

  In this case, the line of reasoning suggested that any Voorwerp-

  producing supernova would have to be exceptional, and there-

  fore exciting. Before we could go searching for the dense remnant,

  the neutron star or black hole that would confirm that a giant

  explosion occurred here, we realized a clue had been overlooked.

  The Voorwerp is large enough that we don’t need to treat it all as

  a single object—we can look at parts of it separately, even from

  our distance of three hundred million light years.

  Once we realized that, a pattern became clear. How excited

  the gas was depended on how far it was from the neighbouring

  Serendipity 195

  galaxy, with the gas closest to the galaxy less excited than that

  which is further away. A small clue, but an important one. If this

  was a hard-boiled detective movie, imagine the camera panning

  slowly to the galaxy, ominously hanging ‘above’ the Voorwerp

  in each of the images we’d spent ages staring at.

  Could it have been the culprit? Like most, and probably all,

  large galaxies, it contains a supermassive black hole at its centre.

  As material falls into such a black hole, if there is enough of it, it can form a disc of material orbiting the central black hole,

  known as an accretion disc. The physics of how material behaves

  in such circumstances is, well, complicated to say the least,

  but it’s clear that such systems can produce powerful jets of

  material moving at high speed. Such jets are common in massive

  elliptical galaxies—M87, the giant at the heart of the nearby

  Virgo Cluster, has a well-studied example which moves at very

  nearly the speed of light—but they have also been found in spirals.

  Volunteers in the Radio Galaxy Zoo project, which tries to pair

  galaxies observed in the radio with their counterparts in the

  infrared, have found just such an object, and in that case as in

  almost all spiral systems with jets, the jet was perpendicular to