Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Crowdsourcing PWI Sept-2011


Published on

Presentation of the crowdsourcing business model to the Professional Women International association. It describes the pros and cons, how to scale with Machine Learning, and the emergence of reputation systems.

Published in: Business, Technology
  • Be the first to comment

Crowdsourcing PWI Sept-2011

  1. 1. • Introduction• Crowd Motivation• Client Motivations and Types of tasks• Scale up with Machine Learning• Quality Management• Workflows for Complex tasks• Reputation Systems• Economic shift PWI - September 29, 2011
  2. 2. Crowdsourcing Crowd or Community (online audience)1 2 3 4
  3. 3. Ex: “Adult Websites” Classification• Large number of sites to label• Get people to look at sites and classify them as: –G (general audience) – PG (parental guidance) –R (restricted) –X (porn) [Panos Ipeirotis. WWW2011 tutorial]
  4. 4. Ex: “Adult Websites” Classification• Large number of hand‐labeled sites• Get people to look at sites and classify them as: –G (general audience) – PG (parental guidance) –R (restricted) –X (porn)Cost/Speed Statistics:• Undergrad intern: 200 websites/hr, cost: $15/hr• MTurk: 2500 websites/hr, cost: $12/hr [Panos Ipeirotis. WWW2011 tutorial]
  5. 5. Crowd Motivation• €,$ = Money!• Self-serving purpose (learning new skills, get recognition, avoid boredom, enjoyment, create a network with other profesionals)• Socializing, feeling of belonging to a community, friendship• Altruism (public good, help others)
  6. 6. Examples: Altruism
  7. 7. Crowd Demography (background defines motivation)• The 2008 survey at iStockphoto indicates that the crowd is quite homogenous and elite.• Amazon’s Mechanical Turk workers come mainly from 2 countries: a) USA b) India
  8. 8. Crowd Demography
  9. 9. Client motivation• Need Suppliers: Mass work, Distributed work, or just tedious work  Creative work  Look for specific talent  Testing  Support  To offload peak demands  Tackle problems that need specific communities or human variety  Any work that can be done cheaper this way.
  10. 10. Client motivation• Need customers!• Need Funding• Need to be Backed up• Crowdsourcing is your business!
  11. 11. Examples of Funding
  12. 12. Client Tasks Goals3 main goals for a task to be done:1. Minimize Cost (cheap)2. Minimize Completion Time (fast)3. Maximize Quality (good) Remember Crowd Motivation! (ex.: Game-ify your task, explain the final purpose)
  13. 13. Examples: Games
  14. 14.[Panos Ipeirotis. WWW2011 tutorial]
  15. 15. Pros• Quicker: Parallellism reduces time• Cheap• Creativity, Innovation• Quality (*depends)• Access to scarce resources: The ‘long tail’• Multiple feedback• Allows to create a community (followers)• Business Agility• Scales up! (*up to a level)
  16. 16. Cons• Lack of professionalism: Unverified quality• Too many answers• No standards• Not always cheap: Added costs to bring aproject to conclusion• Too few participants if task or pay is notattractive• If worker is not motivated, lower quality of work
  17. 17. Scale Up with Machine Learning Build an ‘Adult Website’ Classifier• Crowdsourcing is cheap but not free - Workers cannot do more than xxhours/day, Cannot scale to web without helpBuild automatic classification models using examples from crowdsourced data
  18. 18. Integration with Machine Learning• Humans label training data• Use training data to build model
  19. 19. Quality Management Ex: “Adult Website” Classification• Bad news: Spammers!• Worker ATAMRO447HWJQ labeled X (porn) sites as G (general audience)[Panos Ipeirotis. WWW2011 tutorial]
  20. 20. Quality Management Majority Voting and Label Quality• Spammers try to go undetected• Good willing workers may have bias  difficult to set apart.1. Ask multiple labelers2. Keep majority label as “true” labelUse the probability ofbeing correct as theQuality Indicator
  21. 21. Complex tasks Handle answers through workflow• Q: “My task does not have discrete answers….”• A: Break into two Human Intelligence Tasks (HITs): – “Create” HIT – “Vote” HITVote controls quality of Creation HIT• Redundancy controls quality of Voting HIT
  22. 22. Collaboration: Photo description But the free-form answer can be more complex, not just right or wrong…TurkIt toolkit [Little et al., UIST 2010]:
  23. 23. Collaboration: Description Versions1. A partial view of a pocket calculator together with some coins and a pen.2. ...3. A close‐up photograph of the following items: A CASIO multi‐function calculator. A ball point pen, uncapped. Various coins, apparently European, both copper and gold. Seems to be a theme illustration for a brochure or document cover treating finance, probably personal finance.4. …8. A close‐up photograph of the following items: A CASIO multi‐function, solar powered scientific calculator. A blue ball point pen with a blue rubber grip and the tip extended. Six British coins; two of £1value, three of 20p value and one of 1p value. Seems to be a theme illustration for a brochure or document cover treating finance ‐ probably personal finance.
  24. 24. Collaboration• Exploration / exploitation tradeoff (Independence/or not)– Can accelerate learning, by sharing good solutions– But can lead to premature convergence on suboptimal solution[Mason and Watts, submitted to Science, 2011]
  25. 25. Collaboration: Positive• Building iteratively allows better outcomes for the image description task.• In the FoldIt puzzles, workers built on each other’s results. They recently found in 10 days the molecular structure of a protein- cutting enzyme from an AIDS-like virus.
  26. 26. Collaboration: Negative Group Thinking Effect• Individual search strategies affect group success: Players copying each other make less exploring  lower probability of finding peak on a round
  27. 27. Workflow Patterns• Generate / Create• Find• Improve / Edit / Fix  Creation• Vote for accept‐reject• Vote up, vote down, to generate rank• Vote for best / select top‐k  Quality Control• Split task• Aggregate Flow Control• Iterate  Flow Control
  28. 28. AdSafe Crowdsourcing Experience
  29. 29.
  30. 30. AdSafe Crowdsourcing Experience•Detect pages that discuss swine flu– Pharmaceutical firm had drug “treating” (off-label) swine flu– FDA prohibited pharmaceuticals to display drug ad inpages about swine flu  Two days to comply!• Big fast-food chain does not want ad to appear:– In pages that discuss the brand (99% negative sentiment)– In pages discussing obesity
  31. 31. Adsafe Crowdsourcing Experience Workflow to classify URLs• Find URLs for a given topic (hate speech, gambling, alcoholabuse, guns, bombs, celebrity gossip, etc etc)http://url‐• Classify URLs into appropriate categorieshttp://url‐• Mesure quality of the labelers and remove spammers• Get humans to “beat” the classifier by providing cases wherethe classifier failshttp://adsafe‐
  32. 32. Crowdsourcing AggregatorsAct as Portals• Create a crowd or community.• Create a site to connect a client to the crowd• Deal with workflow of complex tasks, likedecomposition into simpler tasks and answerrecomposition• Works as Broker and Bank, Mediator Allow anonymity Consumers can benefit from a crowd withoutthe need to create it.
  33. 33. Market Design:Crude vs Intelligent Crowdsourcing• Intelligent Crowdsourcing uses an organized workflow to tackle CONS of crude crowdsourcing. Complex task is divided by experts, Given to relevant crowds, and not to everyoneIndividual answers are recomposed by experts into general answer
  34. 34. Lack of Reputation and Market for Lemons“When quality of sold good is uncertain and hidden before transaction, prize goes to value of lowest valued good” [Akerlof, 1970; Nobel prize winner]• Market evolution steps: 1. Employers pays $10 to good worker, $0.1 to bad worker 2. 50% good workers, 50% bad; indistinguishable from each other 3. Employer offers price in the middle: $5 4. Some good workers leave the market (pay too low) 5. Employer revised prices downwards as % of bad increased 6. More good workers leave the market… death spiral
  35. 35. Reputation systems• Challenges: - Insufficient participation - Overwhelmingly positive feedback + Hoping to get a positive ranking in return - Negative feedback avoided for fear of retaliation - Dishonest reports + « Riddle for a PENNY! No shipping-Positive Feedback » - « Bad-mouth » reports• Incentive mechanisms to get honest feedback - pay rater if report matches next; - delay next transaction over time
  36. 36. Reputation systems• “Cheap pseudonyms”: easy to disappear and reregister under a new identity with almost no cost. [Friedman and Resnick 2001] Introduce opportunities to misbehave without paying reputational consequences.Increase the difficulty of online identity changes Impose upfront costs to new entrants: allow new identities (forget the past) but make it costly.• 2-sided Reputation Mechanisms – Crowd: To ensure worker quality – Employer: To ensure their trustworthiness
  37. 37. Economical Shift• From Social Networking to Social Production through Collaborative Innovation  Mass-Collaboration changes how Products & Services are Designed,Manufactured,Marketed• Classical geo-political and economical organisations do not correspond to new economy  Realignment of competitive advantages  Move towards Collaborative Enterprises based on Open Infrastructure
  38. 38. Societal Shift Moral values Reinforcement• Open data access makes actions Transparent• Transparency makes people Accountable• Accountability forces/fosters Integrity• Integrity breeds Community Support Link between Ethical values and ROI
  39. 39. References• Wikipedia,2011• Dion Hinchcliffe Crowdsourcing: 5 Reasons Its Not Just For Start UpsAnymore,2009• Tomoko A. Hosaka, MSNBC. "Facebook asks users to translate forfree“,2008.• Daren C. Brabham. "Moving the Crowd at iStockphoto: The Composition ofthe Crowd and Motivations for Participation in a Crowdsourcing Application",First Monday, 13(6),2008.• Karim R. Lakhani, Lars Bo Jeppesen, Peter A. Lohse & Jill A. Panetta. Thevalue of openness in scientific problem solving (Harvard Business SchoolWorking Paper No. 07-050),2007.• Klaus-Peter Speidel How to Do Intelligent Crowdsourcing,2011• Panos Ipeirotis. Managing Crowdsourced Human Computation,WWW2011 tutorial,2011• Omar Alonso & Matthew Lease. Crowdsourcing 101: Putting the WSDM ofCrowds to Work for You, WSDM Hong Kong 2011.• Sanjoy Dasgupta,,2009•Don Tapscott, Anthony Williams. Macrowikinomics, 2010.
  40. 40. Call For Ideas: If you have a large set of examples or just an idea of application for a program to classify or predict, I would love to hear from you!Questions? PWI - September 29, 2011