Are WE really the experiment? Is there more going on here than is being represented?
-
by
ghansenut
On both this site and the Andromeda Project site, I often wonder if some or all of the experiment has nothing to do with the stated scientific objective. Am I just a cynic or does anyone else get the feeling that there's more going on here than meets the eye? How do we know that WE aren't the ones actually being studied?
Posted
-
by
craigpacker
scientist
Don't worry the science here is legit -- we are wildlife biologists who are interested in the community ecology of the Serengeti -- check out our project website at www.lionresearch.org -- and stay tuned in the future. We will let everyone know all about the patterns that emerge from the data -- how all these different species manage to live together!
Posted
-
by
chalyse
Its nice to see the University of Minnesota's Lion Research group join the Zooniverse family of projects, and even the parent group's privacy statement confirms that no identifiable user data is ever shared. Like any legitimate research science, anonymized and aggregate user interface data may be used to improve all aspects of the experimental models, including usage trends, extrapolations to improve community board use and experience, increased retention of users, etc. Here's the privacy statement of Zooniverse:
https://www.zooniverse.org/privacy?lang=en
exerpt:
"The Zooniverse is a suite of citizen science projects operated by the Citizen Science Alliance (CSA), which support scientific research by involving members of the public - you - in the processes of analyzing and discussing data.. Data from these projects is [also] used to study online community design and theory, interface design, and other topics..."And, here's a nice example of a paper that, in part, explored user-interface data from another Zooniverse project, outlining improvements to user experience, retention, development of specialized expertise, etc.:
http://citsci.syr.edu/system/files/galaxyzoo.pdf
.... I'd be disappointed if the larger experimental objectives did not include maximizing user benefits and improvements to user communities beyond the basic raw data we generate ... it makes each new project all the more enjoyable for those who relish being "armchair interns" with the world's greatest scientists and their fascinating projects :} "no bugs, no heat stroke, no malaria" ... doesn't get any better than that!
Posted
-
by
ahjorth
I came upon this page since I love Zooniverse's Planet Hunters. Zooniverse has a whole bunch of projects where people are better than computers, like:
Analyzing cells for cancer (https://www.zooniverse.org/project/cellslider)
Analyzing kepler date from stars to find planets! (https://www.zooniverse.org/project/planethunters)
Decipher Whale sounds! (https://www.zooniverse.org/project/whalefm)
And many more: https://www.zooniverse.org/projects
Posted
-
by
ghansenut
Don't get me wrong, I really applaud the work and I enjoy being involved. Let me just ask the question that's on my mind: Do the researchers insert confirmed animal photos in the mix in order to "reward" those who are analyzing the photos in order to keep people "in the game"?
For example: In the Andromeda Project they insert "synthetic star clusters" that are "found" when properly identified. I believe that this has two objectives: The stated objective of helping train those involved to identify and mark the appropriate objects. But I think the unstated objective is also to reward the participants thereby motivating them to continue to the next photo for a possible future "reward" (similar to a video game)
So the question is this: Are all the photos "real" in the sense that they have not previously been analyzed, or are some of them confirmed animal photos planted by the researchers?
Posted
-
by
kosmala
scientist
ghansenut: do we insert confirmed photos into the mix to reward people? Short answer: no. We thought about giving known easy animals to each person when they were just starting out, so they could learn the animals a bit at a time -- ease the learning curve, if you will. But, it's too complex to do that, and plenty of people start out not logged in. So no.
Have any photos been confirmed? Short answer: yes. Long answer: The camera traps were initially set out to answer Ali's questions about how carnivores interact and coexist. And initially we thought we'd just look at all the images ourselves. So Ali set up a website to allow undergraduate volunteers to do exactly what you're doing on Snapshot Serengeti. Except the interface was not as cool -- images didn't automate, you had to use a pull-down menu to choose animals, and no filters; there was a PDF guide and a pre-test you had to take. Each image also got seen exactly once. The undergrads worked through Seasons 1 and 2 and part of 3 (over two years!) before Snapshot Serengeti went live. So each image had one classification and there was no way for us to know automatically if there were mistakes, but the classifiers were reasonably well-trained. Does that make those image "confirmed"? I don't know. Ali herself looked at all the images that undergrads classified as lion, leopard, cheetah, and hyena from Seasons 1 and 2. So those ones are definitely "confirmed." We could have left them out of Snapshot Serengeti, but we figured you'd prefer to see them! And there's no clever program deciding when to give them out to people; it's all random.
Posted
-
by
ghansenut
Thank you for that explanation. [/Conspiracy Theory Off]
Posted
-
by
zookeeper
admin
A couple of points to add :
We take huge pride in the Zooniverse in ensuring that all of our projects produce real science in the end - you can see our track record for yourself at zooniverse.org/publications - and we're pretty hard on scientists who approach us to make sure that they will fulfill this condition.
We are also interested in studying how people interact with these sites, so that we can make better ones. In particular, we're interested in ensuring we're not wasting people's time (and so in how to retire things - we currently pull anything which has a consensus classification even though those are probably often the cleanest shots) but also in increasing the efficiency of how we get through data. It doesn't look like we need to be much more efficient than we are here, but we're considering - for example - a project with researchers who have 100 million (!) images of plankton that need classifying. We're going to have to be very careful with what we ask people to classify to get through that lot.
It's actually a fascinating study in itself, because it's not obvious what to do. Imagine we could detect who was best at each kind of animal, and could direct the relevant images your way once they'd be classified by most people. After an initial classification, giraffe experts would get only giraffes, serval experts servals and so on. Now imagine (which might be true) that the variety is a large part of why SnapshotSerengeti is fun. We've just created a system which will systematically drive away our best classifiers, and we need to go back to the drawing board. We hope that by trying many different projects we can begin to understand how to navigate these complex waters and make better projects.
We will always tell you when we're explicitly testing people with known data or as in the Andromeda Project mentioned above, synthetic images. It's funny that that should come up as an example of positive reinforcement - we were deeply worried that it would put people off (based on experiences elsewhere!).
Hope that helps clear things up - thanks for your time!
Chris
Posted
-
by
Veegee
hmmm, everything is a conspiracy nowdays! Not sure how I feel about the synthetic star clusters in Andromeda, almost feel it might be better if it told you "missed one" instead lol
Sure have enjoyed this Serengeti project tho, wish it had lasted longer, but it was so fun, we all just dug in I think. I found most of the animals listed, just a few managed to escape my photos, finally found a porcupine last nite, so that was excitable!Posted
-
by
LoraO
moderator
We are very glad you have enjoyed it!
Posted
-
@ghansenut there will definitely be an aspect of this http://en.wikipedia.org/wiki/Inter-rater_reliability in the experiment.
Posted
-
by
Robcheerio
Don't worry everyone, there is cake at the end!
Posted
-
I'm quite happy to know I can leave the statistics and number-crunching to someone else, and just look at a lot of animals, thank you!
Posted
