AP Computer Science Principles · Unit 5 lecture · Part 5 of 11 · 7 min

5.4 CrowdsourcingMany people, one problem

Mr. Thompkins
Here's some good news for a change. The same connectedness that causes headaches also lets millions of people work on one problem together, including people with no science degree making real discoveries.

Widespread access to information and public data makes it easier to identify problems, develop solutions and share results. Digital collaboration enhances human capabilities: people can work together across distance and time in ways that were impossible before.

Crowdsourcing is the practice of obtaining input or information from a large number of people via the internet. Some online services use the contributions of many people to benefit both individuals and society: navigation apps learn about traffic jams from drivers' phones, online encyclopedias are written by volunteers, and review sites pool the experiences of thousands of customers. Crowdsourcing also offers new models for collaboration, such as crowdfunding, which connects a project or a cause with many small donors.

?Wait, isn't that the same as distributed computing from Unit 4?
No. In distributed computing, many computers each run part of a program. In crowdsourcing, many people contribute information, ideas or work. The question to ask: are people contributing, or are computers computing?

Science has been changed by citizen science. Citizen science is scientific research conducted in whole or in part by distributed individuals, many of whom may not be scientists, who contribute relevant data to research using their own computing devices. A phone has a camera, a GPS and an internet connection, so millions of people can be field researchers. In projects like Galaxy Zoo, volunteers sorted photos of galaxies by shape faster than astronomers ever could alone.

volunteersphonesdatabasescientistssee a birdphoto + GPS + timemillions of recordsmap migrationresults shared back with the publiccitizen science: data from the public, for research
Citizen science: volunteers collect data on their own devices, scientists analyze the combined records, and the results come back to the public.
IdeaWho contributes whatExample
CrowdsourcingMany people online give input, ideas or workDrivers' phones reporting traffic; an online encyclopedia written by volunteers
Citizen scienceMembers of the public, many not scientists, collect data for research with their own devicesVolunteers sorting galaxy photos by shape; logging bird sightings
CrowdfundingMany people each give a little moneyHundreds of small donations pay for a community garden
Distributed computing (Unit 4)Many computers each run part of a program; no people contribute dataServers splitting a huge simulation
Four ideas that sound alike. Watch who contributes what.
?Wait, how do scientists trust data from non-experts?
They design for it. Each item can be checked by several volunteers, and the answer most of them agree on is kept. Odd records are flagged for an expert. Photos with GPS and time metadata, a Unit 2 idea, help confirm when and where an observation was made.
Pause and think
A thousand volunteers count birds in their backyards on one weekend. Predict one strength and one weakness of this data compared with one expert counting birds in one park.
I thought about it. Show me
Strength: far more locations at once, so scientists can see patterns across a whole region. Weakness: volunteers vary in skill, so some birds are misidentified, and backyards aren't a random sample of the landscape. Checking by several people and flagging odd records helps.
Analyze a citizen science design
A wildlife project has 12,000 trail-camera photos. Each photo is shown to 3 different volunteers, who each label the animal. The label at least 2 of the 3 agree on is kept; if all 3 disagree, the photo goes to an expert. (a) How many labels are needed in total? (b) If 600 volunteers share the work equally, how many labels does each one make? (c) Explain why showing each photo to 3 people makes the data more reliable.
  1. (a) 12,000 photos × 3 labels each = 36,000 labels.
  2. (b) 36,000 ÷ 600 = 60 labels per volunteer, an hour or so of work each.
    Why?
    That's the power of crowdsourcing: a job that would take one expert weeks becomes an evening for each of 600 people.
  3. (c) One volunteer's mistake is outvoted by the other two. For a wrong label to be kept, two people have to make the same mistake on the same photo, which is much less likely.
  4. Photos where nobody agrees are probably the hard ones, and those are exactly the ones sent to an expert.
    Why?
    This is redundancy again, the Unit 4 idea, applied to people instead of network paths.

Answer: (a) 36,000 labels. (b) 60 each. (c) Majority agreement outvotes individual mistakes, and the hardest photos go to an expert.

Stuck? Get a hint
  1. Each photo needs 3 labels. How many photos are there?
  2. Divide the total labels evenly among 600 people.
  3. What has to happen for a wrong label to survive a 2-out-of-3 vote?
Check yourself
Thousands of hikers use a phone app to photograph wildflowers and upload each photo with its location and date, so botanists can track when flowers bloom. This is best described as
  1. Adistributed computing
  2. Bcitizen science
  3. Copen source software
  4. Da Creative Commons license
Show the answer

B. Members of the public, mostly not scientists, are contributing data to research with their own devices: citizen science. It is not distributed computing, because people are collecting observations, not computers running parts of a program.

  • Digital collaboration and public data make it easier to find problems, build solutions and share results.
  • Crowdsourcing: getting input or information from many people via the internet. Crowdfunding is one new model.
  • Citizen science: research done in part by distributed individuals, many not scientists, contributing data with their own devices.